Wikipedia:Arbitration/Requests/Case/Magioladitis/Workshop

Main case page (Talk) — Evidence (Talk) — Workshop (Talk) — Proposed decision (Talk)

Case clerk: Amortias (Talk) Drafting arbitrator: Opabinia regalis (Talk)

Purpose of the workshop: The case Workshop exists so that parties to the case, other interested members of the community, and members of the Arbitration Committee can post possible components of the final decision for review and comment by others. Components proposed here may be general principles of site policy and procedure, findings of fact about the dispute, remedies to resolve the dispute, and arrangements for remedy enforcement. These are the four types of proposals that can be included in committee final decisions. There are also sections for analysis of /Evidence, and for general discussion of the case. Any user may edit this workshop page; please sign all posts and proposals. Arbitrators will place components they wish to propose be adopted into the final decision on the /Proposed decision page. Only Arbitrators and clerks may edit that page, for voting, clarification as well as implementation purposes.

Behaviour on this page: Arbitration case pages exist to assist the Arbitration Committee in arriving at fair, well-informed decisions. You are required to act with appropriate decorum during this case. While grievances must often be aired during a case, you are expected to air them without being rude or hostile, and to respond calmly to allegations against you. Accusations of misbehaviour posted in this case must be proven with clear evidence (and otherwise not made at all). Editors who conduct themselves inappropriately during a case may be sanctioned by an arbitrator or clerk, without further warning, by being banned from further participation in the case, or being blocked altogether. Behavior during a case may be considered by the committee in arriving at a final decision.

Motions and requests by the parties

Template

1)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

3)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposed temporary injunctions

Template

1)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

3)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

4)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Questions to the parties

Arbitrators may ask questions of the parties in this section.

The things I am looking for going towards this proposed decision:

1. Is Magioladitis in violation of policies on use of the administrative tools and how should it be remedied?
2. Is Magioladitis in violation of a block and/or the bot policy by editing from their main account using the same methods their bot would?
3. Is Magioladitis editing from their main account using the same methods their bot would?
4. Is the Bot policy sufficiently clear about what cosmetic edits are? If not has the community been sufficiently clear to Magioladitis?
5. Does Magioladitis understand what cosmetic edits are or have they indicated understanding in the past?
6. Are there untapped enforcement venues or methods that could deescalate the dispute now or in the future?
7. Is Magioladitis' edits in dealing with orphaned templates consistent or against policy?
8. Is Magioladitis' working with the community properly to resolve issues that come up?

If you wish to answer these, please include this in your evidence or in your proposals below. -- Amanda (aka DQ) 01:49, 8 January 2017 (UTC)

• @Stevietheman: Thank you for your very detailed background in evidence. Questiosn for you and @Magioladitis::

1. So is the assertion that AWB itself is causing the errors to show vs. any negligence by the bot operator?
2. How close is the gap between the bug fix and software release to the public?
3. Is there a beta branch that would serve Magioladitis better in these fixes?
4. Is it practical to stop Yobot until the changes are deployed and updated with the software?

1. Sub question: Are these bugs being fixed in a timely manner at all?

-- Amanda (aka DQ) 02:09, 8 January 2017 (UTC)

Rich Farmbrough, help me out please--I have a few questions after reading over your evidence.

1. Are you suggesting none of the Yobot blocks were necessary? Is this an opinion shared widely? Because the first response to that would be to think that this practically pulls the rug out from under the case.
2. "Given that a bot is blocked for malfunctioning, rather than malfeasance, it has been normal for blocking admins to say or imply that the bot may be resumed once the issue is resolved." Are we talking about malfunctioning rather than malfeasance? Is that the case for all blocks of Yobot?
3. "It is clear that in all cases Magioladitis unblocked because the issue was resolved." I do not know this to be clear. In fact, I am hoping that one or more of the participants here will provide some discussion/context of/for the blocks and unblocks. Thanks, Drmies (talk) 18:12, 12 January 2017 (UTC)

Although I am not a party, I have posted a limited amount of evidence which I think is related to these questions. — Carl (CBM · talk) 19:23, 15 January 2017 (UTC)

Proposed final decision

Proposals by User:Ramaksoud2000

Proposed findings of fact

Yobot unblocks

1) Magioladitis has unblocked his bot, Yobot, on numerous occasions without the consent of the blocking administrator or the community.

Comment by Arbitrators:

The arbitrators' vote comments in the Rich Farmbrough case five years ago reflect some disagreement at the time as to whether and when a bot operator may unblock his or her own bot. See discussion of proposed principles 7 through 10 here. Have the policies or guidelines surrounding this issue been clarified since that time? Should they be? Or is this a case-by-case situation where too many rules and regulations would be counterproductive? Newyorkbrad (talk) 01:37, 12 January 2017 (UTC)

I believe that for the purposes of this case, there is sufficient clarity. The Misuse of Administrative tools FoF in that case stated, "using the tools to reverse the actions of other administrators, such as unblocking a bot which is believed to be violating bot policy, should not done without good cause, careful thought and usually some kind of consultation." SilkTork's support for that FoF included: "Making clear it is not the unblocking of repaired bots that is the issue, but the occasions when the admin tools were used to unblock in clear disregard of policy and accepted standards". As entered into evidence, there was usually no discussion, and the bot was never "repaired". In addition, Magioladitis has already recognized that unblocking his bot is not an "accepted standard". See [1] where he says "OK but you also have to read WP:BUREAUCRACY", in response to Xeno's "Please do not unilaterally unblock your own bot again". When HJ Mitchell requested that he not unblock his own bot, Magioladitis did not dispute it.

In response to your broader question, I believe that it is a case-by-case issue. I may have missed it, but I don't see an epidemic of bot owners unblocking their own problematic bots. As SilkTork stated, uncontroversial unblocks of repaired bots shouldn't be an issue. A blanket prohibition may have saved some evidence-gathering time in this case, but I don't find it necessary. The reason that people took issues with Magioladitis' unblocks wasn't just blind rule-following. It was repeated issues with the bot after these unblocks, and a refusal to stop unblocking after a problem had been identified. Ramaksoud2000 ^{(Talk to me)} 02:02, 12 January 2017 (UTC)

At the risk of stating the obvious, the operator of a bot is in a technical position to determine that problems with that bot have been fixed. The blocking administrator may not be, and certainly "the community" at large is not. Opabinia regalis (talk) 05:43, 16 January 2017 (UTC)

Opabinia regalis, normally, yes. But if that were the case here, we wouldn't be having this conversation now. A problem with his unblocking was identified early on, yet it continued. Ramaksoud2000 ^{(Talk to me)} 07:50, 16 January 2017 (UTC)

What problem? All the best: Rich Farmbrough, 22:50, 16 January 2017 (UTC).

Comment by parties:

I fail to see how is this supported by the evidence. It is clear that the blocking admin was aware of my action and it was done in communication with them. -- Magioladitis (talk) 00:16, 17 January 2017 (UTC)

"without the consent" is not even true. See for example that the first unblock was after contact with the blocking admin User_talk:Magioladitis/Archive_2#Yobot. (Compare the times + see how fast I responded to an error that was not even related to the "cosmetic" problem + see their response). See also that the second block was mainly that the blocking admin makes an alternative suggestion User_talk:Magioladitis/Archive_5#Empty_sections_and_See_also_.2F_Further_reading related to the task of tagging empty sections. See that the unblock was in cooperation with the blocking admin, unrelated to "cosmetic" changes and focused in a single task. The third unblock discussion: User_talk:Yobot/Archive_2#Taxobox_.28and_other_infoboxes.29_breaking reveals that the error was fixed 1 day before the block and that I contacted the blocking admin before the unblock. the fourth unblock is also related to WikiProject Romania and the problem was caused by the Wikproject itself since the bot did exactly what has been told to. The person who requested the taggin is the one who fixed the list. Still this unblock was unrelated to the discussion here. -- Magioladitis (talk) 20:36, 3 February 2017 (UTC)

It's really troubling the unblocks were used here. I beg the ArbCom to really look into them. Moreover, really examine why Ramasoud2000 decided to add these unblocks to the case opened. -- Magioladitis (talk) 20:56, 3 February 2017 (UTC)

Comment by others:

It is clear that the community permits unblocking of bots blocked for technical reasons when the issue is resolved, per policy. All Magioladitis unblockings were of this type.

All the best: Rich Farmbrough, 22:56, 9 January 2017 (UTC).

Neither of those statements is true, Rich. In my experience, the community strongly frowns upon admins unblocking their own bots, especially when the issues aren't resolved (that this case exists is evidence that they haven't been). It's at the very least courteous to ask the blocking admin for permission or to ask an uninvolved admin or start a noticeboard thread. I don't mean any disrespect to you or to Marios—I think of you both as friends—but the unblocking appears to be a symptom of the same problem: that a handful of bot operators have a sense of urgency about what are mostly very small edits that simply isn't shared by the wider community. We simply can't understand why these edits can't wait for discussion or refinement or even for a few hours while a blocking admin's concerns are assuaged. HJ Mitchell | Penny for your thoughts? 10:34, 11 January 2017 (UTC)

"Automated or semi-automated bots may occasionally not operate as intended for a variety of reasons. Bots may be blocked until the issue is resolved." WP:BLOCK There have been discussions 2009, 2012, 2013, which have supported the principle that this includes the owner unblocking.

I don't see how "courtesy" enters into this. Next you will be saying it is "courteous" to notify someone who placed an {{Unreferenced}} tag, before removing it, having provided references.

I suggest you look at the workload that Magioladitis has undertaken, with others, to fix the CHECKWIKI errors, in order to understand why he wants to get on with things. As normal editors we create a constant stream of work for bots and gnomes (well I certainly do, as I know thanks to BracketBot and RefBot - and indeed you have seen my unadulterated typing). While we need to be able to control bots, and we do, we also need to erect the minimum of bureaucracy.

The unblockings were as follows (note, mostly ancient history, only 2 within the last 5 years)

1. 5 June 2016 (Issue handled)
2. 9 June 2014 (Problem is being resolved. Task won't resume until everything is 100% clear. Blocking admin contacted)
3. 6 April 2011 (2 edits in not Edit war. Involved editor should not block)
4. 25 December 2010 (Issues were fixed before block already)
5. 24 July 2010 (Resolved. Waiting for user's reply to resume work)
6. 16 April 2010 (resolved)
7. 24 March 2009 (Problem fixed)

All except #3 make it clear that the issue was fixed. #3 was concerned with a specific edit which had be reverted and re-applied. Yobot did not re-make the edit, so it appears that one was fixed too.

All the best: Rich Farmbrough, 20:48, 11 January 2017 (UTC).

Template redirect deletions

2) Magioladitis uses unapproved bots on his account to orphan template redirects on a grand scale. He then misuses his administrative tools to improperly delete the redirects.

Comment by Arbitrators:

Comment by parties:

No evidence that any related action was in grand scale. -- Magioladitis (talk) 16:51, 4 February 2017 (UTC)

Comment by others:

If you're talking about the stub templates, I dispute the second part of your statement that the deletion was improper and misuse of his admin rights). See my evidence. עוד מישהו Od Mishehu 16:32, 7 January 2017 (UTC)

Presumes finding 3. In the case of stub templates it seems appropriate, other templates would need careful examination to see if the deletion was appropriate. If, as seems likely, it was appropriate, then it is absurd to call it an misuse of admin tools. It certainly would not be a significant abuse, since the templates could easily be undeleted by any admin. All the best: Rich Farmbrough, 22:54, 9 January 2017 (UTC).

Unapproved bots

Contrary to the bot policy requiring approval and a separate account, Magioladitis has consistently operated unapproved bots, including adminbots, from his main account.

Comment by Arbitrators:

Comment by parties:

Comment by others:

No evidence supports this.

All the best: Rich Farmbrough, 22:52, 9 January 2017 (UTC).

I have added some data in my section of the evidence page, although it is always hard now to reconstruct now what someone did long ago. Overall, I don't see too much of a pattern of adminbots; I think it was just poor judgment to delete hundreds of pages manually with a vague edit summary. Before 2016, I have to go back to 2014 to find another example like it, which used Twinkle. — Carl (CBM · talk) 19:53, 15 January 2017 (UTC)

I very much doubt anyone is claiming Twinkle is a bot, though there have been some pretty outré statements. All the best: Rich Farmbrough, 22:52, 16 January 2017 (UTC).

Of course Twinkle is not a bot framework, which is why I mentioned it! But also, of course, the bot policy applies to bot-like editing of other sorts, even if a bot framework isn't used. As I said, I haven't seen evidence presented of sustained admin-bot-like editing. It looks more like isolated instances of large numbers of deletions to me. The main issue I see in those instances is with the edit summaries, an issue that has also been frequent with Yobot. — Carl (CBM · talk) 23:39, 16 January 2017 (UTC)

Twinkle is unquestionably a full-on bot when used in its automatic editing mode. You can press a button and have 1000s of edits happen while you lean back and drink lemonade. In its semi-automatic mode it can still be disruptive and treated as a bot under MEATBOT. 50.0.136.56 (talk) 22:00, 18 January 2017 (UTC)

Can you tell me how to enable Twinkle's alleged "automatic editing mode" so that I can "press a button and have 1000s of edits happen"? I suspect that you have confused WP:Twinkle with WP:AWB, but perhaps there is a secret automatic editing mode in Twinkle that I have never discovered. WhatamIdoing (talk) 22:00, 23 January 2017 (UTC)

@WhatamIdoing: Twinkle has some admin features that allow batch deletion, undeletion, and protection of a list of articles all at once. It's not exactly an "adminbot", though, and is easy enough to undo. I used to use batch deletion often when dealing with TfDs. Opabinia regalis (talk) 07:13, 24 January 2017 (UTC)

Thanks, but I don't think that's what the IP was talking about (i.e., "editing"). WhatamIdoing (talk) 19:09, 24 January 2017 (UTC)

I agree - I assumed the IP was talking about something other than the Admin features Opabinia mentioned. Doug Weller talk 19:15, 24 January 2017 (UTC)

The automatic deletions is the only fully-automatic option in Twinkle that I'm aware of. It lies in a tricky gray area of the bot policy, as do many other tools (WP:MTC!, for instance). In this case, I think our strong policies on administrative abuse cover things. There are many situations where it's undoubtedly useful. If an administrator were ever to use the option inappropriately (outside a genuine mistake), I would consider that a desysopping offense based on abuse of administrative tools in a high-volume setting. The consideration of Twinkle's D-Batch option seems out-of-scope to this case. ~ Rob₁₃^Talk 19:27, 24 January 2017 (UTC)

My first thought was batch deletion since that came up elsewhere in this case - it can make live edits under some conditions (if you ask it to remove backlinks or file usages). But now that I think of it the unlink function is capable of making lots of automated edits and is not admin-only. However, that has nothing to do with this case, and I suspect you're right that 50 meant AWB. Opabinia regalis (talk) 21:54, 24 January 2017 (UTC)

Unapproved bot tasks

3) Contrary to the policy requiring approval of all bot tasks, Yobot has consistently carried out unapproved bot tasks, such as bypassing template redirects.

Comment by Arbitrators:

Comment by parties:

Where is the evidence of that? Are you all in the same page? -- Magioladitis (talk) 22:51, 25 January 2017 (UTC)

Maybe you need to check Yobot's block log every once in a while. Ramaksoud2000 ^{(Talk to me)} 22:52, 25 January 2017 (UTC)

No proof that I systematically bypassed redirects. I provided proof that I tried to prevent even accidental bypassing by a) removing items from the bypass lists b) asking for a AWB bug fix.I already provided these evidence. Findings should not be based on speculations but on facts. -- Magioladitis (talk)

So what you are claiming now is that despite over 7 years (Wikipedia:Arbitration/Requests/Case/Magioladitis/Evidence#Bypassing_template_redirects) of trying, you were never able to fix this "bug", and continued to operate the bot with this bug, despite numerous blocks and requests to stop? That's not better. Ramaksoud2000 ^{(Talk to me)} 23:16, 25 January 2017 (UTC)

The complains were not for that. We are here a month... We should at least decide what were/are the complains about. There were complains for trivial/cosmetic/minor/non-significant edits, for bad edit summaries, for not fixing what is was supposed to be fixed. -- Magioladitis (talk) 00:19, 26 January 2017 (UTC)

This proves Wikipedia:Bots/Requests for approval/Yobot 44 that the problem was not the unapproved tasks but the fact that some people disagree with approved tasks. Compare to Wikipedia:Bots/Requests for approval/Yobot 16 and to the fact that this time I prosed the task to be done with no general fixes and as a sole task. -- Magioladitis (talk) 14:21, 4 February 2017 (UTC)

Comment by others:

Block evasion

4) Magioladitis evaded Yobot's block by running its program under his account while it was blocked, then attempted to conceal the evasion.

Comment by Arbitrators:

Comment by parties:

See evidence, especially this contribs link where he forgets the remove Yobot's standard edit summary with the bot approval and task number. When he notices, he removes the Yobot portion of the edit summary, but continues running the program. Wikipedia:Arbitration/Requests/Case/Magioladitis/Evidence#Magioladitis_evaded_Yobot.27s_block. Ramaksoud2000 ^{(Talk to me)} 22:11, 30 January 2017 (UTC)

The bot was not blocked for performing this task. -- Magioladitis (talk) 22:22, 30 January 2017 (UTC)

Considering that template redirect bypassing is not any approved task, and that you blame the redirect bypassing on "bugs" in the general program that you have been unable to fix for 7 years, why would you run the general program when it was blocked for being "buggy", as you claim? Either you are being disingenuous and you intentionally enable and disable template redirect bypassing, or you intentionally ran a buggy program after it was blocked for being buggy. Either way, it's not good. Ramaksoud2000 ^{(Talk to me)} 22:28, 30 January 2017 (UTC)

I don't understand. There were no errors in my (less than 20) edits. Bugs occur when trying to run in bot mode due to skip condition. Ramaksoud2000 please read the evidence and the other people's comments. -- Magioladitis (talk) 22:30, 30 January 2017 (UTC)

Comment by others:

Proposed remedies

Note: All remedies that refer to a period of time, for example to a ban of X months or a revert parole of Y months, are to run concurrently unless otherwise stated.

Magioladitis restricted

1) Magioladitis may not make any automated or semi-automated edits from this or any other account. This is to be broadly construed. Basic exceptions would be for use of the rollback tool for vandalism cleanup, and low volume use of standard gadgets such as Twinkle.

Comment by Arbitrators:

Comment by parties:

Copied from User_talk:Magioladitis/Archive_26#Temporary_editing_restrictions and modified to also prohibit bot edits under a bot account. Ramaksoud2000 ^{(Talk to me)} 02:01, 5 January 2017 (UTC)

Comment by others:

If you are minded to go ahead with this you need to make a broader set of exceptions including Hotcat "other than AWB" might be better than "such as Twinkle". I'd also suggest "reasonably construed" rather than "broadly construed" otherwise you license people to get him blocked for the appearance of using semi automation even on blatantly good edits, and we know that didn't go well in a previous case. ϢereSpielChequers 18:10, 5 January 2017 (UTC)

Hotcat is a standard gadget so it should already be allowed under this wording. AWB isn't a gadget. Ramaksoud2000 ^{(Talk to me)} 21:25, 5 January 2017 (UTC)

Reasonably construed sounds good too. Ramaksoud2000 ^{(Talk to me)} 22:16, 5 January 2017 (UTC)

In my years of watching bot disputes, ISTM that both operators and admins often take a binary and sometimes contorted approach to saying whether an editing sequence is automated or not (remember Betacommand and the "pattern of edits" restriction). I can dig up some diffs from old incidents if this is in question. Imho the best way to make a restriction like this is by limiting the number of pages per day the person can edit, without any regard to whether the edits are manual or automated. Then they are free to use automation as they like, as long as they stay within the page limit. Proposals like that have gotten some support in earlier cases and have the advantages of being simple to follow and hard to game. They haven't been implemented yet, but the stuff that's been done instead hasn't worked well, so maybe it's time to give this a try. 50.0.136.56 (talk) 06:51, 7 January 2017 (UTC)

I was blocked for a year by an admin who believed that the removal (or addition, I forget) of a single character from a single article constituted "automated editing". In the face of this kind of stupidity, how do you propose to define "automated or semi-automated".

Who do you propose would take over his tasks? The percentage of requests for assistance that get assistance is already very low.

And more importantly is there any sound reason for this draconian remedy?

All the best: Rich Farmbrough, 23:06, 9 January 2017 (UTC).

Rich, your 1-year block was ridiculous and made by a then-active AE admin with a known approach of treating arb remedies very literally instead of by comparing the disputed edits against the problem the arb remedy sought to fix. That approach eventually met enough community opposition that the admin later moved on to other areas.

If you made an unblock request I'd have certainly supported it if I'd been around, but as I remember I quit Wikipedia for a year or so after one of the sitting arbitrators on ARBRF ran a Facepalm unauthorized bot operation that undid a pile of work that another admin and I had done to clean up a previous unauthorized bot operation (the redlinked /x page above is a remnant of the cleanup). The earlier operation was forgiveable (I now prefer to see it as a one-off error by a generally good admin) but it was the topic of a big ANI that decided it shouldn't have been done and that supported the cleanup effort. I thought about opening a discussion with the arbitrator who rolled back the cleanup, but decided it wasn't worth it. I instead just headdesk'd and quit Wikipedia for a while. I suppose my view that Wikipedia has a "bot problem" to get under control deepened during that period, so I'm more curmudgeonly about it here than in earlier incidents.

Your RFA2 also happened while I was gone, and I would have supported it if I'd been around. I couldn't !vote per se, but I think I could have given a nuanced view of ARBRF and your block, which might have flipped more commenters to your side. Lots of your opposers didn't look into it closely at all: "messy arb case, long block = oppose". 50.0.136.56 (talk) 22:11, 18 January 2017 (UTC)

Magioladitis desysopped

2) For conduct unbecoming of an administrator, and misuse of the administrator tools, including AWB, Magioladitis is desysopped. He may regain adminship through a new request for adminship.; inserted underlined text 19:37, 20 January 2017 (UTC)

Comment by Arbitrators:

Comment by parties:

No evidence of tool misuse. -- Magioladitis (talk) 09:20, 17 January 2017 (UTC)

Comment by others:

Not opposed (not enough info) but would like to see clear evidence that it's justified outside of this bot nonsense. RF didn't need to be desysopped in my opinion. The one time I had (indirect) contact with Magioladitis acting as an admin, it seemed to me he did fine (I looked at some of Magioladitis's posts relating to a particular incident at another editor's request, and didn't see problems). Wikipedia has a long dreary history of "bots making people stupid", i.e. users of otherwise sane judgment seem to completely lose their common sense when it comes to bot editing. So if someone is in bot trouble, maybe it's enough to just keep them away from bots for a while. 50.0.136.56 (talk) 06:57, 7 January 2017 (UTC)

Since there's no substantive evidence of admin abuse, this is a ridiculous proposal. All the best: Rich Farmbrough, 22:57, 9 January 2017 (UTC).

As far as I can see, the only potential admin abuse is the unblocking of the bot, which even if it is a breach of policy (which isn't certain) was almost certainly a good-faith misunderstanding. If it's formally decided that bot operators can't unblock their own bot accounts, I've no doubt he'll stick to it. ‑ Iridescent 10:29, 10 January 2017 (UTC)

I still don't see convincing grounds for this either. 50.0.136.56 (talk) 01:10, 11 January 2017 (UTC)

I'll be proposing my own remedy in this direction because I think it should be considered, at least, but it relies more on WP:ADMINCOND. An administrator who fails to respond to repeated community requests to adjust their behavior over many years in addition to the few cases of questionable use of admin tools starts to look very much like someone who is not conducting themselves as an admin should. In particular, when you compare the concerns he responded to at the massive ANI thread a while back to the concerns here before the committee, there is no difference in the content of the complaints. Acting against the community's wishes after requests to adjust behavior is incompatible with being an administrator. Usually, it's even incompatible with editing the encyclopedia. ~ Rob₁₃^Talk 05:09, 17 January 2017 (UTC)

The way this is (currently) written, it sounds like the proposer is asking the Arbs to say that WP:AWB is an admin tool, which is wrong. WhatamIdoing (talk) 22:05, 23 January 2017 (UTC)

The era when "misuse of admin tools" was the only possible grounds for desysopping is long since over, so arguing about the presence or absence of tool misuse is a red herring. There's simply a subjective question of whether there's been enough generalized misconduct and/or poor judgment to warrant yanking the bit, whether or not "tools" were involved. That said, I feel the same way as before: I don't see the issues here rising to that level. 50.0.136.56 (talk) 22:47, 29 January 2017 (UTC)

Proposals by User:Rich Farmbrough

Proposed principles

Wikipedia is a collaborative project

1) English Wikipedia (Wikipedia) is a collaborative project to build an encyclopaedia. The content is provided, (or found, in the case of PD copyings) almost entirely by volunteers.

Comment by Arbitrators:

Comment by parties:

Comment by others:

If multiple people are interested in a topic and work on an article together, that's collaboration, which comes with the possibility of disagreements and conflict, but it's all in the nature of humans interacting. When someone or something makes a drive-by edit without engaging with the content at all, it can seem more like an invader or pest than a collaborator even if the edit isn't inherently bad. (Also: WP has editors-in-residence who I think get paid, plus some professors edit on "company time" while collecting a salary, etc.) 50.0.136.56 (talk) 05:32, 9 January 2017 (UTC)

A good point, and one I have raised myself in the past. Modified. All the best: Rich Farmbrough, 01:18, 16 January 2017 (UTC).

One thing that's struck me about almost all these bot cases is that bot operations are almost always done by solo coders and operators, i.e. non-collaboratively. This case may be a rare semi-exception. I remember thinking that your own bot writing was non-collaborative and wishing you'd get involved in MediaWiki development instead for a while, since that's a much more collaborative environment than bot coding. (It also would have extended your range as a technical contributor, shown the community that you could code in a more disciplined setting than the current wild west of the client side, and looked good on your CV, which I think was an issue you were concerned with at some point). 50.0.136.56 (talk) 23:23, 18 January 2017 (UTC)

The "Wiki" in Wikipedia means "fast" or "quick". It;s fast because you can just do stuff without a year of committee meetings first. (A charity I know of actually took nearly 30 years to change its name, although everyone agreed it should be changed.)

I have had plenty of involvement in top heavy bureaucracies, I did not and do not want to be involved in more than I have to.

MediaWiki development proceeds at a crawl, bugs are 1o or more years old, and are ignored either because "we have something better coming" (Lua took 3 or 4 years, parser functions were ready to go) or because "no one is available to review the submitted patch" or because of politics.

All the best: Rich Farmbrough, 20:31, 21 January 2017 (UTC).

Patches get in awfully fast depending on what they are and who's willing to implement them. Extended confirmation protection anybody? I was shocked by that. There is indeed a serious reviewing backlog, but that's all the more reason for you to get involved and start writing and reviewing good patches, instead of messing with those silly bots. 50.0.136.56 (talk) 22:51, 29 January 2017 (UTC)

Many types of contributors

2) In order to present Wikipedia properly a process of continuous improvement and corrective actions is applied. These include reverting vandals and fixing typographic, stylistic, and technical errors, as well as adding new information and references, and increasingly, working on non-article pages (non-article edits are in the majority). Each change to a page is called an edit. There is an established (if unscientific) non-exclusive typology of editors: Content creators, Bots, Gnomes, Vandal fighters, Vandals, Trolls... (The article WP:WikiBadger might bear reading.)

Comment by Arbitrators:

Comment by parties:

As a person that started my wikilife by mainly massively fixing redlinks manually I totally support this. -- Magioladitis (talk) 00:19, 17 January 2017 (UTC)

Comment by others:

Editors often gravitate to specialized areas, and m:namespace shift is a thing. But contributing to content is the pons asinorum of becoming a skilled editor. People incapable of contributing content shouldn't be doing quasi-administrative stuff like running bots, reverting other editors (except the most blatant vandals), enforcing "policy" in any way, etc. Wikipedia is about writing an encyclopedia, and the other stuff exists to support that purpose. 50.0.136.56 (talk) 23:13, 10 January 2017 (UTC)

I am sympathetic to this belief, but partly because I have seen examples where those who contributed little or no content caused issues in "high office". However there is no underlying reason that someone who has not contributed much content, capable or not of doing so, should not make a valuable contribution in administrative areas. Perhaps the difficulty of understanding the ecosystem without being part of it makes it unlikely, as does the motivation. But I would strongly welcome a number of facilitative types that I know in RL to resolve disputes, for example. All the best: Rich Farmbrough, 01:27, 16 January 2017 (UTC).

It's not the content contribution per se, but the experience of doing it that makes the person familiar with what content writers deal with. I wouldn't want your facilitative types helping with WP dispute resolution unless they did some content editing first, or maybe in a limited mediation-type capacity between editors who had agreed to work with them. 50.0.136.56 (talk) 23:27, 18 January 2017 (UTC)

Really it is a question of understanding. High court judges rule on copyright without having written a novel. All the best: Rich Farmbrough, 20:31, 21 January 2017 (UTC).

That's because they are part of a legal system, are bound by precedent, etc. Per NOTSTATUTE we are not in a legal system: we have to deal with actual facts on the ground and how they serve or don't serve the project. That means we have to understand how the project itself works and what its tensions are. Only a long immersion in editing can produce that understanding. 50.0.136.56 (talk) 22:54, 29 January 2017 (UTC)

"People incapable of contributing content shouldn't be doing quasi-administrative stuff like running bots" you can contribute plenty via bots without directly adding to content yourself. People have different skills and interests. If someone wants to code and run bots that the community finds helpful, that's a good thing. We shouldn't turn those people away out of some misplaced principle. I run (or at least I ran, it's been knocked out by an API change for now) User:Bibcode bot for years. If some NASA coder had the idea to code this in his spare time instead of me, what would it matter that this coder isn't interested in writing articles directly? The edits the bot would make would still be helpful and desired. Classism between GA-writing and gnoming has no place on Wikipedia. Headbomb {talk / contribs / physics / books} 13:17, 5 February 2017 (UTC)

Different types of contribution welcome

3) Positive contributions are welcome from all. In general these are edits that improve the encyclopaedia, in form, style or content, or are positive contributions to discussions "behind the scenes". There are many policies, guidelines, and essays, some overlapping and contradictory, to help determine what is positive.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Different types of errors are of different seriousness

4) The characterisation of an edit as an "error" is loose, and not always helpful. Both positive and negative edits can be so characterised. We might imagine a rough ranking of edits:

-6 Personal attacks and libel

-5 Copyright infringements

-4 Wrong information

-3 Removing information

-2 Introducing a typo or style error

-1 Making a non-rendering error

0 <-- Neutral point is here

+1 Fixing a non-rendering error:

+2 Fixing a typo or style error

+3 Adding information

+4 Removing erroneous information

+5 Correcting information

+6 Fixing copyvios

+7 Fixing BLP vios

Some types of errors, such as a misleading edit summary, will vary wildly, from being trivial (if form example a spelling error hasn't been corrected as claimed) to potentially serious (if, for example, a copyvio hasn't been removed as claimed).

Comment by Arbitrators:

Comment by parties:

Comment by others:

This isn't too far wrong, but there are a few missing points.

1. The importance numbering is not a linear scale, and in some cases, the ordering can vary as well; and there can be disagreement about how much importance to assign to a type of edit. Some people think these cosmetic and style-adjusting AWB edits are highly important, while others (the non-obsessives, to be cynical) would assign them a value like 0.0001 even if they were recognized as positive. That's close enough to 0 that we'd be fine without them. I liked this post from a big ANI a while back, which ended up getting two regular editors blocked over some microscopic issue from the Manual Of Style. The two editors apparently found this point overwhelmingly important, while nobody else gave a darn. Bot operators similarly often overestimate the importance of the edits their bots make, and I think that is happening here.
2. Some of those edits and actions (like revdelling or oversighting BLP vios, protecting pages, etc) require special privileges that are (supposedly) only granted to users of demonstrated knowledge and wisdom via RFA. Yes they need to be done, but not everyone is allowed to do them, and RFA rejects applicants all the time. Operating a bot is also a special privilege and people who can't do it non-disruptively shouldn't be doing it at all. You yourself posted that half of Wikipedia's mainspace/talkspace edits were from AWB users: are we really going to miss 3% of those edits? There's significant discomfort in the human editor community over the current volume of even non-disruptive bot edits. I don't see how to justify keeping the disruptive ones going. 50.0.136.56 (talk) 07:34, 9 January 2017 (UTC)

Hence the word "rough". All the best: Rich Farmbrough, 20:31, 21 January 2017 (UTC).

Number of errors can be expected to scale with contributions

5) The only people who don't make mistakes don't make anything. Of course more established editors are the smaller the percentage of errors they make, and the less likely they are to be serious. However the error rate should never be expected to become zero.

Comment by Arbitrators:

I've read the evidence on this case, and I'm still not clear on why exactly Yobot seems to be generating more complaints than other bots doing similar work. It could be that 1) Yobot makes more edits, 2) Yobot has a higher error rate, 3) Magioladitis selects lists of articles for Yobot to edit in a way that differs from other operators and results in more perceived disruption from its edits, 4) Yobot has acquired a reputation that lowers the threshold for new complaints, or 5) something I haven't thought of. I don't think we have enough information at the moment to point to edit volume as the primary issue. Opabinia regalis (talk) 23:17, 17 January 2017 (UTC)

Comment by parties:

Comment by others:

I don't think error rate has much to do with this. 1) an informative contribution with a spelling error is much more valuable than an information-free MOS adjustment even with no errors. 2) I don't believe your proportionality argument, at least by edit count. Where I'm from, if you get 3 traffic tickets in 1 year, your license is suspended. It's the exact same 3 tickets whether you drive 5000 km/year (occasional trips to the grocery store) or 300,000 km/year (travelling salesperson). It's not done by tickets per km driven.

I'm not a member of the Eric/Giano C*o*n*t*e*n*t C*o*n*t*r*i*b*u*t*o*r cult, but I do believe that the fundamental unit of Wikipedia participation is the addition of new information to articles. 10 edits adding encyclopedic facts or undoing vandalism, with 3 typos or small errors = positive contribution. 10 fact-neutral bot edits with 1 error = negative contribution. Even with 0 errors the bot edits are still (depending on editor philosophy) possibly of low enough value to not justify the nuisance. 50.0.136.56 (talk) 02:45, 9 January 2017 (UTC)

Indeed, but that is why the type of error is important, as is the type of good edit. What I am trying to illustrate here is that the naive error calculus is useless. If there are 1000 useful edits and 10 "errors" that are "cosmeticbot" errors, then that is a good result. If there are 1000 useful edits and 10 libels, then we may think differently.

The analogy of driving is a poor one, for various reasons. One might better think of failure rates for lightbulbs or something like that. There will always be a dead-on-arrival failure rate, and the rate for different manufacturers and different types of bulbs will be different. But you can be certain that the number of failures will be a monotone increasing function of the number of bulbs you buy.

All the best: Rich Farmbrough, 01:05, 16 January 2017 (UTC).

I also don't think the error rate is the main issue. In this case, the same error continued for years, despite multiple attempts by several editors to inform the bot operator. The error rate would be relevant if there were different, unanticipated errors over time. The issue here, however, is simply that the operator did not take seriously the need to resolve a single kind of error. The error rate for this particular error should have become zero very quickly, just like any other bug that could be fixed in a bot. — Carl (CBM · talk) 01:09, 16 January 2017 (UTC)

There is a huge difference between "the same error" and "the same kind of error". You fail to grasp that Magioladitis has effectively run a large number of different bots, on the same framework. Certainly the framework can be improved, but when one is not a C# programmer that may not be easy (even when an admin insists that you do it or be blocked!).

Taking as a proxy YoBot's block log, mentions of "cosmetic changes" are down to once every two years from several times a year in 2011. I would call this a considerable improvement.

All the best: Rich Farmbrough, 23:33, 19 January 2017 (UTC).

I think that this is a general problem in the community. We definitely have editors who complain about "ten errors" without noting that it's "ten errors out of ten thousand edits". (Yobot has more than four and a half million edits so far; even with "six nines" precision – which is far more than any Wikipedian should expect – you'd expect about five errors.) We do need better statistical literacy and a culture shift in the community around this issue, but I'm not sure that this is especially relevant to this case (despite a couple of people saying "look at the raw number of complaints and/or blocks!"). WhatamIdoing (talk) 22:22, 23 January 2017 (UTC)

We do need a culture shift, but in the other direction, about how often these automated editing sprees are useful regardless of whether there are errors. 50.0.136.56 (talk) 22:56, 29 January 2017 (UTC)

AWB (and other tools) useful and prolific

Percentage of edits in each name-space up to 2012 - dark blue represents article space, dark green article talk

6) Wikipedia is the product of some 800 million edits. (868,215,111) The number of edits, and the amount of change in each varies enormously. Less than 40% of edits are to articles, that is some 240 million edits. AWB has made something over 120 million edits.[2][3], mostly to articles or article talk page. AWB is therefore responsible for something around half of the edits to article/article talk space. Other tools and bots will add substantially to this percentage.

Comment by Arbitrators:

Comment by parties:

I agree with this (and the point below about it hurting Magioladitis' case). I filed this case with the intent that it be limited to Magioladitis' actions. If someone codes a vandalbot with Python, that doesn't mean all Python bots should be banned. The tool should be separated from the operator. Ramaksoud2000 ^{(Talk to me)} 00:43, 9 January 2017 (UTC)

I agree. -- Magioladitis (talk) 15:34, 21 January 2017 (UTC)

Comment by others:

Rich, you do realise that you're significantly harming Magiolatidis's case here, since it raises the obvious point of "if Yobot is only responsible for 3% of AWB edits why isn't it only responsible for 3% of complaints?" ‑ Iridescent 18:16, 8 January 2017 (UTC)

Firstly there is no evidence that it is responsible for more than 3% of the complaints, or indeed any specific percentage.

Secondly I conducted an experiment in 2012 or so which seemed to show that the number of complaints was decreased by not adding "using AWB" to the edit summary.

Thirdly Magioladitis is a pioneer, parts of the AWB infrastructure that he has pioneered have been debugged by him and are now available to others, who will of course not have the errors that he has removed.

All the best: Rich Farmbrough, 20:41, 8 January 2017 (UTC).

Re: the second point, I have often wondered if there has developed a visceral reaction by some to AWB rather than necessarily the edits made by it (after all, a determined manual editor can make all the same edits in one swoop). It seems that some people actually think it's fully automated, and its users are just "setting it and forgetting it". Maybe the tool's name should be changed to be more accurate: "Computer-aided Auto-listing, Auto-suggesting Manual Editor" Stevie is the man! ^{Talk • Work} 20:57, 8 January 2017 (UTC)

No, that has nothing to do with it. It's the speed, volume, and pervasiveness of the edits, and how they make it difficult to navigate article histories. I'll try to write something clearer when I get back in a few days. But if you understand why people are encouraged to use "git rebase" before submitting code patches to an upstream repo, that might help understand how annoying users can find it to look for the human decisions in an article history that's completely full of bot poop. 50.0.136.56 (talk) 04:50, 11 January 2017 (UTC)

When used in a non-meatbot semiautomated manner, like when I'm using it, there is a human decision involved, but other users might think it's "bot poop" because it says "using AWB". I make custom-developed (mostly RegEx-based) Find/Replace edits (to address MOS and technical issues) on top of GenFixes, auto tag and typo fixes. It is indeed a computer-aided manual edit. I review all changes before saving. On top of this, oftentimes my F/R suggestions are incomplete and need manual intervention, as RegEx can't always fix everything perfectly and there's some weird things out there in the wilds of articles. I even often find additional things to break out of AWB to address by regular manual editing.Stevie is the man! ^{Talk • Work} 21:22, 19 January 2017 (UTC)

I think we're here to do what is best for the Wikipedia, not help or harm Magioladitis' case. I am seriously concerned about his behavior to an extent, but I am also seriously concerned about a potential overreach in any attempted correction. I am most concerned about how this site ends up as a result. Stevie is the man! ^{Talk • Work} 20:57, 8 January 2017 (UTC)

I agree. And I think the lack of understanding about software regressions, and the difference between a handful of "cosmeticbot" edits, compared with the sort of issues Erik9Bot and GaneshBot created makes it harder to prevent such overreach. All the best: Rich Farmbrough, 00:35, 9 January 2017 (UTC).

Rich, thanks very much for those stats. I had wanted such numbers myself, and found something on stats.wikimedia.org saying that around 9% of en.wp edits were bot edits, which seemed way too low, so I was thinking of manually surveying a sample myself. I think you've pinpointed the reason it looked low.

Do you have a year by year breakdown? I.e. can you tell how many en.wp AWB edits there were in (say) September 2016, September 2013, and September 2010 (I see they only go back to 2009)? Something really has happened to the editing environment here if more than half the edits are now automated.

My own starting point is that the number of human editors on Wikipedia has been declining steadily since 2008 or so, but the number of edits per month has been increasing, which I see as a sign of a "robot uprising". But I didn't have good numbers, so yours really help. 50.0.136.56 (talk) 03:00, 9 January 2017 (UTC)

I'm not sure who runs the ARB stats server, possibly User:Rjwilmsi will know. Across all projects I suspect the biggest (and biggest waste of effort) number of bot edits were to updating (and then removing) interwiki links. Given that we were "a couple of days of effort" away from "reasonably efficient cross-wiki transclusion" it beggars belief that we had to wait all those years and tens of millions of edits until Wikidata could provide the solution.

All the best: Rich Farmbrough, 21:44, 9 January 2017 (UTC).

I don't know what the deal with interwiki matching was, but I thought of it as one set of bots that worked pretty well. Some inaccuracy was (and still is) unavoidable but I was impressed with how little there was. This is the first I heard that there was a project to handle it server-side. Interesting. I know it's in wikidata now but none of the details as I haven't been keeping up with the times. 50.0.136.56 (talk) 04:43, 10 January 2017 (UTC)

It worked well for us, and the other highly active wikis, but for some small language wikis the vast majority of edits were interwiki bots. All the best: Rich Farmbrough, 01:33, 16 January 2017 (UTC).

Just to answer a few stats related things in the thread above. Article edits are a lot more than 40% of edits, more like 70%. With the increase in editing since 2014 all edits are now back up to 5 million a month (as measured by revisionIDs). Wiki stats measures Article edits only, currently they are running at around 3.5 million a month - roughly 70% of total edits. this chart from Erik Zachte shows 454 million live article edits which would be 60%. So it may have changed over time - or we may have deleted more article edits than I thought. Bot edits have historically been 9% of mainspace edits, but in recent years according to Erik's chart this has dropped to about half that level, due presumably to the rise of the edit filters and the migration of interwiki links to Wikidata. We've know that very active editors, those making over 100 mainspace edits, have been increasing for the last two years. ϢereSpielChequers 20:48, 27 January 2017 (UTC)

I'd like to know how many actual informational edits are being made, where informational edit = something that puts encyclopedic info into an article that wasn't already in it--reformatting/shuffling/fixing spelling errors don't count as informational. That we're seeing more uninformative editing doesn't sound like something to celebrate. Re deleted edits: I remember figuring that around 7% of Wikipedia's edits across all namespaces have been deleted. That was done by counting the number of live revisions in a metadata dump and comparing with the highest revision number, late last year. 50.0.136.56 (talk) 23:01, 29 January 2017 (UTC)

Note: Rich asked on my talk page if I ever located the source of the AWB stats (I haven't), due to concern that the stats themselves might be inaccurate. I haven't been able to pursue this: does anyone else have any ideas? 50.0.136.56 (talk) 23:02, 29 January 2017 (UTC)

Proposed findings of fact

Magioladitis

1) Magioladitis (talk · contribs · blocks · protections · deletions · page moves · rights · RfA) has been an active Wikipedia editor since 2006, and an administrator since 2008. He has extensive experience with and expertise in the use of wiki-automation, including both fully automated bots (such as Yobot and semi-automated tools (such as AutoWikiBrowser). He has made approximately 5.4 million edits, 4.6 million with YoBot, and 800,000 under his own account. In addition he has contributed heavily to discussions on automated editing, and has encouraged others to take over, singly or jointly, the CHECKWIKI tasks. Magioladitis has been for some time one of the primary developers of AWB.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Edsger W. Dijkstra famously wrote:

My point today is that, if we wish to count lines of code, we should not regard them as "lines produced" but as "lines spent": the current conventional wisdom is so foolish as to book that count on the wrong side of the ledger. (EWD 1036, also see EWD 962).

I'm not trying to bag on Magioladitis here, and I get that we can't write an encyclopedia without spending edits. But I think we as a community, on seeing a very high edit count, should be more in the habit of asking what we got for the edits instead of just admiring the high number. 50.0.136.56 (talk) 22:03, 11 January 2017 (UTC)

It is perhaps worth repeating that the idea of General Fixes is to reduce the "edits spent". All the best: Rich Farmbrough, 01:35, 16 January 2017 (UTC).

I just posted a big writeup to /Evidence of a "general fixes" edit that didn't need to be done at all. Maybe a few more are like that too. Maybe a lot more. Maybe even all of them. 50.0.136.56 (talk) 12:33, 18 January 2017 (UTC)

Cosmeticbot non-trivial

2) Despite the assumptions that "COSMETICBOT" is simple to comply with, it can be relatively subtle. Example: This edit, cited in the Evidence page as an example does not count as a cosmetic edit, since it (correctly) changes the rendered spacing between two sections.

Comment by Arbitrators:

Comment by parties:

"removeUselessSpaces" is one of the few functions specifically mentioned in WP:COSMETICBOT as something not to do. Also, how is removing two extra lines not cosmetic? What is your definition of cosmetic? In addition, as demonstrated at Wikipedia:Arbitration/Requests/Case/Magioladitis/Evidence#Whitespace-only_edits, Magioladitis admits numerous times that whitespace-only edits should not be performed. Ramaksoud2000 ^{(Talk to me)} 23:55, 6 January 2017 (UTC)

Unfortunately this level of understanding is all to common. The general consensus is that white-space in the wiki-source is of relatively minor importance, when it doesn't affect the rendered page. Even then I doubt that anyone would claim that white-space does not affect readability. In this case the white-space is not "useless white space" but "damaging white space", since it damages the layout of the article, by inserting extra white-space between the sections.

Your reply supports my contention that people are all to ready to over-generalise and over-simplify - even when they have had the facts clearly explained to them.

All the best: Rich Farmbrough, 00:14, 7 January 2017 (UTC).

Oh and a further note " removeUselessSpaces" is a pywikibot function, not an AWB function. As far as I know Magiolatitis has never used it. All the best: Rich Farmbrough, 00:23, 7 January 2017 (UTC).

I understand what it is. This is the first link of the evidence section. User:CBM all the way back in 2010 complains about trivial editing. Among the edits provided are [4] [5] [6] [7] [8] [9], which just remove "damaging" lines, as you call them. Magioladitis claims that he is working to fix this "bug" to prevent more edits like that from ever happening again. Rich, are you suggesting that even edits that Magioladitis admits fall under the category of cosmetic edits, aren't cosmetic? Ramaksoud2000 ^{(Talk to me)} 00:41, 7 January 2017 (UTC)

Ramaksoud2000 Are you bringing edits that happened 7 years ago? -- Magioladitis (talk) 00:49, 7 January 2017 (UTC)

No. I am showing that you have been aware for a very long time what kinds of edits are not allowed. You are correct that these happened 7 years ago, and 7 years ago you admitted they should not have happened. Yet you never stopped performing them. If you had stopped 6.5 years after you became aware, this case would not exist. Ramaksoud2000 ^{(Talk to me)} 00:52, 7 January 2017 (UTC)

You are mixing different problems... again. This edit is valid whitespace removal since it changes the visual output. Rich is right on that. My mistake not to notice earlier. -- Magioladitis (talk) 00:53, 7 January 2017 (UTC)

So you withdraw your admission? It doesn't really matter. That edit on the evidence page was picked from the edits picked by Rob at User_talk:Yobot#WP:COSMETICBOT_yet_again, which got Yobot blocked the most recent time. There are plenty of additional whitespace-only edits there. You also did not dispute that the edits which caused the block were cosmetic. Ramaksoud2000 ^{(Talk to me)} 01:15, 7 January 2017 (UTC)

Ramaksoud2000 How do you define "cosmetic"? In the section you mention the bot did not work s indented. -- Magioladitis (talk) 01:18, 7 January 2017 (UTC)

Unfortunately CBM does not come with clean hands. He has made hundreds of edits that would be contrary to Cosmeticbot, and hundreds more deliberately making Wikipedia pages worse, for the furtherance of some personal agenda. All the best: Rich Farmbrough, 01:10, 7 January 2017 (UTC).

Comment by others:

The comment by Rich Farmbrough concerning CBM's "unclean hands" is totally irrelevant to this case. The locus of this complaint is not whether normal human editors can make cosmetic changes, but whether bots or editors & bot-operators using AWB in a bot-like manner can do so. It is perfectly clear that the latter is forbidden, just as it is clear that human editors, working at human speeds, can, in the normal course of their editing, fix cosmetic problems should they decide to do so. The fact is that bots/bot-like AWB editing is deliberately hobbled in this regard because of the massive collective affect they can have on the project, and the total lack of effective policing of their editing by outside agencies. Beyond My Ken (talk) 10:55, 7 January 2017 (UTC)

It is certainly irrelevant to this FoF. But it is salient to Ramaksoud2000's comment which cannot be allowed to stand unchallenged. If Ramaksoud2000, or a clark can withdraw that irrelevant comment, then the response can be withdrawn too. All the best: Rich Farmbrough, 12:46, 7 January 2017 (UTC).

Calling CBM's hands "unclean" is not a challenge to Ramaksoud2000's statement, it's a challenge to the integrity of a third party, which is irrelevant not matter which way you look at it, except as a matter of unnecessary rhetorical overkill. I see no reason that R2000 should retract their statement. If you disagree with it, counter its essential factuality, don't throw mud into the air to see who it sticks to. Beyond My Ken (talk) 13:56, 7 January 2017 (UTC)

It sticks to Ramasoud, who has not done his research to establish the context of a dispute seven years ago. I'm sure if you go through any user talk page history you will find editors saying bad things about them. Context is important. All the best: Rich Farmbrough, 18:20, 8 January 2017 (UTC).

Really? I think that if I go through your history I will find misuse of automated tools after being forbidden by ArbCom from doing so, and, I believe, a desysopping for cause. Does any of that have any relevance to whether your suggestions are good or bad? I don't think so, and neither does your remark about CBM's "unclean hands". I think you might examine your own hands, Rick Farmbrough, before making accusations about other people's. All the best. Beyond My Ken (talk) 04:24, 9 January 2017 (UTC)

The point of mentioning CBM's complaint was that Magioladitis agreed with CBM that the edits should not be performed. There is a whole evidence section dedicated to demonstrating that Magioladitis is aware of and agrees with what is considered cosmetic. I'm not sure what Rich's issue with that link is, or why it needs to be "retracted". The person that Magioladitis agreed with is irrelevant. Ramaksoud2000 ^{(Talk to me)} 04:43, 9 January 2017 (UTC)

And that he had such an agreement is irrelevant to whether or not "COSMETICBOT" is easy to diagnose. You, of all people cannot argue that it is, since you failed, despite, (one would hope!) taking extreme care over your selection of diffs, only to make a mistake. That is fine, these things are subtle, although they appear simple. However if you are not "versed in the arts" your statements should carry less weight, since they are bound to be based on a naive approach to AWB, bots, and Wikipedia editing in general.

All the best: Rich Farmbrough, 21:50, 9 January 2017 (UTC).

So, it is your contention that only bot operators are qualified to criticize the editing of bot operators, that the complaints and comments of non-operators should "carry less weight", turning bot operators into uncontrolled "super editors" with no effective check on their actions. But, at the same time, COSMETICBOT, is "too subtle" for even bot operators to interpret correctly, in which case I suggest that COSMETICBOT be replaced with a simple blanket proscription against bots making non-rendering cosmetic changes of any type at any time. They can still do the myriad other helpful edits they do, but would be out of the business of cosmetic edits, which are petty damned easy to define. Beyond My Ken (talk) 05:35, 10 January 2017 (UTC)

Of course I do not make that contention, nor any of the others that you seem to think flow from it, though they do not. Nor the contention in your second sentence - it is understandable by anyone, but it is easy to make a mistake over what is or is not cosmetic. Careful application of thinking apparatus will help - one might assume those posting evidence here would be careful, but who knows? I have seen worse evidence.

The purpose of COSMETICBOT is to discourage mass editing which results in no significant useful change, not to be used as a rod to punish operators whose bots occasionally fail to make the substantive edit.

All the best: Rich Farmbrough, 23:54, 10 January 2017 (UTC).

Ken, I think one issue being argued is whether collapsing whitespace inside wiki markup is cosmetic: see my answer to Newyorkbrad here. Regarding "super editors", I've had similar thoughts, and considered writing a principle referencing the old meta essay section m:Power structure#Technocracy. The section was originally written in the early days of Wikipedia and not edited for a long time, though it has been slightly updated recently. A point of interest was that it contemplated the actual wiki software developers using their control of the servers to win editing disputes. There were basically no bots back then, and getting at the servers has always taken a level of inside access that was at least controlled by the top devs and Jimbo.

The servers are now run by the WMF so basically only WMF staff ops can touch them; deploying new code requires review, testing, signoff by WMF staff devs, etc.; and there's a strong separation between WMF actions and WP editorial issues. Conflicts between developers and editors are rare and are usually about technical issues (Visual Editor *cough*). So I was going to write that the authors of the Power Structure essay had foreseen and remarked on the prospect of rogue programmers controlling Wikipedia through the servers, and that problem was mostly solved by putting the servers under professional staff (there actually were a few incidents in the early days).

But the essay authors hadn't foreseen the widespread easy deployment of bots by literally anyone, enabled considerably by the MediaWiki API implemented around 2006-2007(?) in conjunction with Yurikbot (earlier bots had to use the web interface which was much less convenient). So that basically created a new technocracy of totally self-selected people like Betacommand controlling Wikipedia in ways the actual developers could not. The "fait accompli" principles found in various past arb cases show lots of bots or bot-like editing used in content wars; the amount of automated editing is a headache for human readers and editors in its own right; and even the approved "good" bots have helped Wikipedia's obnoxious bureaucracy scale to levels that couldn't have been reached by humans editing with their own two hands (that last part is my personal opinion but I believe it).

You might be able to write something better with this idea than I can, so feel free to use it. I'll try to put up something about the narrower issue of usability impairment. 50.0.136.56 (talk) 20:31, 10 January 2017 (UTC)

Proposed remedies

Note: All remedies that refer to a period of time, for example to a ban of X months or a revert parole of Y months, are to run concurrently unless otherwise stated.

Magioladitis requested not to unblock Yobot

1) Magioladitis is reminded that unblocking Yobot really annoys some people. As such he is requested not to do so, unless there is explicit consent - in advance or in arrears - from the blocking admin, or from another admin, for the unblock.

Comment by Arbitrators:

Comment by parties:

Yes, I respect that. -- Magioladitis (talk) 00:22, 17 January 2017 (UTC)

I provided evidence that all the blocks were unrelated to the "cosmetic changes" editing. I also underline the fact that no evidence of connection was ever presented apart from numbers ("Yobot has been blocked xx times") just to impress. -- Magioladitis (talk) 21:09, 3 February 2017 (UTC)

Comment by others:

Other parties requested to chill

2) We are talking about a relatively small number of edits that "do not improve the page enough". We are also talking about development of systems that, if the past is anything to go by, will have a major impact on Wikipedia in the future. In this context the significant goal is to look to the future, not to quibble about the past.

Comment by Arbitrators:

Comment by parties:

Comment by others:

It is precisely the possible future "major impact on Wikipedia" of these systems that has some of us alarmed, since it's potentially a negative major impact. We need to get it right, and we're not doing that. 50.0.136.56 (talk) 23:12, 29 January 2017 (UTC)

"potentially a negative major impact" Uh, what? The issue here is cosmetic edits. These are by definition such extremely low-impact edits that the cost of reviewing them exceeds the benefits gained by making them. There is zero 'possible future impact' of any kind with those edits, much less major ones. Headbomb {talk / contribs / physics / books} 13:31, 5 February 2017 (UTC)

I think the main risk of negative impact is indeed the wasted effort in reviewing the edits, although the use of large-scale edits by some bot operators to create "facts on the ground" in favor of certain preferred styles is also a risk. There isn't really a major potential positive impact from the cosmetic edits, because somewhat by definition they don't affect the content of the page. — Carl (CBM · talk) 13:37, 5 February 2017 (UTC)

Proposals by User:Iridescent

Proposed principles

Automated and semi-automated editing is often valuable

1) Fully automated bot editing and semi-automated editing scripts perform an important and valuable function on Wikipedia.

Comment by Arbitrators:

True and worth stating. Newyorkbrad (talk) 01:39, 12 January 2017 (UTC)

Comment by parties:

Comment by others:

This is worth stating explicitly as part of the framework to this case. Just looking at the initial statements, people are already losing sight of the fact that this case is about one specific aspect of the conduct of one specific editor, not a general "creative humans vs soulless machines" or "luddites vs efficiency" (delete according to personal taste) cage match, and that "bots are inherently disruptive and should be deprecated" is not on the table. ‑ Iridescent 12:03, 7 January 2017 (UTC)

I don't think anyone contests this idea, but I don't see its relevance here. It's like having a "biographies of living people are an important and valuable area of Wikipedia content" principle in a case about some editor editing BLP's in inappropriate way, out of a perception that too many people want to delete all of Wikipedia's BLP's. That's a straw person.

What's going on here is more like: there was once a period of widespread irresponsible BLP editing in Wikipedia, and it was a pattern seen across enough different incidents that it had to be addressed at a policy and practice level, so now BLP editing has been greatly tightened up. The Badlydrawnjeff arbitration was also about one aspect of one editor, but it was a focal point of a lot of wider BLP-related concerns.

I'd say bots can be valuable, I'm happy to genuflect daily in the direction of Cluebot; but at the same time I'd say many bots are non-valuable solutions looking for problems, and that implementing and operating bots is a demanding task that should be reflected by high standards being applied to their implementations and operators. A good bot operation is characterized by the operator being cooperative and the bot's edits not provoking significant disputes. Cluebot, Lowercase Sigmabot etc. are examples of this. Yobot obviously is not. Operators who persistently don't meet such standards probably should not be running bots. "Anyone can edit" is a principle that applies to human editors, not to bots. 50.0.136.56 (talk) 20:32, 7 January 2017 (UTC)

I don't consider it a strawman argument. The Badlydrawnjeff comparison isn't valid, as there was never any realistic possibility that it would end with "Wikipedia will no longer host biographies of living people". This case, notwithstanding the supposed "narrow scope", certainly has the potential to set some fairly drastic precedents—in fact, I'm about to nip down a couple of sections and propose one myself—and I do think it warrants mentioning in the result that this is a case of one specific editor, not Bots v Humans. Remember, people referring to whatever's decided here in later years will just be looking at the case page, not the workshop or talkpages, so it's worth making it clear that this is "alleged problems with Magiolatidis" and explicitly not "alleged problems with bots". ‑ Iridescent 17:29, 8 January 2017 (UTC)

Sadly we have reached a stage where this needs to be said. ϢereSpielChequers 16:41, 8 January 2017 (UTC)

• Support if some kind of balance is presented, oppose otherwise. @Brad: I agree that the principle is true, but it's only "worth stating" if there is significant doubt about it in the community. If there is such doubt, then that should be acknowledged explicitly, and it should be an issue for the community to consider rather than something dismissed by a one-sided principle like the above. 50.0.136.56 (talk) 23:37, 18 January 2017 (UTC) (I'm about to leave for several days).

"Often" is plenty of balance. The reason for these "obvious" principles is to provide axioms to argue from. Unfortunately the structure of Principles/Fofs/Remedies does not really support (although it allows) coherent argument. All the best: Rich Farmbrough, 23:35, 19 January 2017 (UTC).

The reason particular axioms are chosen to argue from is because the case is about someone being found going against them. So in a case of an uncivil editor there would be a WP:CIVIL principle; and the case got to arbitration in the first place (rather than stopping at ANI) because there was significant division in the community about the behaviour in question. In a case like this where no one is being particularly uncivil, it would be surprising to find a civility principle, since there would be no factfindings to follow from it.

Similarly a "bots are often useful" principle only makes sense if there's significant division in the community about that. And if there is such a division, it should be recognized. 50.0.136.56 (talk) 23:29, 29 January 2017 (UTC)

Proposed findings of fact

Magioladitis's work has greatly improved Wikipedia

1) Much of Magioladitis's work on Wikipedia, and in particularly his work developing WP:AWB, has been of great benefit to Wikipedia, and made the work of other editors significantly easier.

Comment by Arbitrators:

Comment by parties:

I'm unfamiliar with his work, but this appears to be the case, and it makes this situation unfortunate. I think that this may be why others have been so reluctant to do more than just warn Magioladitis. Surely, if I or another editor had ignored warnings and blocks for so long, I would have been blocked for good a long time ago. Ramaksoud2000 ^{(Talk to me)} 02:32, 8 January 2017 (UTC)

Ramaksoud2000 (I hope it's ok for me to reply here), re if I or another editor had ignored warnings and blocks for so long....: disputes involving bots have historically been incredibly persistent and hard to manage. See for example the three Betacommand arb cases, the Date Delinking case, the RF case, Betacommand's entire department at WP:AN, the antics of editors like Kumioko and OccultZone (both now banned for non-bot-specific reasons but who were also problematic likely bot operators), yada yada. The problem seems to be a hypnotic or addictive effect that bot operation has on some editors, combined with a bot admiration society that supports every bot operation it sees no matter how ill-advised, making it hard to get such drama stopped.

I'm not applying any of this specifically to Magioladitis since I'm mostly unfamiliar with his activity having had no real contact with him up til now. I'm just saying I've seen disputes like this more times than I want to remember, and at first glance this case appears to be more of the same. I think it's desirable for arbcom to take up a practice of ending these disputes forcefully and decisively instead of letting them repeat. I've been considering proposing a workshop principle to that effect. 50.0.136.56 (talk) 04:19, 8 January 2017 (UTC)

Comment by others:

Because so many people—including a lot of people who should know better—misuse it for inappropriate functions, and because its "general fixes" function has acquired such an inappropriate scope creep over time, it's easy to lose sight of the fact that AWB is a fantastically useful piece of software in certain circumstances. Whatever the outcome of this case, Magiolatidis deserves commendation for being one of those who keep it operational since its creator retired a decade ago. ‑ Iridescent 12:03, 7 January 2017 (UTC)

I'm not asking for formal /Evidence but I think a principle like this should be accompanied by some persuasive examples or descriptions, maybe on the workshop talk page. My prior contact with Yobot was one annoying edit (I'll write it up later) and I had no idea until this case opened that Yobot was the subject of repeated blocks and disputes. So I didn't really know about its good side or bad side. One thing I might try is reviewing a sample of Yobot or AWB edits to see how many I'd consider important. If none of them are important then maybe they're collectively not important. 50.0.136.56 (talk) 20:55, 7 January 2017 (UTC)

The most obvious 'evidence' is the number of people who use the software he co-wrote regularly and without incident. (According to the edit count tool, I alone have made over 100,000 edits using it—it really does make life so much easier if something changes its name and you need to make the same search-and-replace across 5,000 separate articles.) I'm not expecting this to come anywhere near making it into the final result, so am not going to waste much time discussing it or providing evidence—it's more there to serve as a reminder that we're talking about someone whose edits are primarily helpful and just has some issues with what is and isn't appropriate, rather than someone sitting in his bunker cackling "ha ha ha, one day my robot army will march triumphant through the smouldering ruins of Wikipedia". ‑ Iridescent 16:15, 8 January 2017 (UTC)

Ok, that's the kind of explanation I wanted, though I didn't realize Mag's contribution to AWB was that large.[10] Still though, according to Rich's figures, about half of the mainspace and talk edits on en.wp since 2009 were made by AWB. That's cumulative: I've asked Rich for a year by year breakdown if he knows where to find one. Automated edits are probably at least 2/3 of the total when you count other bots besides AWB. Does that sound as crazy to you as it does to me? Are you saying that many AWB edits is a good thing? Or should it be higher, like 90%? Maybe we should also have a principle that says human edits are sometimes valuable, since we're in the minority now.

I don't see that number as an "incident" in the sense of a building on fire. It's more like gradual climate change caused by global warming. The editing atmosphere has shifted and I think not for the better. If you really think a principle is needed to say bot edits can be valuable, then it should probably also mention that parts of the human editing community have developed reservations towards them. (That's why you want the principle, after all).

FWIW, I don't think of bot operators as cackling in a bunker. I don't actually understand why anyone becomes so insistent on running contested bot operations, but it comes across as maybe a desire to mark territory by editing as many articles as possible, like how dogs try to pee on as many trees as they can. I.e. it's an in-the-moment instinct rather than a long range evil plan. Maybe someone with more insight or direct knowledge can explain more. 50.0.136.56 (talk) 05:00, 9 January 2017 (UTC)

Drifting slightly off the point, but I can certainly believe that semi-automated edits make up a large proportion of edits, because every vandalism revert with Huggle or Stiki counts as either two or three edits (reversion, warning, reporting). Performing the very unscientific test of searching for the string "using AWB" in the last 500 entries in recent changes I get 6 edits out of 500; performing the same search on my watchlist finds 105 articles out of 1315 to which the last edit was performed using AWB, which seems a more realistic figure. Remember, there are AWB bots which wander around tidying up the templates at the tops of talk pages, which are going to artificially inflate the count quite a bit. Also, bear in mind that a lot of human actions on Wikipedia trigger a bunch of bot edits—when an article is promoted to Featured Article status, for instance, bots will add the yellow star to the article, add the rating on the article's talk page, close the FAC discussion, add the article to the lists on the relevant Wikiprojects and add the author to WBFAN. That's a 5-1 bot/human ratio, but all the bot edits are things which would need to be done anyway (except for WP:WBFAN, which I'd happily delete and salt) and require technical fiddly markup which would waste a lot of time to do manually. This isn't a case of marking territory or bots running wild; it's a case of automating the uncontroversial but necessary tasks which need to be done so as to free up the human editors to do something useful. (If you really want to focus your ire somewhere, cast your eye over WP:AutoWikiBrowser/Typos and note how many of these are completely arbitrary personal preferences about comma and dash formatting, and not "typos" in any sense of the word.) ‑ Iridescent 08:16, 9 January 2017 (UTC)

Thanks, that's helpful, though we did just fine without those bots 10 years ago, an era that was the height of Wikipedia's success. When we promoted an FA we didn't add a lot of "technical fiddly markup" (beyond maybe a template) by hand. We just did stuff with less of the technical fiddly markup, and it was fine. It's like how a modern web site sends you a megabyte of Javascript before delivering any useful information, where a 1990's site would have sent a kilobyte of plain HTML that rendered 10x faster and conveyed the same info just as well. Solutions looking for problems. (I have to go to bed soon, back tomorrow, then maybe away for a few days). 50.0.136.56 (talk) 08:34, 9 January 2017 (UTC)

10 years ago there were more people and fewer articles. About the "fiddly markup" you are absolutely right. And that is why we need to simplify the mark-up where we can, ideally getting rid of all the html-like stuff in articles, reducing the number (thousands) of template redirects to maintenance templates and so forth. Even the simplest of these seem to act as lighting rods.

All the best: Rich Farmbrough, 22:00, 9 January 2017 (UTC).

I still don't understand why we need more bots now than we did in 2007, if there's less human activity to give the bots things to do. Is it now bots creating work for other bots? I feel like Wikipedia is in transition from an anthropocene epoch into a robotocene one. 50.0.136.56 (talk) 06:48, 12 January 2017 (UTC)

No, humans are relieved of clerking and some copyeditng functions. We also have some unnecessarily baroque structures in the back end that are maintained by bots. Probably we will see some simplification over the next few years which will reduce or eliminate some of these tasks. An overhaul of the category system is long overdue, for example.

All the best: Rich Farmbrough, 23:42, 19 January 2017 (UTC).

The reason for more bots (and more semi-automation) is simple: We have more articles, we expect more from our articles, and because editorship has dropped, having bots around means humans don't have to perform tedious edits so they can focus on edits requiring actual thought. It may be true that the FA back then didn't require as much 'fiddly markup', but the quality (and reader-friendliness) of an FA now is also substantial higher than an FA then. Headbomb {talk / contribs / physics / books} 16:24, 3 February 2017 (UTC)

Some "general fixes" are controversial

2) The "general fixes" function of the AWB software was intended for uncontroversial fixes to which no reasonable user would object. Wikipedia:AutoWikiBrowser/General fixes has vastly grown in size over the years, and now includes significant numbers of purported "fixes" for which there is little or no community consensus.

Comment by Arbitrators:

Magioladitis Part of the issue, I think, is that people who see a "cosmetic" edit on their watchlist and get annoyed about it are mostly not the same people who have the background to give you specifics about which of those rules to change and how. We probably all differ on what a "reasonable" user would accept even if they personally find something irritating, but it seems like there currently isn't an effective feedback mechanism to control scope creep in the list of general fixes. Opabinia regalis (talk) 07:36, 17 January 2017 (UTC)

Comment by parties:

If you give me specifics I may can help to clean the WP:AWB/GF. -- Magioladitis (talk) 00:14, 17 January 2017 (UTC)

Re 50.0.136.56: Bypassing redirects should be reduced to the minimum required per many discussions and in the future replaced by a better system. I started helping on this direction. -- Magioladitis (talk) 23:41, 29 January 2017 (UTC)

Comment by others:

In my view, this is what much of this case ultimately revolves around. AWB is great when used for a specific task ("search for every example of 'doe snot' and replace each one with 'does not' or skip with a single keystroke each time), but it comes with "apply general fixes" enabled by default. Many of these 'fixes' are entirely arbitrary matters of personal preference, or an inappropriate treatment of the WP:Manual of Style as prescriptive rather than suggestion. A glance at Wikipedia talk:AutoWikiBrowser/General fixes shows that the AWB devs routinely ignore anyone who complains about, or even queries, this approach. ‑ Iridescent 12:03, 7 January 2017 (UTC)

The points about AWB are informative but I think the "case ultimately revolves around" Magioladitis's aggressive approach to using this bot for unapproved tasks, running it from his main account, running instances with known problems/bugs([11] etc), unblocking and continuing to run it without fixing the problems it was blocked for, and his and Yobot's many blocks and ANI's indicating a disconnect between his editing practices and what the community finds acceptable. It's not about what AWB should or shouldn't do--it's about how it's being used. 50.0.136.56 (talk) 19:45, 7 January 2017 (UTC)

This is a useful comment, and perhaps a little more discussion and advertisement of new GFs could be undertaken. All the best: Rich Farmbrough, 21:08, 8 January 2017 (UTC).

I'm leery of this one. "We, the members of ArbCom, hereby declare that there is no consensus for an unspecified amount of stuff on this very long page", with neither examples nor evidence of actual consensus (a couple of people complaining ≠ lack of consensus), sounds a way to start drama, not to stop it. Also, the title and the resolution are somewhat mismatched. "Some of this is controversial" is not the same as "There's no consensus for this". WhatamIdoing (talk) 02:03, 24 January 2017 (UTC)

This sounds ok to me, especially if specific examples are given per WhatamIdoing. I think CBM's observation is right, that "general fixes" implementers took it on themselves to decide stuff unsupported by the community. For example, my own evidence (sorry about how longwinded it came out) is about Yobot repeatedly bypassing a perfectly good template redirect, when the redirected name was used in the article as an editorial choice of a human. If there's something wrong with the redirect that makes someone think it should be bypassed, the place to discuss that is wp:Redirects for discussion, not "AWB general fixes". See also: wp:Redirects are cheap. 50.0.136.56 (talk) 23:34, 29 January 2017 (UTC)

It's not accurate to say "that "general fixes" implementers took it on themselves to decide stuff unsupported by the community." It would be a great deal more accurate to say genfix implementers took it upon themselves to implement things supported by the community as best as they can tell. Mistakes and errors don't imply they unilaterally decide to implement stuff just because it's their personal preference. Headbomb {talk / contribs / physics / books} 13:59, 5 February 2017 (UTC)

Proposed remedies

Note: All remedies that refer to a period of time, for example to a ban of X months or a revert parole of Y months, are to run concurrently unless otherwise stated.

All general fixes to be disabled unless specifically approved

1) All features currently included at Wikipedia:AutoWikiBrowser/General fixes are to be disabled on English Wikipedia, until specific approval is sought and granted for each one, by means of a reasonably publicised Request for Comment. No further general fixes are to be enabled for AWB on English Wikipedia, unless community consensus is sought and granted for their addition. Editors who customise a version of AWB in order to continue to apply a "general fix" for which community consensus has not been granted, whether on their main or bot accounts, will be treated as unauthorised bots and indefinitely blocked.

Comment by Arbitrators:

I don't see this as out of scope at all and it does appear that it may stop the problem. It might take a long time to deal with this setback but WP:DEADLINE. Any idea how many bots this would affect? What if we specified it just for Magioladitis? Is there a semi-nuclear option? I would also be interested in hearing your thoughts on the second set of questions I posed above Iridescent. -- Amanda (aka DQ) 02:17, 8 January 2017 (UTC)

(IAR and replying here for ease of reading) I'd say the problem isn't specifically with AWB per se, but with the way AWB is configured. If one switches off the "general fixes" checkbox and only uses it for a specific task, it's a fantastic tool (see this run of edits by me yesterday for instance, and I'll note in passing that I rejected more than 50% of AWB's proposed changes in that batch—part of the problem is that certain people assume that all its suggestions are valid and just click save-save-save, or treat the MOS as compulsory and try to use AWB to enforce it). Unfortunately, the default configuration is that when making any change to a page, every single one of these changes is also made. This makes if very hard for anyone subsequently checking diffs to actually see what's been changed, makes it very hard to preview the change to see if the proposed change is appropriate, and when run as an unsupervised bot (as Magiolatidis has been doing, both on his bot and human account) tends to make a lot of inappropriately trivial changes. Because "general fixes" runs in the background even when making manual edits using AWB, unless one specifically turns it off, in my opinion it should be dealt with as a de facto bot that just happens to piggyback onto theoretically human-made edits—and as per my comments elsewhere, these functions were never approved. I would be interested in how many of Yobot's problematic edits aren't the result of it trying to apply an unapproved general fix—I have no intention of going through every edit looking, but I strongly suspect the figure will be zero. On the specific question of bug fixes, I don't have enough technical knowledge to know if they're being addressed appropriately and effectively, although we have plenty of technical people who should know enough to have an informed opinion. If only there were a huge crowd of people employed by us who are paid to know these things ‑ Iridescent 16:32, 8 January 2017 (UTC)

Per AWB Rules of Use, its users are responsible for all the edits, including genfixes. An AWB user is supposed to look at each and every change, and make the decision to keep it or not. I would go further and say that if an AWB user doesn't understand a particular change, they should skip it until they understand it. I realize that not all AWB users will behave this way, but surely they are supposed to. Thus, if an AWB user is doing as they should, there is not "de facto bot" in operation.

As for making it "hard" for those checking diffs, I run into a lot of manual edits (not using AWB or any tool/script) that are hard for me to check, so I honestly can't say we should apply this bias here. If anything, AWB's genfixes are generally reliable cleanups, for even if some of them go too far lint-wise, there is no real damage. And from 2 1/2 years of my own AWB usage, I have found genfixes to be generally conservative in approach compared to tons of issues it doesn't fix that I have written Find/Replace methods to somewhat deal with. On that note, I wonder if what some see as genfixes are really an AWB user's find/replace's + genfixes, and not knowing the difference. Stevie is the man! ^{Talk • Work} 19:33, 8 January 2017 (UTC)

This may be too much, but it does sound like a great deal of the current problem would be solved with a more careful change management process for "general fixes" and for the checkwiki project. Opabinia regalis (talk) 07:31, 17 January 2017 (UTC)

Comment by parties:

I think after the latest bug fixes, no AWB general fixes contain controversial fixes. -- Magioladitis (talk) 08:06, 24 January 2017 (UTC)

Comment by others:

This is a nuclear option, but unless and until the nettle of AWB's scope creep is grasped, any remedies will be akin to fighting a forest fire by focusing on individual trees. While everyone involved has worked in utmost good faith, it needs to be recognised that AWB has metastasised from its original purpose as a tool for making multiple similar edits quickly while still retaining human input, into a whitespace-removing dash-replacing comma-adding reference-moving link-replacing behemoth; that consensus was never sought or granted for this change; and that these changes are often contentious with large sections of the Wikipedia community but are presented as uncontroversial. Although it's not a bot per se, AWB is bot-like enough that it needs to be brought into line with Wikipedia's standard approach to bot editing, and to operate on the principle that any task which has not been specifically approved should be explicitly assumed to be unapproved. At a bare minimum, the default setting for AWB should be "do not apply general fixes" with each fix manually enabled by the user checking a box, such that each user is explicitly taking responsibility for any problems their editing may cause. ‑ Iridescent 12:03, 7 January 2017 (UTC)

This sounds like a good idea to me, though I can understand if it's hard to pass as a remedy in this supposedly narrow arb case. 50.0.136.56 (talk) 21:00, 7 January 2017 (UTC)

IMO limiting the case strictly to "narrow scope" misses the point. Whatever problems Magioladitis and Yobot have been causing are ultimately symptoms rather than the disease. Every accusation being levelled is ultimately some variant of "AWB's general fixes can be inappropriate and don't have consensus"; until that is addressed any sanctions will just be a case of cutting down the tallest poppy while leaving the rest of the field to grow. (I could make a case that since Magiolatidis is one of authors of AWB, and AFAIK the person responsible for "general fixes" in the first place, this remedy would fall into even a narrowly defined scope.) ‑ Iridescent 21:25, 7 January 2017 (UTC)

Well, I'm supportive of the idea as I said. Maybe it can be proposed elsewhere if it doesn't pass here. But I don't agree that the accusations in this arb case are about what AWB does. They're about "Magioladitis insists on keeping running the damn bot after multiple people ask him to stop". That's always a problem no matter what the bot is doing. 50.0.136.56 (talk) 22:33, 7 January 2017 (UTC)

Yeah, but AFAICS every time someone has asked him to stop, it's because he's using to bot to enforce a "general fix" for which there's no consensus. If the bot were just fixing instances of people using semicolons to create bold subheadings, correcting misspellings of the word "parallel", or any of the other entirely legitimate uses of AWB, then we wouldn't be here. (I wouldn't be sorry if the same "is there consensus for this?" nuke-and-rebuild approach was taken towards the AWB typo list as well—a significant chunk of that is not by any possibly measure "typos" but assorted stylistic preferences about dashes, commas and hyphens being slipped through under the radar.) ‑ Iridescent 16:46, 8 January 2017 (UTC)

I support a thorough review but not upfront disabling of the whole of genfixes. If we really want to get basic, the issue is MOS creep/fuzziness, leading to genfixes creep and fixes of some maybe unimportant "problems". Genfixes development is ultimately fixing technical problems and aligning to the MOS, and doing as much of this as can be developed. In the end, there is not an unacceptable level of damage from the genfixes themselves (from long use of AWB, I have found the fixes in general to be pretty conservative, and the "damage" really mostly editor complaints of bots "covering up" vandalism stemming from a decade-old Mediawiki watchlist bug), so the fixes can reasonably proceed until any individual types are ruled out. If genfixes is entirely disabled upfront (and who knows how long that will last), I may reconsider my involvement with the Wikipedia, as I don't feel comfortable with doing volunteer work for an effort that takes sledgehammer approaches. Stevie is the man! ^{Talk • Work} 19:06, 8 January 2017 (UTC)

I think you're missing the point here. Nobody should be using bots or bot-like processes to enforce the MOS, which is very explicitly a list of suggestions and not prescriptive (every proposal in the history of Wikipedia to elevate the status of MOS has been shot down in flames). If all genfixes are doing is enforcing the MOS, then there is not only no consensus for them, the existing consensus is explicitly against them. There may well be consensus to make the MOS prescriptive and to automate its enforcement—consensus does change—but nobody has yet demonstrated it. ‑ Iridescent 19:16, 8 January 2017 (UTC)

Semiautomated editing is not "bot-like" if one is reviewing all the edits, so it is not an "enforcement" -- it is that user's decision to edit and bring an article up to MOS snuff, not critically different than if they were making the edits manually. I object to calling AWB users "bot-like" as a stereotype. People get confused with the 'A' part of AWB, but what is automated is 1) generating and running through a list of articles; 2) suggested changes from genfixes, Find/Replace's the user themselves set up, typos, etc. What is not automated is the user's decision on what changes to save. So, I disagree with the premise of your remarks as stated in their application here. You assert a lack of consensus that doesn't apply to an individual editor's editing. Just because that individual is using a tool to assist doesn't make it special, because again, that editor is deciding what to save. If I decide to fix an article's MOS deficiencies, I can decide to do that. Stevie is the man! ^{Talk • Work} 19:47, 8 January 2017 (UTC)

(edit conflict) Iridescent-- in the single diff that brought me here, Yobot changed (among other things) all the {{cite}} templates in an article to {{citation}}, which is not supposed to be done per WP:CITEVAR. And I sometimes use semicolons for bold subheadings: in fact I just did it in a set of answers to Amanda's questions that I'm still writing. It's not something that I'd want a bot to mess with.

It's possible you're right that this can be solved by unchecking some boxes on AWB--I can't scientifically disprove it, but many years of seeing these disputes makes me pessimistic. When someone is hell-bent to run a bot, they usually get the knee-jerk support of the bot community no matter how much disruption the bot causes, so there's long-term drama.

Stevietheman: in principle AWB users are supposed to review all edits, but in practice they often don't (i.e. they use "manual" AWB editing as a way to bypass BRFA), so that may have brought AWB into some disrepute. And even if each edit is reviewed, I'd still consider them bot-like, but that's more controversial. 50.0.136.56 (talk) 19:59, 8 January 2017 (UTC)

Re: your comment to me, if anyone is using AWB and causing repetitive damage they don't address, then a final remedy is removing their access. Also, AWB assists with going through a list of articles, so in that respect, it is automated, but again, they are making a decision to save, so it's not a bot, but instead a human being technically aided. Think of AWB as computer-aided manual editing -- that's how I use it. I test for many things in my cleanups, and I suppose someone may think I should bot-ize them, but the problem with that is not everything can be bot-ized. Most of the changes I choose to make are based on issues I see in articles as a manual editor, but to make my work more efficient, I use AWB to help me find and fix them in sets of articles I care about (like those included in a wikiproject to which I belong). And a lot of my find/replace's require manual review, because they are not absolutely perfect due to peculiarities in specific articles or deficient regex. Perhaps a subset of my F/R's are bot-worthy, but then, I'm also not seeking to run a general bot to work all across the site. I am wanting to improve particular sets of articles with mostly non-bot-worthy, ultimately manual cleanups, being aided by a tool. Stevie is the man! ^{Talk • Work} 20:17, 8 January 2017 (UTC)

I understand what you're getting at: we're just at different points on a spectrum (that's why I said it was controversial). I'm also coming more from a DR perspective: what to do when a fast, repetitive editing plan gets disruptive. I take a broad view of WP:MEATBOT, which says if something quacks like a bot, it can be treated like one even if it's actually done 100% manually. So if someone editing like that encounters resistance from other users, they should stop and open a discussion. What you're doing with AWB sounds fine, and if nobody is contesting it, there's no problem. 50.0.136.56 (talk) 02:20, 9 January 2017 (UTC)

"Yobot changed...all the {{cite}} templates in an article to {{citation}}, which is not supposed to be done per WP:CITEVAR."

Speaking as the primary author of the guideline you're invoking, let me say: You are wrong. There is nothing in CITEVAR that requires you to use the redirect {{Cite Journal}} rather than {{Cite journal}}, or to use the redirect {{Cite}} rather than {{Citation}}. CITEVAR does require you to stick with CS1 templates rather than switching to CS2 templates (unless a change is discussed and agreed upon). This means that switching from {{cite}} to {{cite web}} is not okay, but that's not what happened. WhatamIdoing (talk) 02:16, 24 January 2017 (UTC)

The whole point of "General Fixes" is to allow as many as possible of those annoying little items to be done in the same edit. It is designed to alleviate the very problems which some in this Arb case complains about: notably "cosmeticbot" changes.

Claims that the devs do not respond to requests for change are groundless: Notably

1. The conversion of "date=1999" to "year=1999" has been stopped.
2. The insertion of blank lines between different level headings has been stopped.

Part of the reason that MoS is kept as a guideline is that we welcome people who write in bad or broken English, use tautologies and misspell or mistype words. However we also welcome those who fix these solecisms - and would do well to remember , as a community, that there are far more people who wish to vomit their content into Wikipedia, than are prepared to spend the effort bringing that content up to our basic requirements.

All the best: Rich Farmbrough, 20:51, 8 January 2017 (UTC).

Bringing a substandard article up to standards requires actually engaging with the content, rewriting for clarity, researching sources, that sort of thing. People who do that sort of rewriting usually also clean up typos etc. (I do that sometimes). Using a bot to make drive-by MoS changes is of almost no help in improving those articles. MOS fanaticism is another massive source of useless disruption on Wikipedia, just like obsessive bot editing. So MOS wanking is a completely unconvincing rationale for calling the bot operations important. Taking care of two birds with one stone sounds good to me. 50.0.136.56 (talk) 01:10, 9 January 2017 (UTC)

Speaking for myself I was looking forward to using my bot to check and add references, remove solecisms, and add and organise content. I may still do some of those things, but five years of grief did dull my appetite somewhat. All the best: Rich Farmbrough, 22:04, 9 January 2017 (UTC).

It is troubling that Mag being one of the primary developers/feature adders for AWB and the semi-opaque nature of the GeneralFixes implementation. I agree with this proposal and suggest that in the future the General fixes should have a public on-wiki log page for each "AWB editor" that is written to enumerating which Fixes are going to be applied to a page and then inserting that revision into the edit summary for the article being edited so that we have a crystal clear idea as to which rules are causing the most problems as we currently do not have the appropriate logs to prevent the same perceived errors occurring. Hasteur (talk) 13:15, 9 January 2017 (UTC)

DeltaQuad: this remedy does seem out of scope to me, as you have decided to restrict the scope to the actions of Magioladitis and Yobot, so the evidence being collected at Wikipedia:Arbitration/Requests/Case/Magioladitis/Evidence only relates to the actions of Magioladitis and Yobot. I could support such a remedy if there was evidence that other users/bots were causing disruption because of the genfixes. However in my experience this is not the case. — Martin (MSGJ · talk) 17:21, 10 January 2017 (UTC)

I often find small robo-fixes annoying though that stops short of "disruptive". Their presence from that perspective is just another of the million annoying Wikipedia aspects that editors have to deal with. I'm sure some other editors have the same view. That's out of scope of the case per se, but it's part of the background that should be taken into consideration. 50.0.136.56 (talk) 23:42, 18 January 2017 (UTC)

If these things annoy you, then I suggest hiding bots from your watchlist. Or if specific bots annoy you, see WP:HIDEBOTS. Headbomb {talk / contribs / physics / books} 13:49, 5 February 2017 (UTC)

This would throw the baby out with the bathwater, DeltaQuad. I actually have multiple tasks that rely wholly on general fixes to clean up formatting after completing a non-cosmetic task (i.e. template orphaning, etc). That's entirely compliant with the bot policy, and I transparently state that I'll be using general fixes in each BRFA where that is the case. This is a problem of behavior, not technical issues. If a bot operator cares to apply effort to ensure they'll only be editing pages where a non-cosmetic task is being completed, they can. For instance, AWB has a "Skip genfixes only" option, which I have checked. Perhaps that option should be required unless explicitly granted authorization to uncheck it at BRFA, instead? That would be more helpful. Banning genfixes entirely would break many AWB bots which have had zero issues. ~ Rob₁₃^Talk 07:07, 24 January 2017 (UTC)

Let me join others in saying this is way overkill and would harm Wikipedia more than anything else. Restricting Magioladitis from applying genfixes is one thing, but the entire community? Something like Wikipedia:Arbitration/Requests/Case/Magioladitis/Workshop#Proposal_by_User:Headbomb would be way more measured. Headbomb {talk / contribs / physics / books} 13:40, 5 February 2017 (UTC)

Magiolatidis to seek broad consensus for proposals

2) Any proposed bot tasks for Yobot or any other bot operated by Magiolatidis must be publicised centrally, must remain open for at least a week, and may not be approved without the agreement of [insert number] editors.

Comment by Arbitrators:

Comment by parties:

The issue with this is that bypassing template redirects and performing whitespace-only edits was never approved by BAG. I also disagree that other participants can foresee cosmetic editing. Only Magioladitis knows how he runs his bot, and he can change its operation at any time. BAG/other participants can only approve based on promises. The only way to deal with the problem that you are describing is WP:BOTISSUE, which instructs administrators to block bots that "operate without approval, operate in a manner not specified in their approval request, or operate counter to the terms of their approval or bot usage policy." Ramaksoud2000 ^{(Talk to me)} 17:57, 8 January 2017 (UTC)

In addition, this doesn't touch on the significant unapproved bot editing from his main account. Ramaksoud2000 ^{(Talk to me)} 18:19, 8 January 2017 (UTC)

Take note that the CHECKWIKI tasks were also approved for two more bots (BG19bot, Menobot). The section fixing tasks were also approved for Dexbot and some other tasks were just a continuation of already approved tasks for other bots. The main problem discussed in this secion is the CHECKWIKI tasks I guess in which multiple editors participate already. -- Magioladitis (talk) 22:14, 18 January 2017 (UTC)

Also note that bypassing redirects was never Yobot's primary task. See also the edit history that proves I was mainly removing redirects and not adding. -- Magioladitis (talk) 15:37, 24 January 2017 (UTC)

Comment by others:

Looking over the various BRFAs for Yobot (listed here), there's a consistent pattern of rubber-stamping of the BRFA on the basis of Magiolatidis himself saying that the edits won't be controversial or cause errors. (I am not implying any fault on the part of the BRFA members, who had no reason to doubt him.) Forcing the BRFA to remain open until a minimum number of participants (I'm thinking ten) have agreed that the task is legitimate and non-problematic won't cause any hardship—if the BRFA takes a month to close, that just means the task runs a month later than it would have otherwise—but will prevent potentially contentious tasks being slipped through, and is also greatly to Magiolatidis's benefit if the bot does then cause problems as he will quite reasonably be able to say that if none of the other participants spotted the issue, it's unfair to accuse him of being sloppy. ‑ Iridescent 17:44, 8 January 2017 (UTC)

It's pretty unusual to see 10 people comment on a routine BRFA, I think. But if BAG is doing its job, it should be vetting the operations more than it does before approving. Surely it knows about all the hassles Yobot has gotten into. 50.0.136.56 (talk) 20:02, 8 January 2017 (UTC)

It's easy to say "if BAG takes a month to close [so what?]" but I have had one-off AWB solutions ready to go, and wiped out by a power failure before BAG had made a move. Conversely I have has tasks that have been finished manually, before BAG could get their act together to respond. This was some time ago, and BAG has changed a lot since then, but please do not trivialise the time and effort invested by those who run bots.

The requirements of BAG are necessary and sufficient. We (the community) need to BAG to be helpful, incisive and timely. Lading them down with additional bureaucracy will not further these goals.

All the best: Rich Farmbrough, 20:59, 8 January 2017 (UTC).

This type of remedy has failed many times in the past. If you have particular reason to think it will be different this time, I'm interested in hearing them. 50.0.136.56 (talk) 20:16, 8 January 2017 (UTC)

Perhaps a better proposal would be "BAG is reminded to check for consensus". All the best: Rich Farmbrough, 20:52, 8 January 2017 (UTC).

I feel that this proposal (in spirit) has potential. It has been my understanding that as a BotOp, you're supposed to come to BRFA with a solid consensus in hand when you want to do a task. I think BRFA verifying that the consensus is accurate and had included the right potentially effected people (i.e. MOS junkies coming to a consensus that dates will now be represented in Y-D-M format without consulting and securing a broader consensus at VP or AN) before approving. Hasteur (talk) 13:08, 9 January 2017 (UTC)

You might look at this venerable post for how well that's worked in the past. 50.0.136.56 (talk) 09:57, 10 January 2017 (UTC)

Arbcom decisions are not precedent setting on future arbcom actions. I would also note that BAG of 8 years ago is very different than BAG of now. I am now calling the question 50, how are you so familiar with significant arcane portions of WikiPolicy? I think it would be appropriate for you to make declarations regarding previous IPs or editor accounts that you have previously edited under as your familiarity with wiki jargon and old cases suggests (in my mind) that you may be a editor who was indef blocked previously in relation to a bot being blocked. I would note that the edit history on this IP only goes back to October of 2016 which means either you're a new hop of an IP address or you're a editor who has previous history that would do with examination to consider the context of all your proposals. Hasteur (talk) 13:17, 10 January 2017 (UTC)

I find your question completely out of order. Unregistered editors are perfectly welcome to participate and 50.0.136.56's contributions seem well considered and valuable. There are various reasons why an IP addresses would change so it is not surprising that the contributions date back only to October 2016. You would do well to assume good faith in the absence of any evidence to the contrary. (If you decide on reflection to remove your question above, please also remove my response.) — Martin (MSGJ · talk) 17:13, 10 January 2017 (UTC)

I appreciate MSGJ's answer so I've removed my own (original text). 50.0.136.56 (talk) 18:22, 10 January 2017 (UTC)

You MSGJ may find it out of order, however it is reasonable to know who is making proposals that are both inproper decorum (see "Second Law of Robotics" proposal and contributions to both the Evidence and Workshop) and what appears to be working to reduce or trivialize the proceedings, I think it is reasonable to know who it is making said proposals. Hasteur (talk) 19:53, 10 January 2017 (UTC)

The idea on Wikipedia is to comment on the edits not the editor, so feel free to support or oppose the proposals as you see fit. I plan to post a few more soon, fwiw. I don't see any improper decorum particularly in the Second Law proposal. On feedback from other users, I withdrew a proposal that would have been suitable for certain other disputes and was used in an earlier arb decision, but in retrospect was inappropriate for this particular case. I haven't tried to trivialize anything. Referring to Asimov's 2nd law of robotics was a concise way of expressing what I think the relationship between humans and bots on Wikipedia should be. 50.0.136.56 (talk) 22:07, 10 January 2017 (UTC)

As a WP:BAG member, I have to say this seems pretty ridiculous to me. WP:BOTPOL already requires that any bots operate with consensus, and that they go through WP:BRFAs. So that's just restating what already exists. However, requiring they stay open for one week seems like unwarranted process-creep. If a discussion happened on a Wikiproject, gains consensus, coding is done, and a BRFA is filed, then what dictates BRFA length is really a matter of BAG discretion and a function of the bot operator's quickness to respond to issues that may arise in a trial, if such a trial is needed. Requiring a specific number of editors get involved in the discussion is likewise silly, not only because BRFAs are a boring as hell thing to monitor and comment on, but because none of the issues that got us here is due to a flaw in the BRFA process. The only' thing that might make sense is that Magio and his bots have bot trials of at least X edits, or something similar (like X edits for Y days). Headbomb {talk / contribs / physics / books} 14:12, 5 February 2017 (UTC)

Proposals by 50.0.136.56 (in progress)

Proposed principles

Wikipedia is an encyclopedia

1) Wikipedia's most fundamental policy is that Wikipedia is an encyclopedia, a reference work whose intended mode of use is humans reading articles. Many secondary uses (not all good) have emerged since Wikipedia started, but they are epiphenomena and shouldn't be allowed to interfere with the main goal of writing the encyclopedia.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Some people seem to want to use WP as a programming playground or a robot theme park. It isn't those. 50.0.136.56 (talk) 08:22, 9 January 2017 (UTC)

This comment is just WP:NOT. It bores me to tears to have to repeat - WP is not those things, but it can be used as those things provided the editing is consistent with our policies. WP is not for pushing one's agenda, but Women In Red certainly has an agenda. That's fine WIR creates (mostly) balanced articles about notable people. We want that. We would also welcome (as Wikipedians) neutral coverage of notable people from any other demographic. We don't care if you are Team Jacob or Team Edward, as long as the articles are neutral.

Similarly as long as your programs/bots contribute positively and non-disruptively, we are fine with it (indeed we have allowed some crazy stuff which has not been disruptive).

All the best: Rich Farmbrough, 20:47, 11 January 2017 (UTC).

No I don't think so. Bots perform some necessary tasks just like admins do. If a new editor shows up saying their main interest is becoming an admin so they can block people and delete pages, that's a sign that they should change their focus or go someplace else. It's the same thing if someone shows up primarily wanting to run bots. If someone wants to contribute to Wikipedia primarily as a programmer rather than as an editor, they should call the WMF, which will even pay them to do that. If they want to run bots on Wikipedia they should have other interests too. I saw Δ for example as a classic disruptive WP:SPA, and I never understood why nobody else seemed to think of him that way. 50.0.136.56 (talk) 12:16, 18 January 2017 (UTC)

We also need good quality page deleters. All the best: Rich Farmbrough, 23:47, 19 January 2017 (UTC).

Bad proposal, again way too judgmental on the motives and intentions of volunteers. If someone wants to code, we have plenty of bot tasks desired by the community waiting to be picked up. Volunteers coders should be welcomed, not driven away with pitchforks and a we don't want your kind here mentality.Headbomb {talk / contribs / physics / books} 14:19, 5 February 2017 (UTC)

Competence is required

2) Wikipedia, the encyclopedia that anyone can edit, presents opportunities for people of all ages and skill levels to make meaningful contributions in both content and technical areas. The growth and learning that comes from stretching and sometimes exceeding one's capabilities is one of the main personal benefits of participating in Wikipedia, and it also benefits the project. So it is healthy for users to attempt difficult challenges occasionally, even if failure and temporary breakage sometimes results. If something goes wrong, they should step back and re-assess rather than cause more disruption.

Sometimes an editor will cause persistent disruption by continuing operating outside their zone of competence against the advice of other users. Such editors should be asked or required to switch to another area of editing. In severe cases, they can be blocked or banned from the project.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Based on WP:CIR. Could use a less verbose formulation. Competence in bot operation is not just technical. It includes avoiding causing disruption. 50.0.136.56 (talk) 08:22, 9 January 2017 (UTC)

If you seem to commit the same errors (or repeated errors) when BotOperating, your competence and operator credentials should be yanked until you can demonstrate that you have improved or implemented procedures to prevent or reduce the occurance of future errors. Hasteur (talk) 13:18, 9 January 2017 (UTC)

The underlying sentiment seems valid, but the wording is sweeping and self contradictory. If you have merely reduced the occurrence, then you will repeat them,, and be "yanked" again. One needs also to look at the severity and frequency, and give due attention to both regression and special circumstances. All the best: Rich Farmbrough, 01:48, 16 January 2017 (UTC).

Production server

3) en.wikipedia.org (the server that hosts the enwiki web site) is a production server, defined on techopedia.com as follows:[12]

- Definition - What does Production Server mean?

A production server is a type of server that is used to deploy and host live websites or Web applications. It hosts websites and Web applications that have undergone extensive development and testing before they are validated as production ready.

A production server may also be referred to as a live server.

Deploying insufficiently tested or buggy code on a production server, or bypassing established release processes even for correct code, is a serious breach of acceptable programmer or operator conduct.

Comment by Arbitrators:

I hope everyone can agree that on Wikipedia, a bot or script that consistently makes too many mistakes (although we'll never have complete agreement on what constitutes "too many") is a problem, without the technical definition of the server type. Newyorkbrad (talk) 01:45, 12 January 2017 (UTC)

Comment by parties:

Comment by others:

Surprisingly, Wikipedia currently has no article about this concept. Deployment_environment#Production has some info that appears to come from some lame "methodology" document but conveys about the same drift. The Techopedia excerpt is from the first Google hit for the phrase.

I made a similar comment in the post-Betacommand 3 discussion that was seconded by the experienced developer John @Nagle:[13] An active, long-running bot that people rely on is essentially a component of the server, so long-term bot operators should be held to standards like those applied to server operators (similarly with programmers).

I pretty much disagree with this, although I sympathise. The whole point of Wiki is that you can make the type of error that would never be allowed in, say a banking environment, which would be replete with reviews and change controls. It is certainly true that there are areas where making a mistake is more costly, notable Arbitration, Edit Filters, Administration and Bots. Of these the easiest to fix are bot errors. I fixed up the Erik9Bot debacle, it was fairly simple - I even managed to deliver the putative benefits.

In the context of this case it's important to remember that Magioladitis stands accused primarily of making a number of edits that did not improve articles enough to be worthy of the cost of an edit. In the a community that has disputes like Israel-Palestine (and even people who revert edits because they are "cosmetic"...) a relatively small number of such edits is insignificant.

All the best: Rich Farmbrough, 22:14, 9 January 2017 (UTC).

Bot errors can be quite difficult to undo, simply because so many pages are affected. I was going to suggest looking at User talk:75.57.241.57/x (some output from a bot that I wrote to help clean up after someone else's bot), but in a stroke of irony it turns out that page was deleted by someone on a bot rampage a couple of months ago, after minding its business quietly for years. Another bot operation we didn't need. No this isn't a banking system, but it's an enormous live server, and a bot screwup here has more impact than even a 1 day total outage of some of the smaller sites I'm involved with, that would get PHB's yelling on the phone. 50.0.136.56 (talk) 06:19, 10 January 2017 (UTC)

And I believe your ironic bot was one I was against. Perhaps you should join BAG, I known I have considered it, but I am not active enough on those pages these days. All the best: Rich Farmbrough, 20:53, 11 January 2017 (UTC).

I'm delighted to hear that Wikipedia has ever had a bot that you were against. I thought that was impossible. :) 50.0.136.56 (talk)

@Newyorkbrad: the issue I'm trying to get at with this proposal is cowboy coding rather than errors per se. Even error-free unauthorized bot operations should be stopped since they present the fait accompli described in many past arb case principles, plus they take unnecessary risk of errors unless they've had some external validation through BRFA. We end up with a Wikipedia controlled by bot operators rather than by the editing community (m:Power structure#Technocracy). 50.0.136.56 (talk) 23:03, 18 January 2017 (UTC)

Error free unauthorised bots are stopped on sight, because there is more to allowing a bot edit than being error free.

1. Useful edits
2. Non-harmful
3. Consensus
4. Bot flagging
5. Avoiding botophobia

The idea that people>bots has a lot of sense behind it, but one has to be careful. People>Gnomes moves towards fascist territory. People>>bots (i.e. bot concerns can be totally ignored) relegates coders to second-class citizens.

All the best: Rich Farmbrough, 20:41, 21 January 2017 (UTC).

Responsibilities of bot operators

4) Like administrators and other editors in positions of trust, bot operators have a heightened responsibility to the community. Bot operators are expected to respond reasonably to questions or concerns about the operation of their bot. An editor who (even in good faith) misuses automated editing tools such as bots and scripts, or fails to respond appropriately to concerns from the community about their use over a period of time, may lose the privilege of using such tools or may have such privilege restricted.

Comment by Arbitrators:

Comment by parties:

Comment by others:

From Betacommand 2. 50.0.136.56 (talk) 08:22, 9 January 2017 (UTC)

This was a bad principle then, although it may be a good principle in general.

Betacommand needed assistance in dealing with the people who came to his page angrily demanding "Y U delete my picture?". Both WormThatTurned and I suggested that if such assistance could be provided by the community, Betacommand's skills and effort could continue to be utilised.

I don't think this is a parallel case, though I do wonder if his explanations are too lacking in chest-beating and garment rending.

All the best: Rich Farmbrough, 22:25, 9 January 2017 (UTC).

I see some similarities in the contents of discussions in the prior DR links. So there are parallels. I was also upset with Yobot a while back because of an annoying edit it made to an article I was working on. So I saw the case opening and went "oh god more of the same". But after a day or two settling into this case, I see it is nowhere near as severe as Beta's problems or yours. For one thing, it occurred to me earlier that Mag. doesn't seem to be under any formal restrictions, while you and Beta were under a lot of them when the arb case opened. Mag made a number of informal commitments that he doesn't seem to have fulfilled, but I didn't see any of the traditional ANI-comments-evaluation-closure discussions ending with a sanction and I don't see anything at WP:RESTRICT. I haven't gone thru everything to confirm it but it's probably worth entering as a FoF if it's true. It may really mean we jumped the gun in opening this case.

Betacommand was hopeless, totally hopeless. Everything that could possibly have worked had already been tried. The amount of good editor effort he burnt was ridiculous, and of far more value than anything he could have contributed. The only reason he lasted as long as he did was that he was supported by some battleground editors/admins who were using him to fight a proxy war about NFCC images.

The principle above (bot operation is a special privilege) applied as much to Beta as to anyone else. 50.0.136.56 (talk) 07:33, 10 January 2017 (UTC)

I wrote a FoF as mentioned above. 50.0.136.56 (talk) 09:47, 10 January 2017 (UTC)

If anything, the wording should mirror WP:BOTCOMM, or at least mention WP:BOTCOMM. Headbomb {talk / contribs / physics / books} 14:35, 3 February 2017 (UTC)

Arbitration in dispute resolution

5) A request for arbitration is the last step of dispute resolution on Wikipedia. With limited exceptions (such as emergency situations, "unusually divisive disputes among administrators", and matters directly referred by Jimbo Wales), it is expected that other avenues of dispute resolution will have been exhausted before an arbitration case is filed. Arbitration is the last resort for conflicts, rather than the first.

Comment by Arbitrators:

Comment by parties:

Comment by others:

From A Man in Black (2009). Maybe there's another case with a better formulation. There are long-term disputes and meta-disputes in Wikipedia that are semi-intractable because they involve genuinely important issues (real-world political conflicts, blocking of major content contributors, etc). There has to be an ongoing balancing act over those rivers of contention in the community.

And then there are huge raging disputes over inconsequential nonsense, often involving the Manual of Style,[14] or nowadays involving bots. That's what we have here. Arbitration remedies in such cases should seek to end the dispute decisively.

Huge raging dispute - I don't think so.

All the best: Rich Farmbrough, 21:26, 20 January 2017 (UTC).

Obsessive point of view (withdrawn)

6) In certain cases a Wikipedia editor will tendentiously focus their attention in an obsessive way. Such users may be banned from editing in the affected area if it becomes disruptive.

Comment by Arbitrators:

There are better and less judgmental ways of making this point. Newyorkbrad (talk) 22:38, 9 January 2017 (UTC)

Comment by parties:

Comment by others:

From EffK (2005). I've always liked this one. 50.0.136.56 (talk) 08:22, 9 January 2017 (UTC)

I'm not sure that's relevant here. Regardless of the letter of the ruling, in spirit it was clearly aimed at those editors who go nuts regarding a pet topic but are fine everywhere else (generally political or nationalist POV-pushers who need to be kept out of their pet issue, but are fine otherwise). Expanding it to cover the general topic of "bot edits" is draconian, and the dispute at hand has certainly not reached the level where WP:DIGWUREN#At wit's end kicks in. If Yobot stuck only to tasks with explicit community approval, and stopped whenever it was challenged, this case wouldn't be happening—we shouldn't extrapolate from "don't bit off more than you can chew" to "never eat anything just in case you choke". ‑ Iridescent 08:42, 9 January 2017 (UTC)

Truthfully I was thinking of that principle about RF, Δ, Kumioko and others. It really seemed to apply to them, especially Δ (Kumioko was maybe more obsessive about non-bot stuff than he was about bots). I agree it might be overstating things in this case. 50.0.136.56 (talk) 09:09, 9 January 2017 (UTC)

This is rather putting the cart before the horse. It is widely agreed that, for example, we have a section entitled "External links" not "Extrenal links". Someone who fixes all occurrences of such a mis-spelling on Wikipedia may or may not be obsessive. If someone else who is obsessive, or litigious, or just plain awkward (and we have plenty of all three categories on Wikipedia) objects, that does not make the first party "tendentious", rather the second party.

All the best: Rich Farmbrough, 22:33, 9 January 2017 (UTC).

Thanks, Iridescent, Rich, and Brad. I think "obsessive" really did apply to editors like Kumioko, but I now agree that it's going overboard here. I've withdrawn the proposal. I still think there are other cases where it should be used, like some of the MOS battles. Iridescent: I think of Rich F as someone with (cough) let's say excessive zeal for bot operations, but who is fine everywhere else; enough that I wish we'd kept him on as an administrator even though I'm glad we stopped his bots. 50.0.136.56 (talk) 22:55, 9 January 2017 (UTC)

FemtoBot runs every day (if I remember). But what stopped me was the tedious officialdom at BRFA, caused by the chilling effect of the now (Vacated? cancelled? whatever..) case. It's quicker just to do most things manually, at least for now. All the best: Rich Farmbrough, 23:24, 9 January 2017 (UTC).

Alternate bot operators

7) Important bots should have backup operators in case something happens to the main operator.[15] A problem with the main operator of an important bot can then be handled by turning the bot over to a backup. Unimportant bots (by definition) can be shut down without major consequences.

Added per ϢereSpielChequers (03:05, 10 January 2017 (UTC)): Bot software released under a free software license can be treated as implicitly having backup operators, if the software is sufficiently documented that new operators can pick it up and run it without the transition incurring an excessive service interruption.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Needs better phrasing and title. 50.0.136.56 (talk) 08:22, 9 January 2017 (UTC)

Important and time sensitive are not necessarily the same thing. A bot can be important and still be something we can live without for a weekend. Time critical important bots need a backup operator. Other important bots need to have source code published under an open license so that when the operator retires someone active at the time can adopt them. There is no great benefit to continually identify a backup operator who may not need to step in for years. ϢereSpielChequers 12:08, 9 January 2017 (UTC)

I have a horror of the "open movement" becoming as dictatorial as everyone else. That is one of the reasons that I have not used WMFlabs - the insistence that "YOU WILL PUBLISH YOUR SOURCE CODE UNDER OPEN LICENSE!"

Yeah, ask nicely and I might share.

All the best: Rich Farmbrough, 22:39, 9 January 2017 (UTC).

This is something Magioladitis has encouraged. I'm sure diffs could be provided, but he has certainly asked me to take over or back up a number of tasks, and has encouraged others to do so. — Preceding unsigned comment added by Rich Farmbrough (talk • contribs) 22:39, 9 January 2017 (UTC)

Rich seems to be saying bot operators should be treated as special perfumed princes/princesses of this supposedly free project. Content editors have to release what they write under CC/GFDL licenses, server programmers have to release code under GPL2, but bot operators can run important project infrastructure while keeping the code to themselves? Bad, bad idea in my book. One of the reasons I wrote this principle is the recurring wail I heard in multiple incidents involving both Δ and Rich, that "we can't block this guy because we neeeeeeeeeeeeed his bot too much!". We should never allow ourselves to be in that position. 50.0.136.56 (talk) 23:05, 9 January 2017 (UTC)

I think you misunderstand. They needed people to run the bots. None of my bot tasks have been particularly complex, and Anomie forked the main one with little difficulty (and my blessing).

But as I remarked elsewhere, with the exception of March 2011, the requests for bot work outstrip supply. And a lot of what I did was not bot work, for example creating the template {{Monthly cleanup category}}, or the infrastructure for {{Lang}}.

It was said at the time "Oh someone else will do it" - well, apart from Anomie, no one else did. I have only just got a few of those things back to normal.

All the best: Rich Farmbrough, 00:00, 10 January 2017 (UTC).

If nobody else thought it was worth doing, that tells me it didn't need to be done. BOTR's do go unanswered but that's usually because they're bad ideas (I'd oppose most of them at BRFA if I hung around there). When someone posts a BOTR that's actually sane, the response is typically like this. 50.0.136.56 (talk) 03:10, 10 January 2017 (UTC)

With that reasoning there would be no Wikipedia. Or pretty much anything. All the best: Rich Farmbrough, 19:59, 10 January 2017 (UTC).

Another reason for bot code release is that one of the supposedly fundamental rights of Wikipedia users is to fork the project. That means they have to get all the content, all the pictures, and all the code, including the bots (at least any that the project uses on an ongoing basis). We are failing part of our mission by not requiring that. 50.0.136.56 (talk) 23:08, 9 January 2017 (UTC)

You will perhaps get free servers donated by WMF? And a cadre of editors to get you started? No, I see the usurpation of Toolserver as part of the hegomonization of the movement by the WMF. All the best: Rich Farmbrough, 23:45, 9 January 2017 (UTC).

The idea of forking is you run your own server. Fred Bauder did that (Wikinfo) for quite a long time. Citizendium is sort of a fork. Veropedia is dead for good reason, but iirc it was self-operated by a few volunteers. The most expensive resource you need to run a full-on fork is the disk space to hold all the Commons pictures, but that's within the reach of a dedicated hobbyist, or almost any institution such as a small school. The rest can be handled by a relatively powerful but not exotic personal computer these days. 50.0.136.56 (talk) 03:21, 10 January 2017 (UTC)

WereSpielChequers--good point, I'll see if I can adjust the wording. I'd consider any open-licensed bot code to implicitly have a backup operator if it's well enough documented that someone else can pick it up and run it if needed. 50.0.136.56 (talk) 23:18, 9 January 2017 (UTC)

I added a sentence to handle this. 50.0.136.56 (talk) 03:23, 10 January 2017 (UTC)

Thanks but the other half of my comment was that important and time critical are not the same. An important but not time critical bot can shut down for a period of time. I wouldn't want to lower our guard against vandalism for a minute, but if a copyright detection bot was down for a few hours would that be a problem? There is already a time delay on that one. ϢereSpielChequers 08:59, 11 January 2017 (UTC)

"...without incurring an excessive service interruption" was supposed to handle that--"excessive" would depend on the task requirements. Do you want something more specific? Feel free to suggest wording. I don't think a principle like this can be too fine-grainedly prescriptive though. 50.0.136.56 (talk) 07:14, 12 January 2017 (UTC)

Not only is general principles about how bots should operate outside the scope of this case unless explicitly related to the actions of Magioladitis and Yobot, but they're also outside the scope of the Arbitration Policy, which forbids the Committee from making judgements about what should be policy. There is currently no policy or guideline requiring alternate bot operators or open source code, and there's a large faction of the bot operating community that would oppose such measures as prohibitive to running bots, which are already difficult to run. ~ Rob₁₃^Talk 23:48, 16 January 2017 (UTC)

I don't know about something like this going through to the final decision, but IMHO it adds some context to the discussion. More context was set in the preliminary statements, which ranged pretty wide. The idea is that disputes are decided against a backdrop of the community's concerns (and the arbs are elected partly for their skill in discerning that backdrop), so it's useful to present context that might not be obvious.

Also this is related to Yobot because of the recurring question "what happens about this bot if we stop its operator from running it?". That came up regarding Yobot. The proposed principle is to make sure that we should never have to worry about this. If a bot is important we should have measures already in place to be sure someone else can run it. If the bot is not important then we can do without it. So we shouldn't let bot operators drag us around. We have WP:OWN for articles and we should have something similar for bots.

I'm not surprised lots of bot operators would oppose this, but they are a self-interested faction who shouldn't be allowed to dominate our decision-making. If they want to take their bots and go home, let them. We are not facing a bot shortage but rather the opposite. I don't sympathize with the sense of entitlement that bot operators alone are exempt from Wikipedia's free-culture principles. Server programmers, template programmers, content writers, and other contributors of every stripe have to abide by them: why are bot operators special? 50.0.136.56 (talk) 11:55, 18 January 2017 (UTC) De-snarked some, 50.0.136.56 (talk) 19:15, 18 January 2017 (UTC)

Actually, we are facing a severe shortage of bot operators. There are many tasks that I'd love to see done, but there aren't sufficient editors with programming knowledge to make them happen. For instance, we need a tool to convert CSV files into JSON files readable on Wikimedia Commons to support their Data: namespace, something that would allow us to create templates on enwiki that automatically update with new data as it becomes available. Hugely useful, and yet I'm unaware of any bot operator willing to work on it. I'm currently working on multiple accessibility-related tasks, in addition to some tagging/stub categorization stuff that I use my bot for regularly. CfD is going to grind to a halt whenever the couple bots supporting WP:CFD/W stop being updated; there's already been many instances where the bots failed and the process essentially stopped for over a week at a time. As for why bot operators should be different, it's often a security thing. Many scripts could be easily hijacked to make mass quantities of disruptive changes to articles. If Cluebot's code were public, vandals could evade detection. If my AWB code for accessibility tasks was public, it could be quickly altered to delete information from thousands of infoboxes. Throwing ready-made scripts out into the open where editing a couple strings turns them into vandalbots is unbelievably high-risk. It would be problematic (or impossible, depending on how far we go with open source code) to have to alter code to make it harder to turn it into a vandal bot. ~ Rob₁₃^Talk 16:55, 20 January 2017 (UTC)

I agree, there are reasons not to place the individual pieces of technology in the open - one is that people will edit to prevent the bot working, precrisely the kind of dispute that we want to avoid. But there are other questions too. I don't want anyone judging my coding - they can judge the edits if they wish. And I certainly don't want to be bullied into giving my code away, nor do I want anyone else to be - both these are reasons for not sharing (most of) my code. I have considered code escrow.

As for CfD, if there's a problem shout me and I'll sort it out - manually. All the best: Rich Farmbrough, 20:53, 20 January 2017 (UTC).

This is a matter of WP:BOTPOL and would categorically object to ARBCOM unilaterally deciding who and how bots may operate on Wikipedia. Headbomb {talk / contribs / physics / books} 14:25, 5 February 2017 (UTC)

Second law of robotics

8) If the operation of a bot (including a BRFA-approved one) encounters significant opposition from human users editors, its operation should stop until a discussion concludes that it can be restarted.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Title inspired by the Three Laws of Robotics. 50.0.136.56 (talk) 01:09, 10 January 2017 (UTC)

Clearly you mean WP:BOTAPPROVAL and WP:CONSENSUS. An Alternative (and more sober) formulation might be Bots are authorized to do specific tasks based on consensus approval from the community. If in the operation of a task, the bot operator encounters non-trivial opposition, the task should be suspended until consensus can be evaluated for the continuation of the task as consensus can change. This indicates where the authority comes from and gives clear indication what the operator must do. Hasteur (talk) 01:21, 10 January 2017 (UTC)

No, I'm saying 1) BRFA isn't a community consensus because most human editors don't participate; and 2) lots of bot incidents happen because people think the bot is doing something outside its authorization. When that happens, it shouldn't take a big bureaucratic process to get the bot stopped. The bot should stop first, and the resumption discussion can happen afterwards. 50.0.136.56 (talk) 01:43, 10 January 2017 (UTC)

It could be worded more elegantly, but I support the principle. When Morwen wrote WP:AGF, she had humans in mind not code snippets. In the case of complaints against bots, the complaint should be presumed valid by default and the onus on the bot operator to convince the community either that the complaint is baseless or the bug in question has been fixed. ‑ Iridescent 10:46, 10 January 2017 (UTC)

How many times should it stop? I have seen human editors derailed for years by repeated opposition after they gained consensus. All the best: Rich Farmbrough, 20:03, 10 January 2017 (UTC).

Remember vandals are human too. Consensus is not unanimity and not all humans are part of our community, but around 500 million are our users. "Community" or "human editors" would be better than "human users". ϢereSpielChequers 09:06, 11 January 2017 (UTC)

Changed to "editors" per WSC. 50.0.136.56 (talk) 07:18, 12 January 2017 (UTC)

Just thinking about the edge cases, further to my note above. Anti-vandal bots probably have more human editors against them than for them. All the best: Rich Farmbrough, 01:53, 16 January 2017 (UTC).

You mean human vandals? I'm a long-term IP editor, I get scrutinized by antivandal bots more than registered editors do, and I haven't had too much trouble with them. Cluebot in particular is a work of brilliance. I get much more frustrated with brainless edit filters, bots like XLinkbot trying to impose content decisions when it has no clue what the link actually points to, etc. 50.0.136.56 (talk) 11:39, 18 January 2017 (UTC)

Hasteur's wording is much better. Headbomb {talk / contribs / physics / books} 14:29, 5 February 2017 (UTC)

Proposed findings of fact

Difficulty of managing bot-related disputes

1) Many seemingly workable remedies have been imposed on bot-related disputes in the past, only to see the disruption continue. (̣Δ 1-3, ARBRF, Date delinking).

Comment by Arbitrators:

Comment by parties:

Comment by others:

Want to expand. Point is that nuanced remedies in bot cases fall apart due to unbelievable stubbornness and wikilawyering from the participants and their supporters. RF was under multiple failed community restrictions before ARBRF, Δ had his own department at WP:AN. The perfect storm is a bot dispute that also involves the Manual of Style. The Date delinking case may have been the worst of those. Arbitrators not familiar with it should probably take a look. It's the next best thing to a bad acid trip. 50.0.136.56 (talk) 08:22, 9 January 2017 (UTC)

Point is that when simple solutions are proposed people turn them down. I have already mentioned the solution to the BetaCommand situation that was ignored. You may look at the workshop in ARBRF to see a proposal that would have addressed all the stated issues, which was roundly ignored. And date delinking was such a stupid conclusion that it chilled any work in that area for a very long time, even where community consensus was clearly demonstrated. The conclusion there should have been, "follow community consensus".

All the best: Rich Farmbrough, 22:48, 9 January 2017 (UTC).

Your suggestion was that Beta receive "community assistance" (front people) for dealing with criticism, but he had already had stupendous amounts of assistance and everything else. He was, at least at that time, completely hopeless as a bot operator. There were some proposals in BC3 to let him keep editing but not run bots, but they were opposed by enough bot enthusiasts (including yourself) to not pass, so he got banned. In retrospect that was the only way it could ever end. (Also, this is Wikipedia, and we don't have minions here).

Any decision should at least tacitly be made on a cost-benefit balance. Part of the cost of any big editing decision on Wikipedia is the dispute resolution workload and other follow-on effects. We have enough experience with BLP's to know how much conflict controversial ones of minor figures cause. We've institutionally decided on encyclopedic grounds to have them anyway, so we get what we expected.

With Date delinking and similar stuff, we said ok to something that seemed to have some small but nonzero benefit and not much cost, never dreaming of the amount of hassle that would ensue. So we got insane drama that we didn't sign up for. It was not part of the original consensus but it turned out to be part of the total package. So it was wise of arbcom to decide "this isn't what we wanted, we have no chance of getting just the good parts, so take the whole thing back and go away please" or whatever. (I actually don't remember how that case came out-- just what a surreal horror show the evidence section was). 50.0.136.56 (talk) 04:13, 10 January 2017 (UTC)

The proposal that I opposed was "Δ is prohibited from editing more than 3 separate articles and their accompanying talk pages in any 7 day period." This neither prohibited him from using a bot to notify users that their possibly non-free content had been so tagged, nor was it a reasonable set of numbers to circumscribe manual editing. One can only imagine the strait-jacket such a restriction would be. All the best: Rich Farmbrough, 23:25, 17 January 2017 (UTC).

IIRC, Δ was under a long-standing ban from anything to do with NFCC enforcement, so he wouldn't have had any reason to leave those notifications. The 3 pages/week was changed to something like 20/day after some discussion. The proposal (with some variations) got considerable support, and its opposition came mostly from proponents of letting Δ continue editing with bots, which years of drama had shown was a terrible idea. So Δ got sitebanned. 50.0.136.56 (talk) 11:02, 18 January 2017 (UTC)

Certainly with different numbers the proposal might have gained traction. Creating a stub and de-orphaning it would have put him over the limit. All the best: Rich Farmbrough, 20:47, 21 January 2017 (UTC).

Robot uprising (TBD)

2) According to Rich, half the edits in mainspace/talk space since 2009 are by AWB. The mind wobbles. I want to get some chronological statistics to put here, so this is a placeholder.

Comment by Arbitrators:

Comment by parties:

According to stats collected by AWB itself since 2009 more than 108 million edits have been done with AWB. See: Marios Magioladitis The Code behind AutoWikiBrowser, 2016.

@50.0.136.56: No idea what lietuval.lt was/is. AWB can be customised to be used by any Wikipedia project, any sister project, any wikia site. aWB also contains customisations in various languages an projects. On of the things I do in international conferences is to ask people to test AWB in their project. Following the consensus / guidelines in some many projects is very very difficult. -- Magioladitis (talk) 08:14, 10 January 2017 (UTC)

If AWB shuts off suddenly it may not send stats to the server. This has happened to me many times in 2010-2012 due to a bug that could freeze AWB and I had to shut off AWB via Task manager. -- Magioladitis (talk) 11:07, 12 January 2017 (UTC)

On the AWB edits count. It can be inaccurate. Some reasons I can think of: In the past the option to "suppress the AWB in edit summary" was available for en.wp too. Maybe stats not working correctly? The Vietnamese Wikipedia number also seems to be very high. -- Magioladitis (talk) 20:04, 12 January 2017 (UTC)

Comment by others:

It wobbleth indeed. Please feel free to check my figures, there could well be significant errors soemwhere - I had expected something like 15-20%. All the best: Rich Farmbrough, 23:28, 9 January 2017 (UTC).

Rich, do you know where those AWB stats come from? Can AWB edits be reliably identified from edit summaries, or do the stats come from data that AWB itself uploads to a server while it's being used? I have the impression it might be the latter. Yes I've been wanting to do some of the math myself, but it's a dreary prospect. My idea is to randomly sample a few dozen edits from each of a few different years, then manually classify them to identify trends. 50.0.136.56 (talk) 00:19, 10 January 2017 (UTC)

Marios, what is lietuval.it and why did AWB edit it 845M times?! My browser didn't reach it or lietuval.lt ("lietuva" means Lithuania. lietuva.lt is a brochure site about Lithuania and doesn't have an obvious wiki anywhere. Similarly the Vietnamese wiki? I wonder what's going on. 50.0.136.56 (talk) 03:57, 10 January 2017 (UTC)

Rich, wait, I looked again. You said there were 240M total article edits (Main=240M), and there were 120M AWB edits spread across article and talk (AWBTalk+AWBMain=AWBTotal=120M), so AWBTotal=Main/2. But you said AWBTotal was mostly Talk edits, while we actually want to know AWBMain/Main. And we don't have that AWBTotal=(Main+Talk)/2 since we don't know the total Talk edits. I think I can find that number. 50.0.136.56 (talk) 04:58, 10 January 2017 (UTC)

I've recently taken over maintaining User:Katalaveno/TBE It shows over 500 million EN wiki edits since the start of 2009, (510 million Nov 2008 - Nov 2016). OK that's all spaces not just mainspace. In combination with the AWB figure of 109m AWB edits since 2009 that indicates circa 20% of all edits, given the dominance of mainspace in editing and the extensive and uncontroversial use of AWB to distribute newsletters in user space I suspect Rich's 15-20% estimate is broadly right. It certainly isn't true that AWB accounts for 50% of mainspace edits since 2009, not of the 109m figure is correct. I'd also point out that AWB has users like myself who are definitely not using it as a bot. ϢereSpielChequers 09:25, 11 January 2017 (UTC)

Thanks, that's interesting. Would it be hard to break that out by namespace, and by edits that are likely to be automated? Or I could try to do it over the next few days. 50.0.136.56 (talk) 07:29, 12 January 2017 (UTC)

I just grepped a metadata dump (December 1) and got about 32M "AWB" edit summaries out of 695M revisions. So if the AWB server is reporting correct numbers, most of the edits aren't labelled as AWB. Maybe they're identifiable some other way--any idea? 50.0.136.56 (talk) 10:47, 12 January 2017 (UTC)

Magioladitis, if AWB sometimes doesn't send stats to the server, that means the true number of AWB edits is even higher than the one on the stats page. The stats page says 109M but directly counting the # of lines with "AWB" in a metadata dump says 32.5M. So it sounds like a heck of a lot of AWB edits are being made with AWB instances that don't put that string in the edit summary. Any idea what's going on? 50.0.136.56 (talk) 18:53, 12 January 2017 (UTC)

Bot users of AWB have a checkbox for suppressing the addition of this string. Stevie is the man! ^{Talk • Work} 20:03, 12 January 2017 (UTC)

Ah thanks, so at least 70% of AWB edits are fully robotic. Is there a way to identify them? 50.0.136.56 (talk) 00:09, 13 January 2017 (UTC)

And non-bot users get that box. All the best: Rich Farmbrough, 01:57, 16 January 2017 (UTC).

Oh ok, so we have no way to know which edits are AWB's. Ugh. 50.0.136.56 (talk) 19:18, 18 January 2017 (UTC)

The fact that automated programs make many edits is neither surprising nor related to this case. ~ Rob₁₃^Talk 21:28, 20 January 2017 (UTC)

Magioladitis is not under any formal community restrictions(???)

2) While Magioladitis is alleged to have broken various commitments he made during informal discussions with other editors, he is not currently under any formal community restrictions.

Comment by Arbitrators:

Comment by parties:

See Wikipedia:Arbitration/Requests/Case/Magioladitis/Evidence#Previous_restrictions_on_automated.2Fsemi-automated_editing_were_quickly_violated. Previous formal restrictions were quickly violated, and nothing came of it. Ramaksoud2000 ^{(Talk to me)} 22:36, 18 January 2017 (UTC)

Comment by others:

This occurred to me earlier today. I don't see anything about Magioladitis at WP:RESTRICT and I don't remember seeing anything in the "previous DR" links with the usual structure of an AN or ANI discussion being closed with a restriction. I didn't check all the ANI discussions that have arisen, so maybe I missed something. Maybe someone else can confirm.

If there really are none, that's different from RF and Betacommand, who had been under all kinds of failed restrictions before they got here. It means maybe we really did jump the gun in opening this case. Neutrality requires that I mention this here. 50.0.136.56 (talk) 09:17, 10 January 2017 (UTC)

Is the RF and BC method (of trying community restrictions first) necessarily any better? We certainly must try lower-level DR methods without restrictions, but I'm not sure if we're better off with advancing to community restrictions. Perhaps an admonishment from ArbCoom may be better, in some cases, than any frmal sanctions. עוד מישהו Od Mishehu 10:07, 10 January 2017 (UTC)

I don't know. It means there was more formal DR before the other cases, for whatever that's worth; and the repeated breaking of pre-existing sanctions means the other two were more demonstrably recalcitrant coming in. RF's case may have taken too long to reach arbcom. Looking at the amount of past conflict with Mag, some kind of intervention does seem to be needed, whether it's a restriction (maybe as simple as just not unblocking Yobot without the other admin ok'ing it or a discussion), Iridescent's idea of adjusting AWB's default settings, etc. 50.0.136.56 (talk) 17:56, 10 January 2017 (UTC)

Magioladitis was blocked several times, which is a sanction, and it has failed to change his behavior. ~ Rob₁₃^Talk 21:29, 20 January 2017 (UTC)

Template

2) {text of proposed finding of fact}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposed remedies

Note: All remedies that refer to a period of time, for example to a ban of X months or a revert parole of Y months, are to run concurrently unless otherwise stated.

Magioladitis reminded

1) Magioladitis is reminded that persistent bad judgment can be incompatible with status as an administrator. (TBD)

Comment by Arbitrators:

Comment by parties:

Comment by others:

I forget how this is usually phrased. I'll look for the stock formulation later. I'm not supporting or opposing R2K's desysop proposal, but imho there should at least be some kind of wp:trout like the above.— Preceding unsigned comment added by 50.0.136.56 (talk • contribs)

It is very hard to put ones finger on what constitutes bad judgement here. I have not followed the minutiae of Magioladitis' talk page over the years, but it s possible that his responses to interlocutors are sub-optimal. It is also possible that your refer to programming decisions, errors clicking check boxes, or judgement selecting tasks.

It might be more useful to offer proposals for improving, or assisting with, whatever areas are seen as falling short. (Actually, here I am at fault, having been a prolific contributor to BRFAs and the BXX pages in general until five years ago.)

All the best: Rich Farmbrough, 23:37, 9 January 2017 (UTC).

It's a softer version of R2K's desysop proposal. There's a bunch of links under "previous dispute resolution" on the main case page, and R2K's evidence presentation calls out lots of examples specifically. The current Yobot discussion on WT:BRFA gives some general flavor. 50.0.136.56 (talk) 03:27, 10 January 2017 (UTC)

Magioladitis restricted

2) I'll save this for now, in deference to Iridescent.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

3) {text of proposed remedy}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

4) {text of proposed remedy}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposed enforcement

Template

1) {text of proposed enforcement}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) {text of proposed enforcement}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposals by User:Od Mishehu

Proposed principles

Automated and semi-automated tools

1) The use of automated tools (generally called "bots") is highly regulated by policy. Specificly, its actions generally need to have community approval. Semi-automated tools may similarly be considered bots if they are, in fact, doing edits without reasonable operator inspection of each edit.

Comment by Arbitrators:

Comment by parties:

Comment by others:

As a point of order, this is WP:BOTPOL/WP:MEATBOT in a nutshell. Headbomb {talk / contribs / physics / books} 14:34, 5 February 2017 (UTC)

COSMETICBOT

2) A bot should not be operated for the sole purpose of improving the cosmetics of the wikitext, such as changing the spacing where it has no affect on the displayed page, or replacing a template redirect with its target.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Unless there is consensus so to do. All the best: Rich Farmbrough, 23:58, 10 January 2017 (UTC).

This is a longstanding point in the bot policy, which the bot approvals group usually does a good job of enforcing. As Rich Farmbrough says, specific consensus can always override a general rule in special cases, but this does not invalidate the general principle. — Carl (CBM · talk) 13:53, 20 January 2017 (UTC)

I think the intent here is constructive, but it's poorly expressed. "A bot should not be operated for the sole purpose" speaks to the purpose of the bot (not the edit) – which means that if I have a minor 'good' reason for using a bot for a few minutes, then I can do purely cosmetic edits with the bot the rest of the day. The example of "replacing a template redirect with its target" is particularly poor, because that means that would ban the bots that clean up after TFD. Finally, I think that it's missing the point that several people have commented on, which is making cosmetic adjustments in addition to a small "productive" edit.

In general, I think that if ArbCom feels a need to address this issue, they'd be better off saying "The Bot policy says, 'Cosmetic changes (such as many of the AWB general fixes) should be applied only when there is a substantive change to make at the same time.'" – possibly followed with a clearer statement, "So if you're making a substantive change, it's perfectly fine under the community's current policy to make thousands of GenFixes in the same edit" (and if The Community™ dislikes that, then The Community™, rather than ArbCom, can change the policy). WhatamIdoing (talk) 19:39, 24 January 2017 (UTC)

This is true if we include Rich's "unless there is consensus to do so", but also kinda besides the point. Yobot's purpose was never to make cosmetic edits. If anything, the wording of the statement should mirror WP:COSMETICBOT more closely. Headbomb {talk / contribs / physics / books} 14:37, 5 February 2017 (UTC)

Accepted practice

3) The aim of many guidelines is primarily to describe current practice, to help editors to understand how Wikipedia works. If something becomes accepted practice, it may be assumed to be legitimate unless it's shown to be controversial.

Comment by Arbitrators:

Comment by parties:

Comment by others:

The first sentence here is taken from WP:Polling is not a substitute for discussion#Policy and guidelines. עוד מישהו Od Mishehu 12:43, 9 January 2017 (UTC)

Sure, the reverse is also true. If something is observed to cause controversy, it shouldn't be assumed to be legitimate just because some wikilawyer claims it's permitted by obscure wiki guideline 27B stroke 6. 50.0.136.56 (talk) 07:02, 11 January 2017 (UTC)

To the best of my knowledge, prior to this ArbCom case, there was no reason to think that deletion of these stub tag reirects was controversial, which is the point I'm trying to make here. While this sentence was taken from an obscure essay, I believe that this is absolutely the way that we work here. עוד מישהו Od Mishehu 06:05, 12 January 2017 (UTC)

I'm not too worried about those deletions either way. The big issue is the many-year conflict shown in the pre-arbcase discussion threads. 50.0.136.56 (talk) 07:32, 12 January 2017 (UTC)

We should probably take something from Wikipedia:Policies and guidelines or another actual policy or guideline to describe what policies and guidelines are. "Describe current practice" is only a single use of the many and nowhere in PAG do I see anything to suggest that they codify current practice "primarily". --Izno (talk) 21:27, 17 January 2017 (UTC)

That stuff used to be written down, but Wikipedia's Vogon contingent has systematically bureaucratized the relevant documents. That doesn't (or shouldn't) change how things work: it just means they put errors into the documents, which happens. 50.0.136.56 (talk) 11:34, 18 January 2017 (UTC)

It looks like that change was made in July 2009. Whether it's better or worse is an open question; what I can tell you is that since then, very few people have asked for help figuring out what the source of policies ought to be. WhatamIdoing (talk) 20:03, 24 January 2017 (UTC)

This idea seems to be a hidden pinciple behind Finding of Fact #8 in the first MZMcBride case. עוד מישהו Od Mishehu 22:07, 30 January 2017 (UTC)

Proposed findings of fact

Magioladitis's conduct with stub tag redirects

1) Magioladitis has been engaged in orphaning stub tag redirects and then deleting them. While the deletion appears to be accepted practice, orphaning them wasn't, especially as it was done by a bot which was operated only for cosmetic purposes.

Comment by Arbitrators:

Comment by parties:

Comment by others:

I don't think deleting with a bot was accepted practice. I don't believe the concept that any editing ok for a human is automatically ok for a bot, even a MEATBOT. This is partly to prevent fait accompli (a principle found in many arb cases). It's not just the mechanization, it's the speed and volume. Plus, adminbots have traditionally been harder to get approved. 50.0.136.56 (talk) 21:40, 9 January 2017 (UTC)

If the deletions are uncontroversial, then there it doesn't matter how fast they were done. Distinguishing between controversial and uncontroversial is a matter of judgement. As far as the stub-template-redirects are concerned they were not controversial, so the judgement in this case was sound.

All the best: Rich Farmbrough, 00:04, 11 January 2017 (UTC).

You don't find out whether something is controversial until after people have had time to react to it. So either 1) you have to do it slowly enough that objecting editors can intervene and stop the action before too many edits/deletions have taken place; 2) the action itself is easily reversible (WP:BRD contemplates the action being 1 single edit); or 3) there's consensus (community RFC) or reasonable evidence of consensus (BRFA if BAG is operating responsibly) for the action before it starts. 50.0.136.56 (talk) 07:39, 12 January 2017 (UTC)

You are quite correct. Stub redirects, I understand, have been deleted for years.

(I'm not sure that I agree, not having examined the practice in detail, but there are many accepted practices I don't agree with.)

All the best: Rich Farmbrough, 21:25, 21 January 2017 (UTC).

AutoWikiBrowser

2) AutoWikiBrowser (frequently calle AWB) is an external program which is very useful for doing repetitive edits on Wikipedia. A significant percentage of all edits to the mainspace are done by this tool, largely for tasks which probably wouldn't get done without it. Magioladitis is one of its developers.

Comment by Arbitrators:

Comment by parties:

Comment by others:

I'll resist making a snarky comment about those "tasks which probably wouldn't get done without it" and simply compare the resulting edits to software bloat. 50.0.136.56 (talk) 07:07, 11 January 2017 (UTC)

I started using AWB in Dec 2015. I use it for a lot of the sort of edits that I made before then, but I now get rather more edits done in a given amount of time. I like to think of my editing as the opposite of software bloat, but I'll concede I'm not an a typical AWB user. ϢereSpielChequers 19:29, 27 January 2017 (UTC)

AutoWikiBrowser's general fixes

3) One of the features of AWB is that it will suggest several "general fixes" and "auto tags", based on rules that most AWB users don't know. A few of the general fixes can be disabled in a separate dialog box, but most of them come as a single package deal that is either "on" or "off". It's quite likely that most AWB users have this feature on, and don't take more than a quick glance at what fixes are done to any specific page.

Comment by Arbitrators:

Comment by parties:

Comment by others:

I added the words 'that is either "on" or "off" for clarity. Hope that's ok. All the best: Rich Farmbrough, 21:27, 21 January 2017 (UTC).

This translates into plain English as "We speculate that an actual majority of AWB users are negligent and not following the mandatory rules for its use". I expect that ArbCom would need to see some actual evidence to support this claim before they could vote on it. WhatamIdoing (talk) 20:12, 24 January 2017 (UTC)

Proposed remedies

Note: All remedies that refer to a period of time, for example to a ban of X months or a revert parole of Y months, are to run concurrently unless otherwise stated.

Admins unblocking their own bots

1) A Request for Comments should be openned on the question of the legitimacy of admins unblocking their own bots - that is, on rules on when it may and when it may not be done. Until the said discussion is closed, Magioladitis may not unblock Yobot.

Comment by Arbitrators:

Comment by parties:

I provided evidence that all the blocks were unrelated to the "cosmetic changes" editing. I also provided evidence that the unblocks were in communication with the blocking admin. I also underline the fact that no evidence of connection was ever presented apart from numbers ("Yobot has been blocked xx times") just to impress. I would like also to note that WhatamIdoing has a point on what it is discussed here. -- Magioladitis (talk) 21:11, 3 February 2017 (UTC)

Comment by others:

I believe that part of the problem here is that there is simply no real agrement on this question. My proposal would actually fix this issue. And since Magioladitis doing this is apparently a significant part of this discussion, he should be prevented from doing this until the issue is resolved.

• Weak oppose This type of RFC always results in choosing the most bureaucratic alternative available. The unblock issue doesn't cause real controversy much of the time. If there's concern about Magioladitis doing more improper unblocks after all this drama, then it should be enough make a bot-unblocking restriction specifically against Magioladitis or Yobot. We have too much policy when we should be instead be trying to get editors to cultivate good judgment (= the ability to make good decisions when there's no policy or algorithm saying what to do). We can't do much about regular editors lacking good judgment, but it's supposed to be part of CIR for admins.

  I think Iridescent's analysis of when it's ok to unblock describes good practice pretty well. 50.0.136.56 (talk) 00:19, 13 January 2017 (UTC)

This has been discussed a number of times, with substantially the same result: if you have fixed the issue, unblocking is fine. In other circumstances you unblock at your peril. All the best: Rich Farmbrough, 02:02, 16 January 2017 (UTC).

The issue is: The bot gets blocked, the operator modifies it in a way he thinks fixes the probelm, and then unblocks the bot. The the bot gets blocked again (possibly for the same bug, although it's impossible to be sure), the operator modifies it in a way he thinks fixes the probelm, and then unblocks the bot. Repeat several times. עוד מישהו Od Mishehu 15:08, 17 January 2017 (UTC)

If someone has been entrusted with both a sysop bit and the privilege of running bots, they should have good enough common sense to know when to stop. If they've reached 3RR against their own bot, it's past that time. 50.0.136.56 (talk) 11:06, 18 January 2017 (UTC)

If you see a few minutes or hours between blocks, then this is a possible scenario. When you see two and a half years between blocks it is not. All the best: Rich Farmbrough, 00:29, 20 January 2017 (UTC).

עוד מישהו, how do you see this working in practice? I see it working like this:

A: I blocked your bot because it frobbed the widget in this article.

B: Okay, I think I fixed it!

C: Great, I'm unblocking the bot for you!

A: Hey, I'm blocking your bot again! It frobbed the widget in this other article!

B: Argh, I thought I had this fixed. Okay, I've tried again.

C: Great, I'm unblocking again!

Do you really see admin C adding any value here? I don't. WhatamIdoing (talk) 20:20, 24 January 2017 (UTC)

Puts WP:BAG hat on. I'll mention here that current practice is that unblocking your own bot is frowned upon in the case of dispute concerning the consensus of a task, but usually uncontroversial in the case of a technical glitch/bug (assuming of course the issue is fixed). When in doubt, checking with the blocking admin is always good. Takes WP:BAG hat off However, AFAICT, there's no such issue here, since we'd have had WP:WHEELWARRING and several WP:AN threads about abuse of admin tools had Magio improperly unblocked his bot. Headbomb {talk / contribs / physics / books} 14:47, 5 February 2017 (UTC)

Improvements to AutoWikiBrowser

2) AutoWikiBrowser (AWB) should be fixed so that the interface allows for shutting off any subset of general fixes and auto tags. The descriptions of all the fixes should be written clearly enough so that the average AWB user on English Wikipedia would be able to understand exactly what it does.

Comment by Arbitrators:

Comment by parties:

T138977 and Wikipedia_talk:AutoWikiBrowser/Feature_requests/Archive_12#Enhanced_Autotagging since 2008. -- Magioladitis (talk) 22:59, 11 January 2017 (UTC)

Comment by others:

It's not really ArbCom's place to demand that volunteers write code or update documentation (you're welcome to do some or all that work yourself, of course). WhatamIdoing (talk) 20:24, 24 January 2017 (UTC)

Agree with WhatamIdoing. Outside of ARBCOM's mandate to dictate policy and software development. Headbomb {talk / contribs / physics / books} 14:50, 5 February 2017 (UTC)

AutoWikiBrowser general fixes discussed

3) All existing general fixes in AWB should be discussed by the community, in separate discussions and gradually, to determine which ones have consensus. No new general fix may be added without prior community discussion.

Comment by Arbitrators:

Comment by parties:

Comment by others:

This kind of thing is very hard to enforce, and it focuses on on piece of software, which could always be replaced with another. I think it is better to focus on the contents of the edits. This mean we can use the 'diff' tool on the wiki to see what was changed, which is much easier than trying to look at source code which may be unpublished or unreadable. — Carl (CBM · talk) 13:55, 20 January 2017 (UTC)

It's probably not "enforceable" but it is probably a good idea. Hence the phrase "should be" - which I hope is to be interpereted as in RFCs. All the best: Rich Farmbrough, 21:34, 20 January 2017 (UTC).

"Separate discussions" for dozens of trivial things means serious RFC fatigue. After a couple of these, we're realistically going to see fewer participants, less discussion, no meeting of the minds, and !votes that are basically "I hate all GenFixes" and "I <3 all GenFixes".

Also, are we-the-community so hopelessly broken that we actually need ArbCom to tell us to start a discussion about things whose consensus is unclear to us? If you personally see value in have a discussion, then you personally should start a discussion about the GenFix that interests you the most! You don't need ArbCom to authorize you talking to your fellow editors. WhatamIdoing (talk) 20:31, 24 January 2017 (UTC)

Let the community decide how it wants to handle genfixes/checkwiki fixes. If it wants to approve things in bulk and instead discuss which fix to demote, that the community's prerogative. And likely a more productive discussion given the vast majority of genfixes/checkwiki fixes are not controversial. Headbomb {talk / contribs / physics / books} 14:53, 5 February 2017 (UTC)

Magioladitis and general fixes

4) Magioladitis may not run any AWB tasks with the general fixes enabled.

Comment by Arbitrators:

Comment by parties:

Comment by others:

This appears to be a major part of the problem here. עוד מישהו Od Mishehu 14:12, 24 January 2017 (UTC)

Magioladitis and automation

5)Magioladitis may not run any automation which he has been part of the development team for, except as autherized by a BRFA or for a limited number of edits necessary for development.

Comment by Arbitrators:

Comment by parties:

Od Mishehu which problem does this solve exactly? -- Magioladitis (talk) 15:27, 24 January 2017 (UTC)

This is designed to deal with the issue of you running AWB on tyour main account when the bot was blocked, which I saw some complaints about. עוד מישהו Od Mishehu 18:12, 24 January 2017 (UTC)

Od Mishehu so you ask that I am prohibited by any automation because...? What wrong did I do from my main account in large scale? If you had some evidence you should have provided. -- Magioladitis (talk) 19:55, 24 January 2017 (UTC)

Hasteur this reply was referring to the controversial issues. There were two longstanding issues: Whitespace between headers and ref ordering. Both fixed after many years. Not just any bug. By Murphy's law there is always a bug. It would be naive to believe to believe that AWB is bug free since there are many bug reports open. -- Magioladitis (talk) 15:32, 26 January 2017 (UTC)

Comment by others:

This clearly includes AWB, which is where the major problems are; it also (in case anyone is worried about this) prevents him from the possibility of WikiLawyering by taking pieces of AWB code, using them for a new program, and saying that it's not AWB. עוד מישהו Od Mishehu 14:14, 24 January 2017 (UTC)

@Magioladitis: Because admissions like this inspire less than zero confidence in your ability to develop defect free code especially in light of having the perception of "Just a quick fix" being the answer for an extended period of time. If you were developing code for your off-wiki job and repeatedly had people express concerns that your code was buggy and your answer was "Just let me fix a bug" for multiple years, would your boss be willing to continue employing you? That's essentially the heart of this dispute. The users of the software (i.e. Community at large) have raised enough issues that your boss (ArbCom) is having a frank discussion with you about the quality of your code. Hasteur (talk) 14:51, 26 January 2017 (UTC)

Proposed enforcement

Template

1) {text of proposed enforcement}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) {text of proposed enforcement}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposals by user:WereSpielChequers

Proposed principles

Bot edits are useful, but can inadvertently hide vandalism

1) Vandalfighting and Bot maintenance are both useful to Wikipedia, but the two have long been in conflict because valid bot edits can hide recent vandalism from watchlisters. This could be resolved or greatly reduced by coding maintenance bots to only edit articles which have not changed in a given period of time.

As NewYorkBrad has pointed out this could alternatively be done by changing the watchlist system. But that would be a complex system that required watchlists to differentiate between anti vandal bot and other bots, and there would also be the problem that simple undos wouldn't work, you would need to revert to the version prior to the vandalism, and thereby possibly lose something else done by the maintenance bot. Having maintenance bots ignore the most recently edited articles is not the only way to solve this problem, but it is likely to be the simplest. It also enables incremental improvements to the process in that every recoded bot would be a step in the right direction.

Comment by Arbitrators:

... or by changing the effect of the watchlist preference "hide bot edits" to "show last non-bot edit on the article" rather than "omit the article if the last edit was a bot edit." (Except for vandalism-reversion bots, as showing just the vandalism and not the bot reversion of it would mislead human editors to think that there's vandalism reversion still to be undone, and thus to duplicate work and confusion—how do we address that?) But your suggestion would not require changes to wikicode and would therefore presumably be easier to implement. Newyorkbrad (talk) 17:43, 9 January 2017 (UTC)

Do we have (or can we get without extraordinary effort) information on what percentage of maintenance bots' edits affect articles that have been edited within the past hour/day/week/whatever-time-period? Relatedly, do we have information bearing on how long an edit should stay on watchlists to maximize the chances that bad edits, but not so bad as to be detected by the anti-vandalbots, are reverted? Newyorkbrad (talk) 18:25, 9 January 2017 (UTC)

Dating maintenance tags is generally done after about 15 minutes with no editing. All the best: Rich Farmbrough, 00:17, 11 January 2017 (UTC).

Comment by parties:

T127173 Created in February 2016. -- Magioladitis (talk) 12:07, 11 January 2017 (UTC)

Comment by others:

@Newyorkbrad - it's hard to identify bot edits reliably, but maybe there are some usable approximations. I can think of ways to get those numbers with considerable effort. It's doable though. 50.0.136.56 (talk) 05:25, 10 January 2017 (UTC)

There is a gadget that enables one to watchlist ignore bot edits, while still seeing the previous non-bot edit. Every time I mention this I have to rely upon a stalker to identify the gadget. All the best: Rich Farmbrough, 00:15, 11 January 2017 (UTC).

How does it do that? If you mean just bot-flagged or registered bot edits, that's unhelpful. The vast majority of automated edits are unflagged. What I want to see is edits that change the article content. Maybe that can be done semi-reliably by having a script check the diffs, hmm.

This was fun: [16] [17][18][19][20] a group of 5 edits to the same page appearing in the middle of a multi-thousand edit bot run. Each one changes the article length by 0 bytes. To save some clicking, the first revision had a list of judges named Douglas (John Douglas, Charles Douglas, etc). The next 4 edits each permuted the names in the list to bring them closer to and finally into alphabetical order. It looked like somebody's bot had a sorting algorithm that saved the page after each pass, which is Daily WTF material and terrifying. After a while I decided they were probably manual edits, i.e. the person was goofing around or hadn't had enough coffee yet, no big deal either way. But I can't know for sure. It messes up some ideas I had about identifying bot sequences by noticing the account hopping around between lots of pages quickly. 50.0.136.56 (talk) 07:34, 11 January 2017 (UTC)

This was manual: https://en.wikipedia.org/enwiki/w/index.php?title=Justice_Dunn&action=historysubmit&type=revision&diff=755166549&oldid=755166513 explains it. I had a system planned for keeping lists ordered, but someone delete the key template. Oh well. All the best: Rich Farmbrough, 02:07, 16 January 2017 (UTC).

Yeah, I remember looking at that in deciding the sorting was probably done manually. I don't think any automated tools are needed for keeping lists like that ordered, especially when ordering is an editorial decision (and a decision leaving a list slightly out of order on purpose is possible too). If a given list should be sorted automatically, it should be written as one of those tables with sorting buttons at the top. 50.0.136.56 (talk) 22:23, 18 January 2017 (UTC)

Does that work without something being clicked? And maybe not a good idea for large lists. Sorting them hundreds of times rather than just once is wasteful.

All the best: Rich Farmbrough, 21:36, 20 January 2017 (UTC).

Invitations to block need to be formalised

2) Admins who operate bots have long encouraged other admins to freely block their bots if they misperform. Implicit to that is that the bot operator is free to unblock their bot when they've fixed things. Clearly not every editor who looks at the consequent blocklogs and possibly not every admin who issues such blocks accepts that such unblocks are legit. ARBCOM would do everyone a service by clearly stating that in future such self unblocks are only acceptable where the blocking admin has stated "feel free to unblock yourself when you have fixed your bot". Ideally in the blocklog when setting the block.

Comment by Arbitrators:

Comment by parties:

I think this might be a good idea as a general principle. -- Magioladitis (talk) 19:56, 27 January 2017 (UTC)

"we know exactly what Magioladitis will do" bad faith? I am still waiting to understand which "edits that are contested here". I don't understand with BU Rob13 never becomes explicit when it comes to this question. Is this because the people who contest some edits do not really agree with each other? -- Magioladitis (talk)

BU Rob13 I'll resume my CHECKWKI when the case resolves because even you agreed that CONSENSUS > COSMETICBOT. There seems to be an agreement to that. On the other hand you mentioned some AWB functions randomly but this does not mean there is consensus what exactly are cosmetic edits. The set everyone defines is different. Are we making circles in this discussion? Why we can't agree on what our views are at least? I explained that a vague definition does not help here. If there s something good out of this procedure we follow is that at least are rules will become more specific. -- Magioladitis (talk) 23:22, 20 January 2017 (UTC)

Comment by others:

Seems like instruction creep to prescribe this. If the bot was blocked because its normal operation was being contested, or if the blocking admin made it clear that the operator shouldn't unblock the bot, then unblocking it anyway is obviously not good. Otherwise, observed practice has generally been that if the bot is blocked for a specific problem and the operator clears up the problem for real, their unblocking the bot usually doesn't cause controversy. 50.0.136.56 (talk) 21:43, 9 January 2017 (UTC)

At the heart of this Arbitration request is the record of Magioladitis unblocking his bot and whether that is taboo or OK. If we say unblocking is OK provided a bot is successfully fixed then we enter into all sorts of mess re bots that run multiple programs and whether two bugs are related or not. That would be bad enough if we were all programmers and had access to view the source code. I can only see three viable solutions.

1. No self unblocking with no exceptions.
2. No self unblocking except with explicit consent "OK to unblock your bot yourself when you've debugged".
3. No self unblocking if consent is withheld "please don't unblock yourself, happy for any other admin to unblock".

My preference is for the second. ϢereSpielChequers 08:11, 11 January 2017 (UTC)

I don't think unblocking Yobot is the heart of the arb request. It's one of a bunch of issues raised, and maybe an easy one to shake a policy finger at. The heart of the request is there's been a series of incidents stretching over 7(?) years that didn't seem headed towards letting up. 50.0.136.56 (talk) 00:54, 13 January 2017 (UTC)

This is a rather good description of things. I could look at any one of the many issues or incidents here and see something wholly fixable. But it all adds up, both across different incidents, and across the long time period. It shows no signs of stopping. In fact, we know exactly what Magioladitis will do once the case ends in the absence of restrictions, since he's plainly stated on his user talk and elsewhere that he plans to continue the edits that are contested here. As stated in WP:A/G, "arbitration aims to "break the back" of the dispute". My proposed restrictions aim to do just that. ~ Rob₁₃^Talk 21:55, 20 January 2017 (UTC)

I have explained it to you every which way, Magioladitis. At WT:Bot policy, I spelled out exact AWB genfixes that are currently cosmetic. I've cited specific edits in Evidence and Evidence talk. I've provided an easy-to-apply definition multiple times. As for bad faith, yes, I believe you're acting in bad faith at this point by refusing to follow community policies, but I'm not assuming that. You've literally said on your user talk that some of the most recent objectionable edits would be continued after this case closes. ~ Rob₁₃^Talk 22:51, 20 January 2017 (UTC)

I had a look at your recent comments at WT:BOT in the hope of figuring out the definition of cosmetic edits, and I see confusing responses. You defined cosmetic changes as "An edit which doesn't change the visual output of the page", to explain your opposition to removing hidden categories – except, of course, that removing categories does "change the visual output of the page" (for those of us that view hidden categories). Renaming files after a file move, on the other hand, doesn't appear to change the visual output of the page, but you declared that to be a non-violation of COSMETICBOT. And then, in the RFC, you made a really useful list: "TemplateRedirects, FixSyntax*, SimplifyLinks, FixCitationTemplates*, FixSyntaxRedirects, FixCategories, SimplifyReferenceTags, FixImages, FixLinkWhitespace*, RemoveEmptyComments" – ten examples of edits that you personally consider cosmetic. You marked three of those with an asterisk to indicate that you only sometimes consider those cosmetic edits. One in your list (RemoveEmptyComments) isn't really a GenFix (according to subsequent discussion), and another which (FixSyntaxRedirects) is non-cosmetic ("make busted attempt at creating a redirect actually work"). #REDIRECT User talk:BU Rob13 is a redirect, but this:

#REDIRECT [[User talk:BU Rob13
]]

produces a numbered list. And that's just your off-the-cuff list; it's not a list that has broad support in the community. I think that there is room for both good-faith disagreement about what ought to be covered by COSMETICBOT and good-faith confusion about what the community does or doesn't want to have covered. WhatamIdoing (talk) 21:04, 24 January 2017 (UTC)

@WhatamIdoing: The definition, in general, is that a rendered change should be made. There are exceptions, most of which proceed from common sense, such as changes that do not render a change but are required for accessibility reasons for those using screen readers - a device which doesn't render anything. There are further exceptions allowable by consensus. There's consensus, for instance, that we should not keep many file redirects and should instead move files without redirects in most cases. In that case, the actual edit changing a file name from redirect to main file doesn't render a change, but it's necessary to get rid of the file redirect, which consensus says we should do. As for the two genfixes, I went by WP:GENFIXES. RemoveEmptyComments was included on the list, so I included it in mine. FixSyntaxRedirects should have had a *. The problem with citing individual genfix tasks is that if you look at the GENFIXES page, each fix is actually a set of many fixes. For instance, FixSyntaxRedirects contains the busted redirect fix. That's clearly ok. It also contains a fix to remove an unnecessary namespace, which is clearly not ok if the redirect works the same without the namespace there (although honestly, I'm not too sure what it means - I don't know of a situation where a namespace isn't mandatory in a redirect?). Similarly, FixCitationTemplates and other ones I asterisked have both rendered and non-rendered changes.

There's definitely some room for confusion, which is why AWB rule of use #4 tells operators that they must stop and seek consensus on any changes that are challenged under that rule. Magioladitis has not done that. In fact, when this has been brought to large venues, the community has repeatedly told him many of his edits are unacceptable and must stop. He's ignored that. WP:BOTREQUIRE states "performs only tasks for which there is consensus", which a reasonable bot operator would interpret to mean that they should stop their bot and discuss when edits prove controversial, more or-less the same as with the AWB rules of use. These tasks have continued without such consensus-gathering for the many years the community has dealt with this. Our bot policy states "Should bot operators wish to modify or extend the operation of their bots, they should ensure that they do so in compliance with this policy. Small changes, for example to fix problems or improve the operation of a particular task, are unlikely to be an issue, but larger changes should not be implemented without some discussion. Completely new tasks usually require a separate approval request. Bot operators may wish to create a separate bot account for each task." This is partially so that the community can examine additions to a task to ensure that they comply with community standards. Instead, Magioladitis has implemented entirely new CHECKWIKI fixes without consulting anyone, even in the face of substantial opposition, and despite his original BRFA containing a specific list of fixes that would be implemented at the request of the BAG member who eventually approved that set of fixes (who has recently clarified they did not intend to approve the edits that have been scrutinized). In short, my point is that while this policy (like WP:GNG and many other policies and guidelines) may require some interpretation (see my remedy 4 below), the actions here fall well outside what can be expected of a bot operator and an administrator. We don't need to understand any bit of COSMETICBOT to know that events should not have proceeded as they did following this discussion. We just need a basic appreciation for WP:CONSENSUS, which is itself enshrined deeply in our bot policy. ~ Rob₁₃^Talk 21:20, 24 January 2017 (UTC)

I agree that, as a first approximation, "a rendered change" should be acceptable, but I also see on the /Evidence page that half the links you provided produced a rendered change, and you still listed them as evidence that changes were in violation of this requirement.

So if I were a bot op, and you showed up on my talk page with that kind of "evidence" that my bot was misbehaving, I doubt that I would be convinced that there was consensus against that kind of edit. I'd probably conclude that you just hadn't looked at the diff closely or misunderstood it – and I'd probably think it quite reasonable to continue reaching that same conclusion no matter how many editors made the same mistake in interpreting the diffs. "Some people misunderstand what's happening in a diff and decide to yell at you about their mistaken understanding of reality" is not evidence of an unclear consensus that should prompt a bot operator to have a discussion about whether policy-compliant edits should be made. Instead, that's evidence that the diffs system needs work (e.g., so that it's not so easy for even highly experienced editors to overlook an appearance-changing edit when other GenFixes are going on). WhatamIdoing (talk) 05:06, 25 January 2017 (UTC)

This is a matter concerning WP:BLOCK and WP:BOTPOL policies, and it is not in the mandate of ARBCOM to decide what those policies should be. Headbomb {talk / contribs / physics / books} 14:59, 5 February 2017 (UTC)

Proposed findings of fact

Conflict between maintenance bots and vandalfighters

There is a longstanding conflict between bots doing minor fixes and vandalfighters patrolling their watchlists for vandalism. Watchlists show the most recent edit to an article, if a vandal vandalises an article immediately after a good edit that will prompt a bot edit then the vandalism is likely to be hidden by a bot edit.

Comment by Arbitrators:

Comment by parties:

I agree with that. This is the reason I gave part of Ramaksoud2000's profile above. On the other part of watchlists I am still waiting for evidence that me or Yobot (recall: 4 million edits. there should be enough examples) ever hid a vandalism that the vandalfighters would catch. -- Magioladitis (talk) 21:14, 3 February 2017 (UTC)

Comment by others:

Cosmetic edits are those that don't make even a cosmetic change

There has been some argument as to what is and is not a cosmetic edit in terms of wikipedia jargon. Longstanding convention if originally applied rather informally is that counterintuitively a cosmetic edit on wikipedia is one that doesn't even make a cosmetic change visible to a reader. This sometimes causes confusion, not least when an edit changes the appearance of an article in some browsers but not others. Editors who wish to change the policy on this should seek consensus to change the policy not try to enforce a stricter definition on individual bots and bot operators.

@ NewYorkBrad. My understanding from my observations is that Wikipedia:COSMETICBOT assumes cosmetic edits to be edits that are not visible to any readers. Arguments tend to be over edits that are not visible to the critic but allegedly would be visible to people with other hardware or software. But it is possible that the parties may know more and have a different understanding than me.

Comment by Arbitrators:

Does "cosmetic" as used in "don't make cosmetic-only bot edits" mean edits that affect only the cosmetics of the article visible to readers, or that affect only the cosmetics of the mark-up window? Or is there a lack of clarity on this point? Newyorkbrad (talk) 17:38, 9 January 2017 (UTC)

Comment by parties:

Comment by others:

@ NewyorkBrad - Cosmetic traditionally means an edit that doesn't change what a human reader sees on the screen. E.g. if the markup says "foo   bar", (three ordinary spaces between the words, not the nonbreaking spaces that I put in to make that show up properly), browsers will render the words the same way as if there were just one space. So crunching the three spaces to one is arguably a cosmetic change, since browsers will light the exact same pixels after the edit.

The contra argument if I understand it is that the spaces can affect the pacing of screen readers (text-to-speech devices that read the page out loud), so adjusting them is non-cosmetic: "fixing" them lets a sight-impaired reader hear the sentence the "right way". Some editors have said that screen readers acting like that aren't respecting the HTML standard so they should be considered buggy. There have been long ongoing discussions (not just in the current dispute) about this spacing issue, but I haven't followed the issue so I might have gotten this wrong.

ISTM that if those changes really are non-cosmetic, then they are contextual as well: as a writer I might put in extra spaces to produce good pacing for screen readers, and bots shouldn't mess that up. Or alternatively, if that reasoning is rejected (i.e. extra spaces are presumptively human errors), then maybe the spaces can be fixed on the server side, by having the mediawiki html renderer take them out (possibly as a user preference). That would be a one-time fix in one single place that handles the problem on every page forever, which is far preferable to having bots editing millions of pages on a continuous basis. Again maybe someone can explain if I'm mistaken about this. 50.0.136.56 (talk) 22:45, 9 January 2017 (UTC)

I don't believe there have been any bots - or people - for a long time who have been making substantial numbers of edits to fix these sorts of white space issues. I did receive a well-intentioned communication last year from someone who proposed such a plan, but I was able to persuade them that this would not be looked on favourably by the community, and was probably not a good way of doing it.

All the best: Rich Farmbrough, 00:13, 11 January 2017 (UTC).

Good to hear. I remember it being a topic of discussion at the beginning of the case. 50.0.136.56 (talk) 11:32, 18 January 2017 (UTC)

There are diffs on the evidence page that show the bot (correctly) removing extraneous blank lines between sections. Multiple blank lines (thing "empty paragraphs", not "extra spaces between words") change the appearance of the page. It's usually an accident, but it's sometimes done intentionally by editors who don't know how to use the {{clear}} template and are trying to get images to line up nicely (for their own screen/browser/font combination – it's a bad idea to do this, but most people don't know that). WhatamIdoing (talk) 21:09, 24 January 2017 (UTC)

@NewYorkBrad: Generally speaking, the 'cosmetic' in WP:COSMETICBOT refers to edits that changes the appearance of the wikitext in the edit window, but not the appearance of the actual rendered page. The draft for a revised WP:COSMETICBOT section should make things clear. Headbomb {talk / contribs / physics / books} 15:04, 5 February 2017 (UTC)

Agree, it refers to the mark-up window. More importantly, that meaning of "cosmetic" is what was presented at the AN on Magioladitis in January 2016 here, where the original post stated "The most common complaint is that he makes many trivial ("cosmetic") changes which do not change the appearance of an article ...", so it's the definition he should be expected to know. ~ Rob₁₃^Talk 15:14, 5 February 2017 (UTC)

Proposed remedies

Magioladitis reminded

1) Magioladitis and other admin bot operators are reminded to only unblock their account or bot accounts when explicitly invited to do so by the blocking admin.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Would @Iridescent: comment? I thought the problem with Mag's unblocks was the specific circumstances they were done under, but I didn't have the impression that unblocking was badly perceived per se. It's just turning a machine on and off. 50.0.136.56 (talk) 00:07, 10 January 2017 (UTC)

It's not really my area; my impression would be that it's not a good field for a hard-and-fast policy, as it will depend on circumstances.

• If a bot is blocked because it's consistently causing a specific error, it would be legitimate for the operator to unblock once the bug has been fixed;
• If a bot is blocked because there's a concern it's acting outside its approved remit, or that its approved remit is having unintended consequences, it would be bad form for the bot operator to unblock even if they'd addressed the issue, without at least confirming with the blocking admin that they're satisfied the issue has been addressed. Although in this case the unblocking would be inappropriate, I wouldn't consider it rising to the level of admin abuse, just bad manners;
• If a bot is blocked for a reason the bot operator considers inappropriate (the bot is being blamed for something it didn't do, or someone is objecting to the task despite there being an unarguable consensus to carry it out), the bot operator should under no circumstances unblock the bot themselves, but should explain to the blocking admin (or someone else if they can't be found) the the block is in error and let someone else make the call. If there's one thing Beta, RF, Lightmouse et al have taught us, it's that bot operators make very poor judges of whether their bot is actually doing what the community wants it to; in these circumstances, I would consider a bot operator unblocking their own bot to be admin abuse.

If a bot is blocked for whatever reason, whether valid or not, then just shifting the bot across to the operator's human account—which is one of the accusations here—is obviously totally inappropriate, and it probably should be written into policy somewhere that if it can be demonstrated beyond reasonable doubt that anyone's pulling that stunt it should trigger an automatic permanent ban from Wikipedia. ‑ Iridescent 11:07, 10 January 2017 (UTC)

Running a bot on a human account is not (usually) allowed in any case. Doing the same job with a human account, however, is. It would be absurd to say that there are edits that bots may make, but humans may not.

Certainly if someone claims that the job is detrimental to WIkipedia, that should be taken into account. But there are always Wikipedians, and often admins to say that anything is detrimental. I had no less than two Arbitrators tell me that creating articles denied others the opportunity to create better ones, and that fixing spelling errors discouraged people from making their first edit.

So any suggestion of an automatic permanent ban for making perfectly good manual edits - which would be permissible to any other editor - does not really make sense.

All the best: Rich Farmbrough, 00:29, 11 January 2017 (UTC).

This case does not have scope for other admin bot operators. It's good to be reminded, but reminders have been used in past cases as evidence that users have not been suitably reminded, which would bite bot operators who are otherwise acting correctly at the time of this case. --Izno (talk) 21:34, 17 January 2017 (UTC)

IMO if this is a general problem, then someone should take this idea (or Iridescent's description of the standard sensible practices) over to WP:BOT and make an explicit policy WP:PROPOSAL to add this new restriction on bot operators. ArbCom does not create policies for the community. WhatamIdoing (talk) 21:12, 24 January 2017 (UTC)

As a procedural note, no remedy should be offered that targets "other admin bot operators", as the scope of this case is explicitly on Magioladitis only and other bot operators were not parties in this case. If this sentiment is shared by the Committee, they could word it as "Magioladitis is reminded that unblocking one's own bot account without an explicit invitation to do so by the blocking administrator is a violation of the unblocking policy and the policy on involved administrator actions. Unless these policies are amended by the community, unblocking one's own alternate or bot account would be considered abuse of administrative tools." The last sentence applies to everyone without specifically "reminding" them as if they're a party to the case. ~ Rob₁₃^Talk 15:18, 5 February 2017 (UTC)

Magioladitis is restricted

2) Magioladitis is restricted in that his bot edits may not edit articles that have been manually edited in the last 48 hours.

Comment by Arbitrators:

Comment by parties:

I would like to note that no Evidence was presented that my editing (not even Yobot's with the 4 million edits!) ever hid vandalism. In fact some of the fixes revealed vandalism which fixed. It's interesting how all the story was built in an hypothetical scenario which still nobody confirmed by any evidence. -- Magioladitis (talk) 20:17, 3 February 2017 (UTC)

Comment by others:

This is interesting and it would be nice if it applied to almost every bot, but it's hard even for humans to tell manual from automatic edits. It has to be harder still for a bot. It would also be great if the different bots could combine their changes into a single composite periodic edit instead of having them all editing nonstop. 50.0.136.56 (talk) 22:23, 9 January 2017 (UTC)

I'm assuming that if this or something like it works and reduces tension between Magio and the watchlisters then other bot operators will take note or at the least find people suggesting this at BAG. But first it would be good to see if it can work, and what the right interval is. We have over 5 million articles and only about 5 million edits per month, so at any one time over 90% of articles haven't been edited for 48 hours. I'm assuming that bot operators will be able to code their bots to differentiate between articles that have been edited in the last 48 hours or not and edited by an account with the bot flag or not. We could go more sophisticated and ignore extended confirmed edits, but my suspicion is that it isn't just the vandalfighters who are bothered about this but also other watchlisters who want to see others edits because they are collaborating and want to know when the other has responded to their change. I do think we could ignore reverted vandalism though, so if the only thing in the last 24 hours is replacing a section with Kilroy woz here and then reverting that I see no problem in bots editing afterwards. ϢereSpielChequers 08:34, 11 January 2017 (UTC)

Again the single edit is a nice idea - and precisely why stuff like General Fixes exists. However there are two drawbacks: firstly people complain that there are too many changes, and they cannot understand the diff. Secondly the more you wrap in one edit, the more scope for error.

True bots should, by and large, have no problem detecting the nature of previous editors. Helpful Pixie Bot had custom delays for a number of editors who had requested a minimum period after their edit before the bot edited. (And if even an "unresponsive" bot runner can do that...)

Whether this is something AWB should implement as a hybrid, is up to the devs.

All the best: Rich Farmbrough, 23:37, 17 January 2017 (UTC).

It's hard even for humans to detect the nature of previous editors. How are bots supposed to do it? 50.0.136.56 (talk) 11:30, 18 January 2017 (UTC)

I don't see how delaying bot edits after edits by editors who request it helps anyone. The idea of giving watchlisters the opportunity to catch vandals would mean a time delay after edits by IP editors and newbies. Or after everyone if they also want to see other goodfaith editors. ϢereSpielChequers 22:25, 23 January 2017 (UTC)

48 hours is certainly too long. I base this view on the fact that if an editor hasn't saved an edit during the last 30 minutes, then that editor is probably off wiki. There may be a delay in reviewing edits, but it's unlikely to be ~100 times as long as an editing session. WhatamIdoing (talk) 21:22, 24 January 2017 (UTC)

48 Hours is admittedly a bit arbitrary, but remember this is for people checking their watchlists. 48 hours means that if you check your watchlist every day or so you can ignore such bot edits. I'd concede something a bit longer would be more logical, 7 days would mean this works for people who check their watchlists once a week. I don't see how 30 minutes would make much of a difference, you'd have to be editing almost at the same time as the vandal. ϢereSpielChequers 19:41, 27 January 2017 (UTC)

I'm not recommending 30 minutes (or any specific number).

I think you significantly overestimate the number of vandals that are reverted via someone checking a watchlist once a day. Most vandalism is handled by ClueBot, Huggle, and other action-within-seconds tools, and secondarily by RecentChanges patrollers.

Also, the problem (of "hidden" edits) can be completely solved (i.e., for all bots and all editors, rather than just one) by changing watchlist configurations. There are several configurations we could adopt as default for all editors that wouldn't produce this problem, including not hiding bot edits or showing all edits separately. WhatamIdoing (talk) 19:47, 3 February 2017 (UTC)

I don't see how you can say I overestimate something without knowing how big a problem I estimate something to be:) I'm aware that watchlisting has become much less important as a vandal fighting tool, I know that from experience with my own watchlist and discussions with others. I check my watchlist quite frequently, I don't see the vandalism that I used to, no dispute that less gets past the edit filters, bots and patrollers than once did. But clearly enough does get past to be seen as a problem by the people filing this Arbitration request. Not hiding bot edits on its own doesn't solve the problem as you potentially need to check the history of all those articles. expanding the watchlist display to cover all edits should work fine for those with short watchlists and who don't watchlist very active pages. It wouldn't work for me and I doubt would work for those who raise the complaints, if their watchlists were short enough for that to work they would probably have taken that option already. ϢereSpielChequers 17:06, 4 February 2017 (UTC)

Proposals by User:Magioladitis

Proposed findings of fact

Ramaksoud2000

1) Ramasoud200's contribution to Wikipedia consists almost exclusively of reversions.

Comment by Arbitrators:

Comment by parties: We have to examine why Ramaksoud2000 without any prior interaction with me decided to go to Arbcom. This is worrying behaviour. The findings are in consistent to my evidence. -- Magioladitis (talk) 10:26, 24 January 2017 (UTC)

MSGJ There a lot of reverters in Wikipedia who dedicate their dy only in reverting edits. I 've been doing this for some days and then I had the impression that every single anonoymous IP was a potential threat. I was a new page patroller for a time and I stopped when I noticed that there was a rally who is going to revert faster. That days every new edit was a pain because it meant I had more pages to check. -- Magioladitis (talk) 10:42, 24 January 2017 (UTC)

The evidence in fact support the "there are different types of editors". Conflicts occur because people have different experiences while interacting with Wikipedia. - Magioladitis (talk) 10:51, 24 January 2017 (UTC)

Comment by others:

This case is not about Ramaksoud2000 so you would do well to remove all such "findings". If you have a problem with Ramaksoud2000 then this is not the correct forum. — Martin (MSGJ · talk) 09:32, 24 January 2017 (UTC)

2) Ramasoud2000's communication with other consists almost exclusively via automated messages.

Comment by Arbitrators:

Comment by parties: This is mainly to support the concept of Ramaksoud's approach. -- Magioladitis (talk) 13:56, 2 February 2017 (UTC)

Comment by others:

Minor edits and vandalism

1) No actual proof that minor edits hide vandalism.

Comment by Arbitrators:

Comment by parties:

Comment by others:

This point interests me. I've seen several editors complain about this. Then someone shows them how to set their watchlists in Special:Preferences, and they are happy again. While everyone wants a "magically know what I'm interested, and only show me that" setting (for which, see the ORES filters), I'd honestly be surprised if experienced vandalfighters actually had problems finding their way past minor edits.

That said, I don't think it's likely that ArbCom will decide that a statement like this is relevant. WhatamIdoing (talk) 21:30, 24 January 2017 (UTC)

Bot edits do hide (potential) vandalism under certain conditions; see phab:T11790 for the problem and a proposed fix. Opabinia regalis (talk) 21:56, 24 January 2017 (UTC)

Bot edits (and minor edits, and several other things) "hide" all edits (some of which may be vandalism or otherwise unwanted) under certain conditions – from individual editors only, if and only if those individual editors (a) have certain settings for their watchlists and (b) only review edits via their watchlists. "I don't personally choose to display those edits on my own watchlist" (including "I never realized that the enormous forest of options at the top of Special:Watchlist meant that I might want to change the settings") is not the same thing as "Bot edits hide vandalism from the whole community". It doesn't really matter if I personally see those edits via a single, specific mechanism (watchlists); it only matters if someone (e.g., someone with Huggle) sees them. IMO Magioladitis is correct: ArbCom has been given no evidence of even a single instance of a bot edit hiding vandalism from the entire community. If someone wants ArbCom to agree that this happens in practice, then I think someone needs to produce some evidence of this happening RSN. (That evidence could, for example, be a diff of some vandalism, followed closely by a bot edit, and the vandalism persisting for a couple of weeks, plus one or more editors saying that they scrupulously check all changes to the vandalized article and only missed it because their watchlist preferences were set to hide any article most recently edited by a bot. It shouldn't be impossible to produce this evidence if it's actually a serious problem.) WhatamIdoing (talk) 03:55, 4 February 2017 (UTC)

Yobot's blocks

1) First Yobot's block in 2009 is not related with others.

Comment by Arbitrators:

Comment by parties:

Comment by others:

2) 2nd-4th Yobot's block were talk page related

Comment by Arbitrators:

Comment by parties:

Comment by others:

CHECKWIKI project

1) CHECKWIKI project is a prominent project by the community. Editors are encouraged to participate in it

Comment by Arbitrators:

Comment by parties:

Comment by others:

2) There are no concerns about BG19bot, Menobot, Dexbot and XqBot which work fine.

Comment by Arbitrators:

Comment by parties: This page is to present things bases on the Evidence. -- Magioladitis (talk) 16:00, 21 January 2017 (UTC)

Comment by others:

This is not correct.

• The first three sections of the talk page for Dexbot are about concerns with cosmetic edits. [21].
• I had been very impressed with BG19bot, which had been running much better than Yobot. However, an edit appeared on my watchlist the other day which showed that it is having some of the same errors. I am currently trying to wait for the bot operator to respond with a plan to fix that issue. It again seems to be a mixture of the bot making edits when it should skip them, and the bot intentionally carrying out edits based on a vague BRFA.

— Carl (CBM · talk) 15:46, 21 January 2017 (UTC)

Proposed principles

CONSENSUS > COSMETICBOT

1) Consensus on performing a series of edits overwhelms COSMETICBOT

Comment by Arbitrators:

Magioladitis, by this do you mean that the consensus on a series of edits can supersede or outrank the policy? If so, do you think that is because the policy is flawed, or is being understood improperly, or more that there are times it should be overruled in an IAR-type way in individual circumstances? And if the latter, how do we decide what those circumstances are? Newyorkbrad (talk) 01:47, 12 January 2017 (UTC)

As an empirical matter, it's clearly the case that there's a lot of variability in what people mean by "cosmetic". And it's also the case that, with the working definition of "something that doesn't affect the rendered page", there are legitimate and uncontroversial bot tasks that make such edits. (Well... let's at least say "not controversial because of cosmeticness" ;) However, I'm unclear on how this applies to the pattern of facts in the current case, which seem to involve instances of "cosmetic" editing without explicit consensus. Is this just a statement for background, or did I miss something? Opabinia regalis (talk) 07:56, 17 January 2017 (UTC)

Comment by parties:

Newyorkbrad I mean this: Wikipedia_talk:Bot_policy#Add_a_footnote_that_the_community_can_allow_a_certain_cosmetic-only_edit_by_consensus_in_COSMECTICBOT_section. COSMETICBOT right now i far from being explicit. I also tried other approaches on the same page. BURob13 opposes because they disagree with "blanket statements". COSMETICBOT is by far a "blanket statement". If there is something to work is to make the restrictions more explicit. Otherwise, even the approval for the bot that removed Persondata as sole edit should not have been approved and should have been done "in addition to other edits" as some people demand(ed) from Yobot. Under COSMETICBOT right now we can't even mass rename a bad filename etc. Even teh task to remove duplicated arguments from templates conflicts with COSMETICBOT. There are so many examples for that. -- Magioladitis (talk) 08:18, 12 January 2017 (UTC)

Comment by others:

You have to be careful what you think is consensus. BRFA is a discussion between bot users, that non-bot users never pay attention to. If the bot is affecting 1000s of pages on the site and the humans who edit those pages start complaining, that means BRFA approval didn't actually reflect a community consensus. So you should stop the bot and open a wider discussion before continuing. 50.0.136.56 (talk) 07:53, 11 January 2017 (UTC)

Actually, I opposed because it's already common practice and obvious from WP:CONSENSUS. I only opposed the blanket statement of allowing all maintenance category related edits as an exception to COSMETICBOT. I have no problem with specific tasks being approved by the community over WP:COSMETICBOT. That doesn't mean COSMETICBOT is flawed. It means that, like any policy, we have a default assumption that cosmetic-only tasks are not okay, and the community can find consensus for exceptions. WP:CONSENSUS is indeed stronger than WP:COSMETICBOT or any of our other policies (except those with legal considerations), but that's irrelevant to this case, where you never obtained broad community consensus to allow either (a) a high rate of error on Yobot, above and beyond what can be expected from typical bot operation, and; (b) regular cosmetic-only edits from your main account without consensus that such edits can be made with AWB. Note that CONSENSUS > COSMETICBOT (which I broadly agree with) means consensus specific to using AWB or a bot to implement a type of cosmetic edit, not just consensus that the edit is valid. As noted on my evidence talk section, there's lots of precedent for situations where an individual edit is good, but the circumstances in which they're made become not-so-good. But then again, this whole thing is an argument about policy alterations, something ArbCom should not intrude upon. ~ Rob₁₃^Talk 20:17, 13 January 2017 (UTC)

@Opabinia regalis:

In this case we are mostly dealing with edits that are inadvertently only cosmetic, and generally few in number. COSMETICBOT is really a red-herring - the substantive issue is that the bot made a few edits that did not meet the stated goal.

Similar edits have occurred over a number of years. This sort of thing can happen despite worthy efforts to avoid it. For example the Blue Screen of Death still occurs when running Microsoft operating systems, some 23 years after its first appearance. (Ships are still sinking, planes are still crashing.) Such scenarios are the results of a class of failure, rather than a specific bug.

All the best: Rich Farmbrough, 23:49, 17 January 2017 (UTC).

This is all true enough, but not directly pertinent to the question I was asking. (I'm also not sure it's a good idea to compare errors in a commercial OS to errors in software developed and used by volunteers.) The contention AFAICT is that Magioladitis is taking insufficient precautions to avoid "cosmetic" edits under circumstances where it's known that consensus is against them. This proposal states that under some circumstances, cosmetic edits do have consensus. True, but not directly relevant unless some of Magioladitis' tasks actually fall under that second set of circumstances where cosmetic edits are at least tolerated. I was asking for clarification because I might have missed an example where that's the case. Opabinia regalis (talk) 23:58, 18 January 2017 (UTC)

@Opabinia regalis:

I was answering the part "However, I'm unclear on how this applies to the pattern of facts in the current case..." - shorter version is "COSMETICBOT is really a red-herring as far as this case is concerned."

All the best: Rich Farmbrough, 22:12, 19 January 2017 (UTC).

The idea of this is correct. The issue with this RFARB is that there is not consensus in these particular cases for the edits that are being made. However, if there was a BRFA that specifically authorized a particular edit, backed by a discussion demonstrating clear consensus, then COSMETICBOT would not apply to that edit - I have not seen anyone disagree with that idea. — Carl (CBM · talk) 19:35, 19 January 2017 (UTC)

No, you're right. We all agree with that idea in principle. But in practice, we'll claim that previous discussions don't show "clear consensus" for whatever it is that we disagree with, or that the "wrong people" made the decision (e.g., see the IP's comments about "BRFA is a discussion between bot users, that non-bot users never pay attention to", which is wrong, because I'm not a bot user and I sometimes participate in those discussions). WhatamIdoing (talk) 21:35, 24 January 2017 (UTC)

Wikipedia is based on good faith

1) Every editor and every administrator should act in good faith. Everyone is here to help for build a digital Encyclopedia.

Comment by Arbitrators:

Comment by parties:

Comment by others:

I don't think there's been any real bad faith or assumptions of it in this case. There's been good intentions but poor judgement going on for a long time, and imo there hasn't been enough sensitivity to broader bot-vs-human tensions in the editing community on the policy and DR side. With luck, this case can relax the conflict without having to hammer anyone. 50.0.136.56 (talk) 11:19, 18 January 2017 (UTC)

While I won;t go hunting for specific language, there has been a willingness to impute a "carry-on regardless" attitude. Fortunately this seems at odds with the record.

There has also been general ABFing around some of Magioladitis proposals defining COSMETICBOT, suggesting that people in general would attempt to wiki-lawyer them into carte blanche.

All the best: Rich Farmbrough, 19:50, 18 January 2017 (UTC).

The "carry-on-regardless" attitude is well illustrated by evidence here and in other cases such as yours. It's an observation not an imputation. 50.0.136.56 (talk) 22:32, 18 January 2017 (UTC)

• My case is not relevant, I may or may not be "uncivil and unresponsive", that has no bearing on Magioladitis behaviour.
• Magioladitis has pointed to specific tickets he has raised, and I could add many bug tickets before Phabricator raised by him, to address issues raised to him.
• Magioladitis has fixed issues raised with him, either in his settings, or in the code itself.
• Magioladitis has encouraged others to take over or share his tasks.
• Magioladitis has encouraged discussion about what changes bots should make.

It is perhaps easy to look at those instances where a bug or a timing error caused a handful of changes not to take effect, and assume that the remaining minor changes were the intended change. From there it would indeed follow that he was being recalcitrant. But it is patently clear that the assumption is false.

All the best: Rich Farmbrough, 22:54, 19 January 2017 (UTC).

There is a reason that BRFAs are flexible

1) Minor changes and tweaks to the bot behavior usually do not need to be reviewed by the community at large.

Comment by Arbitrators:

Comment by parties: This is added as principle. -- Magioladitis (talk) 13:23, 22 January 2017 (UTC)

Comment by others:

• Adding entirely new fixes to a task that are wholly unrelated to existing aspects of the task are not "minor changes and tweaks", however. ~ Rob₁₃^Talk 10:32, 22 January 2017 (UTC)

Proposed remedies

Ramaksoud2000 reminded

1) Before engaging more people in a discussion, try approaching the person in question.

Comment by Arbitrators:

Comment by parties:

Before I filed this case, you unfortunately declined to answer my questions at Wikipedia:Bot Approvals Group/nominations/Magioladitis 2, provided an unsatisfactory response at User_talk:Ramaksoud2000/Archive_3#subst_in_references, and declined to answer my question on your talk page at User_talk:Magioladitis/Archive_30#Automated_editing. Ramaksoud2000 ^{(Talk to me)} 23:33, 18 January 2017 (UTC)

I replied to your question my by BAG re-nomination in an appropriate way. Your questions here on the other hand were unrelated and most important you have already opposed by BAG re-nomination before my reply. It's also worth noting that the questions were also bases on conclusions. Exactly as the evidence I brought say.

In my talk page Bgwhite replied for me. If you check my talk page history a lot of people reply instead of me exactly because they understand how much time the whole process takes. -- Magioladitis (talk) 01:02, 19 January 2017 (UTC)

RU Rob13 who is "you" and who is "he"? -- Magioladitis (talk) 09:40, 19 January 2017 (UTC)

Comment by others:

One can only approach you so many times about the same issue before it's time to escalate. He listed over 25 of those times in the initial case request. Every editor on the project doesn't need to issue someone a warning before we're able to try to solve the issue. ~ Rob₁₃^Talk 06:53, 19 January 2017 (UTC)

They were not about the same issue. All the best: Rich Farmbrough, 22:56, 19 January 2017 (UTC).

Proposals by User:BU Rob13

Proposed principles

Purpose of Wikipedia

1) The purpose of Wikipedia is to create a high-quality, free-content encyclopedia in an atmosphere of camaraderie and mutual respect among contributors. Contributors whose actions are detrimental to that goal may be asked to refrain from them, even when these actions are undertaken in good faith; and good faith actions, where disruptive, may still be sanctioned.

Comment by Arbitrators:

Comment by parties:

Comment by others:

From Wikipedia:Arbitration/Requests/Case/The Rambling Man. ~ Rob₁₃^Talk 20:28, 13 January 2017 (UTC)

Good faith and disruption

2) Inappropriate behavior driven by good intentions is still inappropriate. Editors acting in good faith may still be sanctioned when their actions are disruptive.

Comment by Arbitrators:

Comment by parties:

Comment by others:

From Wikipedia:Arbitration/Requests/Case/Michael Hardy. ~ Rob₁₃^Talk 20:30, 13 January 2017 (UTC)

Isn't this redundant to the previous? I see nothing covered here but not there. עוד מישהו Od Mishehu 09:50, 15 January 2017 (UTC)

I don't disagree but these are usually separate principles in past cases. A sensible merge would be fine. ~ Rob₁₃^Talk 01:10, 16 January 2017 (UTC)

These are from separate cases, which explains why they can both be found; I doubt you could find a single case with both. עוד מישהו Od Mishehu 04:06, 16 January 2017 (UTC)

@Od Mishehu: See principles 1 and 4 from The Rambling Man. ~ Rob₁₃^Talk 23:29, 16 January 2017 (UTC)

Administrator conduct

3) Administrators are expected to lead by example and to behave in a respectful, civil manner in their interactions with others. Administrators are expected to follow Wikipedia policies and to perform their duties to the best of their abilities. Occasional mistakes are entirely compatible with adminship; administrators are not expected to be perfect. However, sustained or serious disruption of Wikipedia is incompatible with the status of administrator, and consistently or egregiously poor judgment may result in the removal of administrator status.

Comment by Arbitrators:

Comment by parties:

Comment by others:

From Wikipedia:Arbitration/Requests/Case/Rich Farmbrough. ~ Rob₁₃^Talk 20:20, 13 January 2017 (UTC)

Actually, shorter and more succinct version from Wikipedia:Arbitration/Requests/Case/The Rambling Man. ~ Rob₁₃^Talk 20:27, 13 January 2017 (UTC)

Automation tools

4) An automation tool is a technology designed to facilitate making multiple similar edits that would be unduly time-consuming or tedious for a human editor to perform manually. Common automation tools include bots (independently running processes that modify Wikipedia content in a fully or partially automated fashion), scripts (software components utilized to automate or semi-automate certain types of editing), and various other technologies.

The use of automation tools on Wikipedia is subject to numerous restrictions, and certain tools require approval from the Bot Approvals Group before an editor may use them.

Comment by Arbitrators:

Comment by parties:

Comment by others:

From Wikipedia:Arbitration/Requests/Case/Rich Farmbrough. ~ Rob₁₃^Talk 20:24, 13 January 2017 (UTC)

This of course was nonsense then and is nonsense now. The use of gadgets is not subject to Bot Policy, for example. Nor is search and replace.

Moreover it is pretty irrelevant, now, as then. We are here talking almost exclusively about AWB.

All the best: Rich Farmbrough, 22:34, 16 January 2017 (UTC).

Yobot is an AWB bot, and it's clearly at issue here. WP:BOTASSIST is part of the bot policy but impacts non-bot accounts. AWB is a script. This is wholly relevant to the current case. ~ Rob₁₃^Talk 23:31, 16 January 2017 (UTC)

AWB is not a script. The definition of script in the principle is wrong, the definition of bot is sloppy. The definition of "automation tool" is too narrow in some respects, it doesn't include other reasons for automation, such as timeliness or accuracy. Automation can also be used for single edits.

All the best: Rich Farmbrough, 00:32, 18 January 2017 (UTC).

I agree with Rich about this. Finding a precise definition of bots or automation that lets us identify what's being used in a given edit is impossible and irrelevant. The issue is the many-layered disruption resulting from prolonged fast editing no matter how it's done (wp:meatbot). I wish we could get away from this useless hairsplitting and just define "fast" as anything that edits more than N pages a day, and make it subject to BOTPOL or an analog of it. 50.0.136.56 (talk) 11:24, 18 January 2017 (UTC)

"Common automation tools include X, Y, and Z" could be read as saying "All X, Y, and Z are automation tools", which is wrong. It might be clearer to say that "Some X, Y, and Z are automation tools". Also, the definitions are poor (e.g., mw:Citoid is a "software component[] utilized to automate or semi-automate certain types of editing", but it's a MediaWiki extension, not a script). OTOH, those poor definitions might be sufficient for the purposes of an ArbCom case. WhatamIdoing (talk) 21:42, 24 January 2017 (UTC)

Users of automation tools

5) Like administrators and other editors in positions of trust, users of automation tools have a heightened responsibility to the community, and are expected to comply with applicable policies and restrictions; to respond reasonably to questions or concerns about their use of such tools; and to respect the community's wishes regarding the use of automation.

An editor who misuses automation tools—whether deliberately or in good faith—or fails to respond appropriately to concerns from the community about their use may lose the privilege of using such tools or may have such privilege restricted.

Comment by Arbitrators:

Comment by parties:

Comment by others:

From Wikipedia:Arbitration/Requests/Case/Rich Farmbrough. ~ Rob₁₃^Talk 20:24, 13 January 2017 (UTC)

Since the definition of "automation tool" - an ugly and clunky neologism if ever there was one - is so deeply flawed, anything using that term should be thrown on the same scrap heap.

All the best: Rich Farmbrough, 22:37, 16 January 2017 (UTC).

I thought it was rather clear. A tool which automates. ~ Rob₁₃^Talk 23:33, 16 January 2017 (UTC)

So you would argue that users of tabbed browsers "have a heightened responsibility to the community"?

All the best: Rich Farmbrough, 19:59, 18 January 2017 (UTC).

The issue is editing in a fashion that an unassisted human would not. Thus from ARBRF, "For the purposes of this remedy, any edits that reasonably appear to be automated shall be assumed to be so." That principle is perfectly workable if treated sensibly (NOTBURO). In your instance (the 1 year block) it wasn't treated sensibly, which is unfortunate, but shouldn't be seen as saying the principle itself was wrong. The AE environment has improved since then. 50.0.136.56 (talk) 22:30, 18 January 2017 (UTC)

Since when does a tabbed browser automate? If Chrome can make automated edits for me, please do tell me how. ~ Rob₁₃^Talk 23:05, 18 January 2017 (UTC)

There we have it. "Automation tools" are not defined as making edits for you, rather facilitating multiple edits. All the best: Rich Farmbrough, 23:26, 18 January 2017 (UTC).

"Designed to facilitate", meaning it must be an intended purpose of the tool. If you can point me to some evidence that Chrome's developers sat down and discussed how to add tabs so Wikipedians can edit faster, I'll gladly withdraw that definition. ~ Rob₁₃^Talk 23:29, 18 January 2017 (UTC)

Tabbed browsing is designed to facilitate faster use of the web: editing is a specific example. Moreover with enough oomph it is possible to edit much faster with tabbed browsing than with AWB, rendering any distinction that includes AWB but excludes tabbed browsing ludicrous on its face.

All the best: Rich Farmbrough, 23:03, 19 January 2017 (UTC).

That's simply untrue. AWB can do complicated edits at rates of at least 30/minute. Tabbed editing cannot edit at a sustained rate anywhere approaching that. And "designed to make browsing faster" is quite different from "designed to facilitate multiple edits on Wikipedia". ~ Rob₁₃^Talk 23:40, 19 January 2017 (UTC)

The devs for all of the major web browsers consider Wikipedia in their design. A browser that works poorly on popular websites like this one is not going to be successful. MediaWiki devs (staff and volunteers) routinely push bug reports "upstream" to browser developers. Wikipedia really is something that web teams look at in great detail: they all want to be the fastest, smoothest, best environment for visiting Wikipedia.

That said, the use of tabs in web browsers pre-dates Google Chrome by about a decade, so they probably made that choice because all the other popular browsers were already doing it, rather than because it would be handy for Wikipedia. But you really should not underestimate the value that browser devs put on this website. WhatamIdoing (talk) 21:53, 24 January 2017 (UTC)

Oh, I definitely agree. It's just also true/obvious that tabbed browsing was clearly intended as a usability/feature add so users didn't need multiple windows open, not something explicitly for Wikipedia. ~ Rob₁₃^Talk 22:07, 24 January 2017 (UTC)

Involved administrators

6) Administrator tools are not to be used in connection with matters in which the administrator is involved. In circumstances where an administrator is involved, the administrator should not take administrative action but should instead report the issue to a relevant noticeboard, perhaps with a suggestion for appropriate action, to be dealt with by another administrator. In limited circumstances, such as blatant vandalism or bad-faith harassment, an involved administrator may act, but such exceptions are likely to be rare.

Comment by Arbitrators:

Comment by parties:

Comment by others:

From Wikipedia:Arbitration/Requests/Case/Rich Farmbrough. Alteration made to the heading for simplicity. ~ Rob₁₃^Talk 20:24, 13 January 2017 (UTC)

This is about the unblocks? I wouldn't go overboard about that, and I'm one of the bigger bot critics here. I like Iridescent's analysis. 50.0.136.56 (talk) 11:26, 18 January 2017 (UTC)

This fails to comprehend that the blocks and unblocks weren't about a dispute. They were simply a mechanism to stop and restart the bot. All the best: Rich Farmbrough, 23:29, 18 January 2017 (UTC).

Fair enough. Changed "disputes" to "matters", which is equally true. ~ Rob₁₃^Talk 23:41, 19 January 2017 (UTC)

Clearly there are conflicting opinions as to whether unblocking a bot once you think you've fixed the bug is one of the acceptable exceptions. I've offered a compromise but I suspect we need an RFC on this. I suggest that the invitation to block the bot freely - I'll unblock once I've fixed the bug needs to get explicit consent from the blocking admin. ϢereSpielChequers 19:51, 27 January 2017 (UTC)

Misuse of administrative tools

7) Administrative tools are provided to trusted users, and should be used with thought. Per WP:TOOLMISUSE, using the tools to reverse the actions of other administrators, such as unblocking a bot which is believed to be violating bot policy, should not be done without good cause, careful thought and usually some kind of consultation.

Comment by Arbitrators:

Comment by parties:

Comment by others:

From Wikipedia:Arbitration/Requests/Case/Rich Farmbrough. ~ Rob₁₃^Talk 20:24, 13 January 2017 (UTC)

Outdated groupthink from half a decade ago. All the best: Rich Farmbrough, 22:39, 16 January 2017 (UTC).

... stating that we shouldn't reverse admin actions without consulting the admin who made the action is groupthink? No. It's both common courtesy and standard practice to prevent wheel-warring situations. ~ Rob₁₃^Talk 23:32, 16 January 2017 (UTC)

Have you even read any of the previous discussion I have pointed out? All the best: Rich Farmbrough, 20:01, 18 January 2017 (UTC).

Improperly unblocking a bot might be tool misuse some of the time, but it should generally be ok to block a bot at a much lower bar than blocking a human. Similarly, a mistaken block of a bot is a much smaller error than one of a human. 50.0.136.56 (talk) 11:28, 18 January 2017 (UTC)

I can support the sentiment and ideas behind that, even if the exact wording has issues. It should be realised, though, that people will abuse block history to "win" in disputes. All the best: Rich Farmbrough, 23:31, 18 January 2017 (UTC).

Role of the Arbitration Committee

8) The Arbitration Committee may not create, alter, or destroy policies or guidelines established by community consensus. The community may alter existing policies as laid out in the procedural policy for doing so, but the Arbitration Committee decides cases based on the policies or guidelines in effect at the time alleged violations occur.

Comment by Arbitrators:

Comment by parties:

Still it would be good if the ArbCom states the obvious: The COSMETICBOT is very broad and open to interpretations. The ArbCom can underline the fact that it comes into conflict with other rules. -- Magioladitis (talk) 18:26, 19 January 2017 (UTC)

Comment by others:

Originally written. The amount of editors who provided "evidence" claiming that ArbCom should discard an entire section of the bot policy because a few people dislike it was staggering. It is important to assert the actual role of the Arbitration Committee, which is to interpret existing policies (based on relevant community discussion) and enforce them as necessary to limit disruption. The Arbitration Committee should be especially careful to avoid destroying policy by failing to enforce it in situations where the community fails to do so. ~ Rob₁₃^Talk 18:16, 19 January 2017 (UTC)

@Magioladitis: See my finding of fact #1. While I doubt you'll appreciate my other finding of facts, that one I wrote specifically in an attempt to compromise on the "confusing" bit. It states that editors unfamiliar with the policy may find it unclear, and I plan to also propose a remedy with the Committee encouraging the community to review the section. I don't personally think either is necessary, but perhaps making it even more obvious will prevent this in the future. ~ Rob₁₃^Talk 18:41, 19 January 2017 (UTC)

Proposed findings of fact

Interpretation of COSMETICBOT

1) The bot policy states that "Cosmetic changes (such as some of the AWB general fixes) should be applied only when there is a substantive change to make at the same time." In the absence of a specific definition of "cosmetic", the word may be interpreted as commonly defined in the English language and through consensus. Merriam-Webster [defines https://www.merriam-webster.com/dictionary/cosmetic] "cosmetic" to mean "done or made for the sake of appearance", and this is the common usage of the word in the English language. The community has commonly defined "cosmetic-only" to mean "edits which do not alter the rendered appearance of the page", but there are exceptions to this. The current definition may be somewhat confusing to editors unfamiliar with its application.

Comment by Arbitrators:

Comment by parties:

"done or made for the sake of appearance" is a good definition probably because it excludes edits such as closing unclosed tags and other potential errors. Maybe this is a good definition afteral. None of the CHECKWIKI tasks are done just for the sake of (the code?) appearance. -- Magioladitis (talk) 18:14, 19 January 2017 (UTC)

The reason that we added this, is that we noticed that history merge in other Wikipedias was causing errors. -- Magioladitis (talk) 19:13, 19 January 2017 (UTC)

Also explaining why every single error is in the CHECKWIKI list is not exactly my role since CHECKWIKI is a community project that a lot of people contribute and the list is affected by discussions between Wikipedia editors. -- Magioladitis (talk) 22:46, 19 January 2017 (UTC)

Comment by others:

I hope this is something everyone can agree with, including Magioladitis. I do not personally find the definition to be vague, but perhaps new editors would find it vague, so I have conceded that. I recall being somewhat confused about it early on, although I believe competent good-faith editors can figure it out by asking a couple questions and taking constructive criticism. ~ Rob₁₃^Talk 18:11, 19 January 2017 (UTC)

I'd love to hear you explain how changing {{tl|Template:TEMPLATENAME}} to {{tl|TEMPLATENAME}} is anything other than done for the sake of appearance of the wikicode. Same for removing underscores from wikilinks, etc. These tasks literally do not change the visual output of the page. They only change the appearance of the code. That is their sole purpose. ~ Rob₁₃^Talk 18:56, 19 January 2017 (UTC)

I don't know. It's possible that correcting the redundant namespace has a slight performance benefit. It's true that we don't have to worry about WP:PERF (unless Ops tells us to), but editing something to improve performance isn't actually prohibited. WhatamIdoing (talk) 22:08, 24 January 2017 (UTC)

I love this proposal, because it so neatly encapsulates the confusion. According to the dictionary, a cosmetic edit is one that does change the appearance (only), and the community prohibits edits that don't change the appearance – on the grounds that these non-appearance-changing editors are merely cosmetic changes.

Also, the "quoted" definition that the community allegedly uses so commonly cannot be found on any page except this one. If we're going to "quote" a definition, then let's quote a definition that is widely accepted enough to be found in a policy or guideline on the subject. And if none exists, then let's not quote it – or even give it too much attention. WhatamIdoing (talk) 22:08, 24 January 2017 (UTC)

@WhatamIdoing: Cosmetic refers to the cosmetics of the wikicode. It is contrasted with "substantive edit" in the policy. Interpretation of any policy is subject to consensus on what it means, and Magioladitis had half a decade of community discussion about his editing to learn how the community applies the policy. When a large-scale discussion at AN results in "Don't do this", an editor (especially an admin) must stop doing that. That's how consensus operates. I agree the definition in the policy could use some work, but an administrator who runs bots, was a member of the Bot Approvals Group, and was the subject of many large-scale discussions at community noticeboards on this topic is expected to learn something, not cover their ears and feign ignorance. It's either a competence issue or disruptive editing after half a decade of activity. Either way, restrictions have become necessary. ~ Rob₁₃^Talk 23:58, 24 January 2017 (UTC)

Interpretation of AutoWikiBrowser Rules of Use

2) The AutoWikiBrowser Rules of Use #4 states "Do not make insignificant or inconsequential edits. An edit that has no noticeable effect on the rendered page is generally considered an insignificant edit. If in doubt, or if other editors object to edits on the basis of this rule, seek consensus at an appropriate venue before making further similar edits." This defines edits which fail to change the rendered page as an insignificant edit, but this definition is not exhaustive. If objections are raised to certain types of edits, the onus is placed on the editor making edits with AWB to gain consensus that the edits do not violate this rule before continuing.

The Rules of Use section makes clear that repeated violations of the rules will result in losing access to AutoWikiBrowser.

Comment by Arbitrators:

Comment by parties:

Comment by others:

These rules were imposed by the early designers and maintained by the AWB community, primarily for the purpose of preventing AWB falling into disrepute. While I support them, they relate to the software/en:wp:project and do not have the force of community rules such as policies.

All the best: Rich Farmbrough, 23:09, 19 January 2017 (UTC).

The community decided to implement WP:PERM/AWB and provide administrators the power to revoke AWB permissions for violations of the rules of use. That is a community endorsement of the rules of use. ~ Rob₁₃^Talk 23:42, 19 January 2017 (UTC)

PERM/AWB doesn't mention the rules of use or anything about revoking permissions. The previous system said that applicants had to agree to follow the rules, but the new "community endorsed" version doesn't. WhatamIdoing (talk) 22:16, 24 January 2017 (UTC)

AutoWikiBrowser access for administrators

3) AutoWikiBrowser is enabled by default for administrator accounts. As such, access to AutoWikiBrowser is an administrative tool. Based on current policy, abuse of a single administrative tool may result in de-adminship.

Comment by Arbitrators:

Comment by parties:

Comment by others:

See also WP:ROLLBACK, where it explicitly states abuse of rollback may lead to de-adminship as a whole. This is similar. ~ Rob₁₃^Talk 18:39, 19 January 2017 (UTC)

I don't like this definition at all. Admins can be de-sysopped for abusing anything. That doesn't mean that anything they can use is an admin tool. An admin tool is a tool that only admins (or all admins plus relevant similar user groups) have access to. WhatamIdoing (talk) 22:19, 24 January 2017 (UTC)

Glad to see that this is still "may result" not "will result" or "should result". ϢereSpielChequers 19:56, 27 January 2017 (UTC)

Cosmetic-only edits

4) Cosmetic-only edits have been made regularly by both the Magioladitis and Yobot accounts. Examples of cosmetic-only edits can be found in BU Rob13's evidence section for the Magioladitis account and the archives of the Yobot account's talk page.

Comment by Arbitrators:

Comment by parties:

This sentence lacks definition of cosmetic. Moreover, if we agree that consensus > cosmeticbot, this may actually be OK. -- Magioladitis (talk) 08:11, 24 January 2017 (UTC)

I really encourage the Arbs to look the "famous Yobot archives" carefully so that they can see which complains were about bugs, which were just misunderstandings and which were "cosmetic-only" runs. I provided evidence that the nature of the complains varies. I recent example of a compain that was only a misunderstanding can be found here. Note that since Yobot stopped complains were transferred. -- Magioladitis (talk) 07:49, 25 January 2017 (UTC)

Comment by others:

Will follow this up with a principle relevant to the Arbitration Committee's role here. ~ Rob₁₃^Talk 18:11, 19 January 2017 (UTC)

Regularly? Perhaps you mean "frequently".

All the best: Rich Farmbrough, 23:10, 19 January 2017 (UTC).

Indeed. ~ Rob₁₃^Talk 23:43, 19 January 2017 (UTC)

I had a look at all of the links in BR Rob13's section on the /Evidence page. They are:

• [22] [23] [24][25][26] all fix the visible spacing between the end of the page and the stub tag, in compliance with WP:LAYOUT (and also make one other change). That means that these edits are not cosmetic-only according to Rob's definition of something that changes what the reader sees.
• [27][28][29][30][31][32] all replace a redirect (Template:Refs) with the most common template name (Template:Reflist).

If this is the only "cosmetic" issue, then it seems to me that a far more appropriate approach would be to have an RFC in which editors decide whether or not this is a violation of COSMETICBOT and update the policy accordingly. Given the determination that some editors have to use the (pointless and marginally slower) reflist template when none of its features are being used, specifically on the grounds that if it's in all the articles, it'll become familiar to new editors), then it's possible that we would be surprised and discover that this specific type of edit really was wanted. WhatamIdoing (talk) 22:31, 24 January 2017 (UTC)

@WhatamIdoing: My section is very light on the evidence of cosmetic-only editing because it was covered by other evidence sections when I returned from a vacation and began to compile evidence. My evidence focused on issues that weren't yet covered to avoid redundancy. ~ Rob₁₃^Talk 23:54, 24 January 2017 (UTC)

• Your section was 50% non-cosmetic edits being mis-identified as cosmetic-only edits. And the other half may actually have consensus, because The Community™ decided a few years ago that they wanted {{reflist}} itself practically everywhere, even where its use is pointless.
• This proposal specifically says that the evidence can be found (only) in that section on the /Evidence page. If you intended to include other evidence (e.g., indisputable violations from recent history), then you'll have to re-write it. WhatamIdoing (talk) 05:01, 25 January 2017 (UTC)

• "and the archives of the Yobot account's talk page". I'll allow the arbitrators to word where the evidence is as they see fit. Most of the evidence, honestly, is in the 25+ discussions linked in the original case request. ~ Rob₁₃^Talk 05:25, 25 January 2017 (UTC)

Previous dispute resolution

5) Many administrators and editors have attempted to individually address bot policy and AWB rules of use issues with Magioladitis over a period of at least six years. Additionally, Magioladitis and Yobot were the subject of discussion at the Administrators' noticeboard/Incidents and the Administrators' noticeboard multiple times. These discussions are listed in the initial case request.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Awareness of community standards

6) Based on previous dispute resolution and his roles as an administrator, member of the Bot Approvals Group, and developer for AutoWikiBrowser, a reasonable editor in Magioladitis' position would be aware of the community's standards for AWB and bot edits. At a previous AN discussion, Magioladitis responded clearly to issues related to the bot policy and AWB rules of use. Ramaksoud2000's evidence section demonstrates Magioladitis' awareness of various types of cosmetic-only editing based on comments made from 2010 to 2016.

Comment by Arbitrators:

Comment by parties:

From the findings it is shown that certain types of edits that look minor are important. Community standards are no clearly shown in today's version of COSMETICBOT and it's nice that people for the first people tend to agree that the role of consensus should be underlined. -- Magioladitis (talk) 22:44, 19 January 2017 (UTC)

Ramaksoud2000 Again not all blocks are related and not all blocks are of the same nature. The evidence are clear in that. -- Magioladitis (talk) 22:51, 24 January 2017 (UTC)

Comment by others:

It is important to acknowledge that, even if the word of the policy was not clear to him, an editor in Magioladitis' position should be reasonably aware of the standards of the community. It is unacceptable for an editor, let alone an admin, to attempt to wikilawyer their way out of responsibility for editing against community consensus based on claiming ignorance of the definition of "cosmetic" even after it's been explained repeatedly to him. I'm falling short of proposing a finding of fact that Magioladitis has demonstrated an WP:IDONTHEARTHAT attitude, but I believe it's very true nonetheless. ~ Rob₁₃^Talk 18:55, 19 January 2017 (UTC)

I don't think that you should be making side accusations of "wikilawering" etc.. (I'm not sure how "very true" differs from "true" either.)

All the best: Rich Farmbrough, 23:13, 19 January 2017 (UTC).

I've declined to include that in a proposed finding of fact because I doubt the community would benefit from the Committee formalizing that reality, but the fact that wikilawyering has occurred by disputing the meaning of a word in the English language is very relevant to this case. ~ Rob₁₃^Talk 23:45, 19 January 2017 (UTC)

I find myself thinking about the difference between "some people complained, based upon their personal interpretations of what's cosmetic" and "the community agreed enough to get their standard written down in the policy". I see the first happening; I haven't seen the second happening. I have a hard time blaming someone for not agreeing that a couple of complainers represents the One True Voice of the Community™, especially given that people hold different views and are even self-contradictory on occasion (whether by accident or because they've changed their minds over time). And that's not counting the people who don't want bots to do any cosmetic edits at all, even when they're doing obviously substantive ones. WhatamIdoing (talk) 22:37, 24 January 2017 (UTC)

user:WhatamIdoing, running any type of bot on a non-bot account is completely prohibited. Running any unapproved bot tasks (bypassing template redirects) on a bot account is completely prohibited. Even if he never bothered to read WP:BOTPOL, do you not think that a reasonable editor would have re-evaluated the situation after the 23rd block? Ramaksoud2000 ^{(Talk to me)} 22:45, 24 January 2017 (UTC)

The policy says that "Contributors should create a separate account in order to operate a bot". It does not say that contributors must do this – only that it's recommended.

I think that if a bot operator had someone like Rob saying that adjusting the visible spacing between the end of the article and the stub tag is "purely cosmetic", and also someone like me telling the same bot-op that the same edit is not cosmetic, that I'd be entitled to be confused – no matter whether those opinions were delivered as unwarranted blocks or as friendly messages. The mere fact that Rob and I disagree about those edits is evidence that there's no consensus about whether that's cosmetic. WhatamIdoing (talk) 23:16, 24 January 2017 (UTC)

WhatamIdoing, you misunderstand. All bot tasks must be approved. That is absolute. No bot tasks that run on non-bot accounts are ever approved. Thus, no bot runs on non-bot accounts are ever allowed. The major issue is running unapproved bot tasks, both on his account and Yobot. Bot approvals exist to prevent cosmetic editing like bypassing template redirects. Nobody is allowed to circumvent the system. Pull up any edit. If he had approval, it's OK. If he didn't, it's not. Very simple. Ramaksoud2000 ^{(Talk to me)} 04:01, 25 January 2017 (UTC)

Where to begin?

• Bot tasks above a certain (e.g., non-testing) size need to be approved.
• The use of AWB in non-bot accounts is fairly common. You may argue that AWB isn't a bot, and you would be technically correct, but most of the diffs complained about in this long discussion seem to be using AWB and Twinkle (although that may not be clear to everyone).
• The bot approval process exists for many reasons. Preventing edits that are exclusively cosmetic is merely a minor goal for that process. OTOH, if an edit needs to be made for other reasons, then making cosmetic edits at the same time is generally acceptable.
• Sometimes, bypassing templates is necessary and desirable, e.g., just before the redirect is going to be deleted at TFD. So regardless of whether you consider bypassing a template to be cosmetic, bots are actually allowed to bypass templates, and under some circumstances, they're allowed to bypass them while making no other changes to the page. WhatamIdoing (talk) 05:19, 25 January 2017 (UTC)

Negligence as bot operator

7) Magioladitis has stated that he "run[s] Yobot in huge untested links" and claimed this was the only reason Yobot encountered errors. Bot operators are permitted to make limited changes to their bots without further formal approval, but they are responsible for testing changes and for all resulting edits. Based on Magioladitis' explanation, this practice continued even after repeated blocks of Yobot for the resulting errors. Running Yobot untested after changing how it operates or the list of pages it operates on was negligent, and this practice does not remove Magioladitis' responsibility for the edits of his bot.

Comment by Arbitrators:

Comment by parties:

Comment by others:

I somewhat doubt this will pass since it strays from the point a bit, but I think it should at least be discussed/considered by the Committee. It may very well be worth having a finding of fact formally stating that "but I ran the bot totally untested" is not a defense when faced with over half a decade of policy violations. In the face of errors and policy violations, the community expects a bot operator to permanently adjust how their bot operates (or, in this case, how they operate when running their bot) to address the issues. ~ Rob₁₃^Talk 19:13, 19 January 2017 (UTC)

Deletions

8) Magioladitis deleted several template redirects without discussion after changing all instances of the redirect to the main template's name (see BU Rob13's and Ramaksoud2000's evidence). This bypasses a deletion discussion at Redirects for discussion. Orphaning and deleting template redirects without discussion has no basis in any deletion process defined by the community. Deleting pages out-of-process to bypass a deletion discussion constitutes abuse of the administrative tools.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Two points. First, that is not to say this is a serious abuse of the administrative tools, just that it is abuse to bypass deletion discussions. Second, I'm not calling the deletion of the 400+ stub templates abuse because editors have shown that other administrators encouraged such deletions, and it appears to be an (undocumented) standard practice. Separately, I will propose a finding of fact relating to the speed at which deletions took place. ~ Rob₁₃^Talk 19:21, 19 January 2017 (UTC)

Rapid editing on main account

9) Magioladitis has edited on his main account at speeds far exceeding community standards for semi-automated editing with AWB (see BU Rob13's evidence). Editing at extremely high rates with semi-automated tools on a non-bot account constitutes a bot policy violation. Editing with AWB at rates which make it impossible to individually review each edit violate AWB Rules of Use #1, which states in part that editors must "review all changes before saving".

Comment by Arbitrators:

Comment by parties:

I never said I deleted them separately. I used the D-batch option the same way I deleted the templates closed by RU Rbo13 as "delete all" in a TfD. Twinkle allows it. Here is the list I used [33] -- Magioladitis (talk) 19:40, 19 January 2017 (UTC)

Comment by others:

Based on analysis of Ramaksoud2000's evidence. I'm falling short of formally accusing Magioladitis of running a fully-automated bot from his main account in this finding of fact, but how else would one do this? Even if he had 270 tabs open and Twinkle boxes open waiting for him to hit submit, it's impossible to click "Submit" at a rate of 9 deletions/second. This is physically impossible. I've checked the deletions, and it's not a case of multiple redirects being deleted with one click; they all had different targets. Not only does this look like an automated script; it looks like he intentionally included (TW) in the edit summary to obscure this fact. I know this is a rather bold/extreme claim, but I genuinely have no idea how else this many edits in this short a period of time makes sense. Note that I've limited the "end time" of edits appearing in this query to 23:16:30 (see URL), so these are all in the first half of the "23:16" minute. ~ Rob₁₃^Talk 19:37, 19 January 2017 (UTC)

@Magioladitis: Apologies. I was unaware of that feature and had not heard you previously mention it. I've rescinded my suspicions and will alter this finding of fact to only reflect edits made using AWB. ~ Rob₁₃^Talk 19:52, 19 January 2017 (UTC)

Superseded by later edits. I have some experience with this kind of thing, from when I did administrative work. I am don't think it is at all hard to do. Here is one way: make a list of the pages that should be deleted, and have a Twinkle applet that can perform the deletion with one click, using a stored deletion summary. Then go down the list, clicking all the pages. It is easy to click 10 in a second. With the right applet, it is not necessary to wait for one to finish before clicking the next one. This is not to say that the deletions followed best practices, but if there really was a consensus to delete these stubs (which is not well documented in the deletion summary), this is basically the way I would expect to see them deleted. — Carl (CBM · talk) 20:07, 19 January 2017 (UTC)

@CBM: As this is less than clear-cut, I removed all references to the deletions from this finding of fact. Instead, it focuses on other rapid editing using AWB. ~ Rob₁₃^Talk 20:12, 19 January 2017 (UTC)

OK, I will also strike out my comments as superseded. — Carl (CBM · talk) 20:14, 19 January 2017 (UTC)

Note that it is possible for a developer of AWB to keep their own version of that software that makes it run in bot mode no matter which account it is run from. This could be done as simply as setting a single flag differently or modifying an if-then statement and recompiling (note: I haven't reviewed the AWB code in this respect -- I just know that this can be done, as a former software developer who often used flags for changing software behavior). I don't say this to accuse Magioladitis of doing this, because I have zero evidence of it, but to highlight an inherit conflict of interest which doesn't allow us to determine in certain terms how AWB was used as a meatbot. We have to give the benefit of any doubt here and say that 'Save' was clicked about as quickly as it became enabled, without any or any reasonable edit review. Stevie is the man! ^{Talk • Work} 22:16, 19 January 2017 (UTC)

@Stevietheman: Very early on, before I was aware of WP:BOTASSIST, I made some edits along the lines of something that eventually became a bot task at the rate of clicking Save when it became available. (As soon as I was made aware of BOTASSIST, of course, I immediately stopped that practice. I was less than 2 months old as an editor at that point.) I can say that the speed of edits here is entirely consistent with that method of semi-automated editing as a "human bot" with no oversight of the edits whatsoever. I would say it's unlikely that a "bot mode" was enabled. ~ Rob₁₃^Talk 22:21, 19 January 2017 (UTC)

Totally understood. However, note that AWB's bot mode has a 'delay' function, which sets "The delay in seconds before saving the page after loading" A human bot can be mimicked. Again, this is just within the realm of possibility. Stevie is the man! ^{Talk • Work} 22:38, 19 January 2017 (UTC)

1. speeds far exceeding community standards for semi-automated editing

No community standard speed limit at that link.

1. Editing at extremely high rates with semi-automated tools constitute a bot policy violation.

No it doesn't. Bot policy specifically allows it.

All the best: Rich Farmbrough, 23:22, 19 January 2017 (UTC).

"In general, processes that are operated at higher speeds, with a high volume of edits, or are more automated, may be more likely to be treated as bots for these purposes." That explicitly takes speed and volume into account. No-one can accuse Magioladitis of having few edits, and the speed at times has been as fast as humanly possible using the tool, which is a rate of 30 edits/minute. Further, to edit at that rate with AWB, it is humanly impossible to review the contents of each edit, especially given that Magioladitis runs general fixes, which would require scrolling around to review multiple sections that have been edited, etc. As such, we have a situation where the process is being operated as automated as it can be without being automated, as high speed as it can be given the constraints of a human being and the program, and extremely high volume by the editor with the tenth most edits on the project. If this is not the time where BOTASSIST applies, when is? Citation needed on your claim that the bot policy allows editing at extremely high rates of speed with semi-auto tools. That appears to directly contradict BOTASSIST. ~ Rob₁₃^Talk 23:53, 19 January 2017 (UTC)

Rich is right. One of the main purposes of the bot policy is to allow high-speed editing. This is a case of WP:Policy writing is hard; you mean something reasonable, but what you wrote doesn't accurately convey your meaning. You probably meant to say something like "Unless you have permission, editing at extremely high rates with semi-automated tools constitute a bot policy violation." WhatamIdoing (talk) 05:23, 25 January 2017 (UTC)

Conduct unbecoming

10) Due to a failure to abide by the bot policy or the AWB Rules of Use and repeated disregard of requests by the community to alter his editing patterns, Magioladitis has demonstrated poor judgement and disrupted the project. Based on current policy on administrator conduct, such sustained behavior may be incompatible with adminship.

Comment by Arbitrators:

Comment by parties:

Comment by others:

failure to abide by the bot policy or the AWB Rules of Use not demonstrated. All the best: Rich Farmbrough, 23:23, 19 January 2017 (UTC).

Wikipedia:Arbitration/Requests/Case/Magioladitis/Evidence, including the sections that highlights Magioladitis' past acknowledgement that some of the edits from his account and the bot account which have persisted for years should not be occurring. This is a bit of a crazy thing to contest, given that past Magioladitis even agrees. ~ Rob₁₃^Talk 23:54, 19 January 2017 (UTC)

Bot unblocks

11) Magioladitis repeatedly unblocked Yobot. Yobot is an account operated by Magioladitis, and the unblocking policy explicitly states that "unblock[ing] one's own account" is never acceptable and may lead to desysopping on a first offense. Magioladitis is involved with his bot in an administrative sense.

Comment by Arbitrators:

I don't see a consensus either way. I'd like to see one coming out of the discussion generated by this case, but we can't make one. Doug Weller talk 19:26, 7 February 2017 (UTC)

Comment by parties:

Comment by others:

We have already thrashed this one out numerous times. Community consensus is that unblocking one's own bot after a technical fix is acceptable. All the best: Rich Farmbrough, 23:24, 19 January 2017 (UTC).

We have most certainly not found consensus for that at any centralized discussion. You would need to show a discussion with "widespread consensus" that this is acceptable to alter the unblocking policy and involved to include an exception for bot accounts, as per WP:PGCHANGE. ~ Rob₁₃^Talk 23:56, 19 January 2017 (UTC)

The "real policy" is what the community accepts as standard practice, in practice – not what's on the page. I think that the community has accepted this, even if it hasn't documented it on a particular page (yet). Also, as I commented above, it's not clear what value another admin adds here. Either you have an unfounded belief that he fixed the bot (in which case, it doesn't really matter who pushes the unblocking button – it's still happening on the bot-owner's say-so), or you have an unfounded belief that he didn't (in which case, you'll find out within minutes whether it's actually broken). You have no basis for your belief in either case, and there's no practical way around that problem. WhatamIdoing (talk) 22:45, 24 January 2017 (UTC)

Whether such unblocks are legit has been contentious in the past; I recall that the Rich Farmbrough (sp?) arb case had a FoF on the subject matter that was retracted later. I suspect that there isn't an agreement on whether it is appropriate to unblock in such circumstances. I submit that this proposed finding should not be included as it's trying to set down a policy where no consensus on one exists yet, which is not within the scope of Arbcom. Jo-Jo Eumerus (talk, contributions) 14:41, 7 February 2017 (UTC)

Bot blocks and evasion

12) Magioladitis made similar edits on his main account to Yobot's approved tasks while Yobot was blocked. Bot accounts are typically blocked for technical issues, and making similar semi-automated edits from another non-bot account is not block evasion. Such edits are not against any policy, provided that they abide by all other policies and rules regulating semi-automated editing.

Comment by Arbitrators:

Comment by parties:

Re to Carl: Not all Yobot jobs were in question. Moreover, the blocks were not for tasks in question but exactly because Yobot did not do the jobs e.g. simplifying a wikilink was expected and Yobot failed to fix that. -- Magioladitis (talk) 21:36, 22 January 2017 (UTC)

Comment by others:

This was a common claim in Evidence and the case request, but there's just no policy basis for it, and the Committee should acknowledge that. A block of a bot does not mean the contributor himself or herself is blocked from editing, and in the absence of such a block, there can be no evasion. ~ Rob₁₃^Talk 19:07, 22 January 2017 (UTC)

If a bot is blocked because of a disagreement over a particular task, the operator certainly cannot run the same bot job off their main account instead. The operator may need to make a limited number of edits as part of testing, but conducting large-scale runs at bot-like speed from the main account can't be acceptable when the bot was blocked exactly to stop those edits from being made. — Carl (CBM · talk) 19:13, 22 January 2017 (UTC)

Bot-like speed would be a bot policy violation, which is why I specified they must abide by all other policies. That includes WP:CONSENSUS, etc. My point is that if the quabbles are over a bot making cosmetic-only edits and Magioladitis takes up the task but skipped the cosmetic-only ones, it would not be a violation. If he failed to skip them, it would be an AWB rules of use violation, not block evasion. Our other policies cover this. ~ Rob₁₃^Talk 19:19, 22 January 2017 (UTC)

If the bot operator put in a bot request for a particular job - which is to say they accepted that the job should be done by a bot - and then they perform the task on their main account while the bot is blocked, I do view that as block evasion. Of course, there is a difference between running a bot job and just performing a few testing edits. But Magioladitis has been run ordinary Yobot jobs on his main account during the arbcom case, while Yobot was blocked; see the evidence talk page. — Carl (CBM · talk) 20:53, 22 January 2017 (UTC)

I don't think edits done by bots and edits done by humans must be mutually exclusive. For instance, I have a bot approved to close peer reviews, but clearly those edits could be done by a human. We have bots to clerk WP:CFD/W, but those edits could be done by humans (and often are when the bots break). Again, if the edits themselves are problematic, as they often were in this case, that's a separate issue. Same deal if the edits are in high volume without a bot flag, as that may be against the bot policy or consensus. There's no such thing as evading a block on specific types of editing though. If we consider an editor to be evading a block on their bot, that would mean they can't make any edits. A block of a bot isn't a pseudo-ban on all types of editing the bot did. ~ Rob₁₃^Talk 07:29, 23 January 2017 (UTC)

I agree with Rob. This proposal seems like an accurate statement of the actual policies and practices. WhatamIdoing (talk) 22:48, 24 January 2017 (UTC)

Proposed remedies

Note: All remedies that refer to a period of time, for example to a ban of X months or a revert parole of Y months, are to run concurrently unless otherwise stated.

Magioladitis desysopped

1) For failure to maintain the standards compatible with adminship and abuse of administrator tools, Magioladitis is desysopped. He may regain the tools at any time via a successful request for adminship.

Comment by Arbitrators:

Comment by parties:

No abuse was ever proved. -- Magioladitis (talk) 20:08, 19 January 2017 (UTC)

BU Rob13 did you take time to check the unblocks and the communication I had with the people who blocked me? -- Magioladitis (talk) 20:58, 3 February 2017 (UTC)

This is not even related to the ArbCom case opened. the only relation I see that vandalfighters like Ramasoud2000 never really check editor's history. I gave analysis of the blocks and the unblocks and none was related to COSMETICBOT. -- Magioladitis (talk) 23:03, 3 February 2017 (UTC)

Comment by others:

Sadly, first choice. The behavior over half a decade is far below what's expected of any editor on the project – a basic willingness to abide by community consensus. That bar itself is far below what is expected of an administrator. On top of all of that, we have the unblocks of his own bot and deletions outside of appropriate deletion processes. After writing up the findings of fact, I'm surprised at how clear-cut this looks to me. I see no other choice. ~ Rob₁₃^Talk 20:04, 19 January 2017 (UTC)

You unblocked your own bot multiple times, in contravention of WP:INVOLVED and the unblocking policy, which states you must never unblock your own accounts. You deleted template redirects with zero discussion after orphaning them yourself to avoid a deletion discussion that would never have gone through. That is abuse. Facts are facts whether you like them or not. ~ Rob₁₃^Talk 20:14, 19 January 2017 (UTC)

Noting the text of this is borrowed in part from the proposed decision of Michael Hardy. ~ Rob₁₃^Talk 20:42, 19 January 2017 (UTC)

Neither of these have been demonstrated. Clearly this has been the aim of Rob/Rama for some time, but it would be a travesty to allow it. All the best: Rich Farmbrough, 23:26, 19 January 2017 (UTC).

See both evidence and the rest of my workshop sections. I've never met Rama before this, nor have I worked with him on any aspect of this case. The filing was unexpected, but welcome, given how long the issues have been going on. I didn't expect at the start of the case to wind up proposing this particular remedy, but the evidence discovered throughout the case leaves no other choice. While you're questioning my motivations based on nothing, your motivations here are fairly clear. ~ Rob₁₃^Talk 00:00, 20 January 2017 (UTC)

Clearly there is a dispute as to whether admins can invite others to be free in blocking their bot and then unblock that bot when they think they've debugged. Resolving that dispute in either direction, or calling on the community to resolve the issue would be a more sensible option than desysopping an admin. ϢereSpielChequers 20:10, 27 January 2017 (UTC)

That is not even half of the issues here. There's editing against consensus after repeated blocks and warnings, violations of the bot policy, out-of-process deletions to circumvent deletion discussions, etc. The unblocks of his own bot are abuse, in my opinion, but they are a small fraction of the overall issues here. In my opinion, as stated above, the editing against consensus is the worst of it. That's something we would sanction a typical editor for, and we expect far better from our admins. It's difficult to read the AN thread on Magioladitis and reach any conclusion other than that there was a clear consensus that he should stop making the sorts of edits that eventually landed us here. (And yes, I've read through the unblocks and communication.) ~ Rob₁₃^Talk 22:53, 3 February 2017 (UTC)

Making edits which a consensus of editors determined to be in violation of COSMETICBOT in a bot-like fashion is well within the scope of this case. The scope was approved as Magioladitis and Yobot's bot and bot-like editing and conduct surrounding that, as far as I'm aware, and this issue is very much interconnected with all that. ~ Rob₁₃^Talk 23:10, 3 February 2017 (UTC)

Magioladitis restricted

2) Magioladitis is restricted from using automated or semi-automated tools from any account with the following exceptions:

1. Magioladitis may use any counter-vandalism tools, including Twinkle in its entirety.
2. Magioladitis may use tools wholly developed by other editors which are (i) designed not to make insignificant or cosmetic-only edits, and (ii) are unable to be modified to do so without further development.
3. Magioladitis may operate any tool on an approved bot account to make edits related to a successful bot approval approved after this remedy goes into effect. This includes semi-automated tasks.

This restriction may be enforced through the use of edit filters.

If, in the opinion of an uninvolved administrator, Magioladitis engages in prohibited activities, he may be blocked for a duration consistent with the blocking policy. The first four blocks under this provision shall be arbitration enforcement actions and may only be reviewed or appealed at the arbitration enforcement noticeboard. Should a fifth block prove necessary, the blocking administrator must notify the Arbitration Committee of the block via a Request for Clarification and Amendment so that the remedy may be reviewed.

Nothing in this remedy prevents enforcement of policy by uninvolved administrators in the usual way.

Comment by Arbitrators:

Comment by parties:

No evidence that my editing ever hid vandalism or something similar to that. -- Magioladitis (talk) 20:52, 19 January 2017 (UTC)

Comment by others:

Some text from Wikipedia:Arbitration/Requests/Case/The Rambling Man. This may seem a bit convoluted, but I'm trying to avoid the mistakes of the RF case (to the extent I perceive them as mistakes, anyway). No-one wants to bar Magioladitis from all bot work. I'd like to see a clear path forward to continue bot work and semi-automated editing with sufficient restrictions to guarantee that this doesn't continue to waste the community's time. This remedy is that path. It allows Magioladitis to use any semi-automated tools that cannot cause the issues under discussion here. It also allows him to get specific tasks approved in the future which can be carefully reviewed to ensure they abide by the bot policy. Magioladitis should be strongly encouraged to request clarification if he's unsure about whether a specific tool is appropriate under this remedy. Note that the "wholly developed by other editors" bit is to prevent gaming the system where Magioladitis develops a tool which he claims is not intended to make insignificant edits but just has many bugs. This is largely what's been going on with Yobot for years, so that bit's important. ~ Rob₁₃^Talk 20:42, 19 January 2017 (UTC)

Oy vey. This response is something. The bot policy, rules of use, etc. do not depend on whether your edits hid vandalism. All that matters is whether you violated community consensus against allowing such edits repeatedly and with complete disdain for consensus. And you did exactly that. ~ Rob₁₃^Talk 20:55, 19 January 2017 (UTC)

This is actually one section where I had hoped for your comment, Rich Farmbrough. Are there other exceptions necessary to make this workable? As mentioned above, I'm very interested in restricting without being too restrictive, as was done in your case. Do you have any comments to that effect? ~ Rob₁₃^Talk 00:02, 20 January 2017 (UTC)

Pinging WereSpielChequers as well, as they had expressed an interest in a sensible set of exceptions in another workshop section. Thoughts? ~ Rob₁₃^Talk 00:10, 20 January 2017 (UTC)

This would not be enforceable by edit filter alone. You would need to decide which tools met the criteria 1, 2.i and 2.ii.

Also you still need to simultaneously trust him to declare his tool use in detail and not trust him to follow the restriction.

(Although if such a restriction were imposed Magioladitis might welcome such an EF to prevent accidental breaches.)

All the best: Rich Farmbrough, 21:05, 21 January 2017 (UTC).

I agree that an edit filter isn't a perfect means of enforcement (or even fully necessary). As I've attempted to explain below, and not done too successfully I believe, the idea of an edit filter was solely to allow the Committee to justify not desysopping if they'd like to. It's very difficult to justify retaining someone as an admin if they have abused or misused an aspect of the administrative toolset. The edit filter is an attempt to divorce AWB from the administrator toolset to give the arbitrators a more complete array of options here. For this broader restriction, it would be up to admins at AE to enforce this restriction, as with any other. I do not consider an edit filter desirable here, except for AWB in the event that Magioladitis is not desysopped. If Magioladitis requests one to prevent accidental breaches, that is another story, and I believe arbitrators should allow that upon request. ~ Rob₁₃^Talk 02:43, 22 January 2017 (UTC)

This seems to be predicated upon your belief that AWB is somehow "an admin tool", even though any non-admin with 250 edits can get access (but admins without a Windows box can't use it). You can get AWB access even if you haven't edited enough to be able to touch the article Israel. WhatamIdoing (talk) 22:53, 24 January 2017 (UTC)

WhatamIdoing, AWB access is like rollback. Anybody can get it, but it can easily be removed in case of abuse like this. However, it can't be removed for admins, just like rollback. The only way to remove both is desysopping. That's why rollback and AWB are considered admin tools. Ramaksoud2000 ^{(Talk to me)} 22:59, 24 January 2017 (UTC)

No, that doesn't make them admin tools. It might necessitate de-sysopping (if one thought that a stated ban were insufficient), but other behaviors, such as uploading naked pictures without consent, might also necessitate de-sysopping. Not all causes of desysopping involve abuse of "admin tools", and not all tools that admins can use, and can get de-sysopped if they abuse, are "admin tools". WhatamIdoing (talk) 23:12, 24 January 2017 (UTC)

@WhatamIdoing: Rollback is a tool that all admins can use, plus a subgroup of editors that admins provide the right to (via a user group). AWB is a tool that all admins can use, plus a subgroup of editors that admins provide the right to (via the CheckPage). What am I missing here? ~ Rob₁₃^Talk 23:52, 24 January 2017 (UTC)

You are missing the fact that "a tool that admins can use" is not the same thing as "an admin tool".

The edit window is "a tool that all admins can use". The edit window is additionally "a tool that admins have been de-sysopped for misusing". None of that makes the edit window "an admin tool". It makes the edit window a non-admin tool that is available to both admins and non-admins under specified circumstances (e.g., the account isn't blocked).

When the policies talk about abuse of admin tools, the primary meaning is abuse of tools that aren't available to non-admins, such as page protection, blocking, and deletion. The primary meaning in the policy is not tools that require a couple hundred edits to articles.

Or, to put it another way, if you happen to think that de-sysopping is appropriate, you don't need to waste time and effort trying to stretch the definition of "admin tools" to include a tool that has always been used by many non-admins (unlike rollback, which used to be exclusively available to admins). There are plenty of other reasons why admins get de-sysopped; it is not necessary to try to prove that AWB is an admin tool and that AWB was abused. WhatamIdoing (talk) 04:43, 25 January 2017 (UTC)

I just don't see any meaningful difference between the rollback situation and the AWB situation. Both are tools available to non-admins under specific situations but available to all admins. The only way to remove AWB access from an admin is to desysop (or set up an edit filter, which is possible, but expensive - you could restrict rollback in a similar way, theoretically). ~ Rob₁₃^Talk 20:11, 27 January 2017 (UTC)

The easiest way to get editors to stop using AWB is to tell them to stop using it. You don't need to make it technologically impossible; just issue a TBAN. WhatamIdoing (talk) 04:00, 4 February 2017 (UTC)

I also don't believe editors who require topic bans should be administrators, so that wouldn't change anything for me. ~ Rob₁₃^Talk 16:56, 4 February 2017 (UTC)

Magioladitis prohibited

3) Magioladitis is prohibited from making any insignificant or cosmetic-only edits from any account other than the main Magioladitis account. Further, insignificant or cosmetic-only edits may be made from the Magioladitis account at rates no greater than one edit per minute.

For the purposes of this remedy, an "insignificant or cosmetic-only edit" is any edit which: i) does not alter the visual rendering of the page, and; ii) does not correct any non-visual accessibility-related errors.

All edits made by a bot account related to an ongoing trial approved as part of the bot approvals process are exempt from this remedy.

If, in the opinion of an uninvolved administrator, Magioladitis engages in prohibited activities, he may be blocked for a duration consistent with the blocking policy. The first four blocks under this provision shall be arbitration enforcement actions and may only be reviewed or appealed at the arbitration enforcement noticeboard. Should a fifth block prove necessary, the blocking administrator must notify the Arbitration Committee of the block via a Request for Clarification and Amendment so that the remedy may be reviewed.

Nothing in this remedy prevents enforcement of policy by uninvolved administrators in the usual way.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Intended to further supplement remedy 2, but not replace it (for now). In the future, it might be possible to lift remedy 2 and keep this in place as part of a transition back to normal editing. For now, this is intended to make plain that, even for approved bot tasks, continued cosmetic-only "errors" will not be tolerated indefinitely. Bugs should be identified and corrected in the trial process. Note that this does not have an exception via consensus, unlike the bot policy. Given that Magioladitis has repeatedly interpreted "consensus for an edit" to be equivalent to "consensus to make that edit at high volumes with semi-automated or automated tools", I don't think an exception via consensus is wise. ARCA is always a possible venue for amendment. ~ Rob₁₃^Talk 20:53, 19 January 2017 (UTC)

Community encouraged

4) The community is encouraged to discuss how to further define cosmetic-only edits in the bot policy to avoid future confusion by non-technical or newer editors.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Note that this does not contradict a remedy being offered. Magioladitis is neither non-technical nor new. His positions and past involvement in discussions about his conduct make it clear that he was in a position where he should understand the policy, perhaps better than almost anyone else on the site. Still, it's become obvious that non-technical arbitrators and editors have had some difficulty grappling with this because they don't have that sort of experience, so perhaps further discussion is a good thing. Maybe if the definition was clearer, this could have been handled more effectively by the community at AN/ANI by allowing editors outside the "bot area" to have educated opinions on how to resolve the issue. ~ Rob₁₃^Talk 20:59, 19 January 2017 (UTC)

Proposed enforcement

Revocation of AWB via edit filters

1) Should Magioladitis violate any restrictions in this case, his access to AWB may be revoked through the use of an edit filter. The activation of such an edit filter shall be an arbitration enforcement action and may only be reviewed or appealed at the arbitration enforcement noticeboard.

Comment by Arbitrators:

Comment by parties:

The desyssop proposal in the section above is not related to misuse of AWB. -- Magioladitis (talk) 18:04, 15 January 2017 (UTC)

It is related to the misuse of AWB. Also, I don't see how conduct requiring a specialized edit filter to stop is conduct becoming of an administrator. Surely, if a hypothetical admin was abusing the deletion tool and declined to stop, Arbcom wouldn't create an edit filter that prevented the admin from deleting (if such an edit filter could be created. Not too familiar with edit filters). Ramaksoud2000 ^{(Talk to me)} 19:53, 15 January 2017 (UTC)

OK so where do you focus? To the misuse of the tool or that there is something deeper behind it? -- Magioladitis (talk) 19:56, 15 January 2017 (UTC)

Arbcom don't create edit filters. Edit filter managers do that. All the best: Rich Farmbrough, 22:45, 16 January 2017 (UTC).

On the how question T111663 -- Magioladitis (talk) 12:16, 18 January 2017 (UTC)

Again this action is not related to the evidence. -- Magioladitis (talk) 10:29, 22 January 2017 (UTC)

Comment by others:

A possible route if desysopping is considered too severe. This may not need a separate enforcement section (are these ever even used?), but it could be incorporated into any sensible restriction involving AWB. Some language taken from remedy #4 of Wikipedia:Arbitration/Requests/Case/The Rambling Man. ~ Rob₁₃^Talk 20:38, 13 January 2017 (UTC)

And how exactly would you craft an edit filter to do this? All the best: Rich Farmbrough, 00:50, 16 January 2017 (UTC).

Target username and then edit summaries including "using AWB", which cannot be removed from the edit summaries of AWB edits from non-bots. ~ Rob₁₃^Talk 01:08, 16 January 2017 (UTC)

I believe "using AWB" is optional for all users. I'm also pretty sure that this is a bad use of edit filters.

Apart form anything else it reeks of Bad Faith. If such a decision were taken it would be enough to ask Magioladitis to cease using the tool.

All the best: Rich Farmbrough, 22:40, 16 January 2017 (UTC).

If we get to the point where an administrator is using something tied to the administrative toolkit and needs it removed, the usual argument is that their admin status must go. I'm trying to provide a potential alternative. I don't think requiring other editors to monitor his edits is helpful when an edit filter can be used. I've run this past other edit filter managers, and they agree it can definitely be done. AWB edit summaries are mandatory for non-bots; it can only be turned off on the bot screen. I've double-checked this using my admin and bot accounts. The option is on the Bot tab. ~ Rob₁₃^Talk 23:28, 16 January 2017 (UTC)

It's a weak argument, however usual it may be. If we respect our fellow editors, then we can work with them collegially, rather than trying to coerce them, we AGF.

If however you ABF then, as AWB developer Magioladitis can leave whatever summary he likes.

All the best: Rich Farmbrough, 23:08, 18 January 2017 (UTC).

How is an edit filter supposed to know if an edit is from AWB? I'll look at the rest more carefully tomorrow, but again it seems excessively detailed. 50.0.136.56 (talk) 11:29, 18 January 2017 (UTC)

OK, I've re-read. Rob, I don't think you understand how wp:edit filters work, unless you mean the filter should just notice "AWB" in edit summaries. I don't think desysopping is required just because an admin has a history of making mistakes with a specific tool. E.g. admin who has too many deletions overturned and gets in drama over them could simply be restricted from doing deletions, while letting them continue their work in other areas if that work is good. In any case, there's no way to prevent someone from modifying AWB to bypass its whitelist or changing its edit summaries. 50.0.136.56 (talk) 22:51, 18 January 2017 (UTC)

Uh ... I'm an edit filter manager. I've created several edit filters, and I most certainly understand how they work. Your claim that you, as an IP editor, understand edit filters better than multiple edit filter managers is quite bold. I can ensure you that an edit filter could be written to prevent edits from Magioladitis' user account with "using AWB" in the edit summary. It's quite an easy and inexpensive filter to write. If Magioladitis were to act in bad faith to re-write the code of AWB to allow him to use it around an edit filter enacted as an arbitration enforcement action, that would obviously be a bright-line action for both desysopping and a ban. As for desysopping, your claim is at odds with both common practice and policy. WP:ADMINCOND expects that administrators act at a high level of conduct in all areas, including where they do not exercise their administrative tools. The idea that an administrator could even outright abuse their administrative tools in one area and retain them with a restriction is so far from both policy and practice that I will not bother to link to arbitration cases contradicting that notion. The arbitrators are well aware of the precedent and policy. ~ Rob₁₃^Talk 23:02, 18 January 2017 (UTC)

See above. There would need to be so many changes to policy (which Arb Com is not supposed to make) to ban someone form editing the source code of an open source project. So why bring edit filters into it?

How would you tell if Magioladitis was using AWB or some other piece of software? Would you insist he has Team Viewer installed on his computer, so that someone could monitor his actions?

All the best: Rich Farmbrough, 23:13, 18 January 2017 (UTC).

A really trivial point: while local edit filter managers are certainly the experts on local usage, the English Wikipedia is not the only place one can learn about Mediawiki extensions, so 50.0.136.56 may well know in detail how they work :)

Rich, I think the underlying idea isn't to ban someone from editing the source code of an open source project, but to ban them from using the resulting modification on the English Wikipedia, which is maybe unusual but hardly a huge departure from precedent in terms of editing restrictions. Opabinia regalis (talk) 00:06, 19 January 2017 (UTC)

Correct. Note that I haven't proposed any restrictions yet. Still thinking on those, but they will come. At the moment, I'm just pointing out that if we restricted the usage of AWB or other semi-automated tools, we could enforce that technically without desysopping. As upset as people are getting over this enforcement idea, this is a de-escalation from what I see as the default action of desysopping if AWB access must be removed, since it's bundled into the administrative toolset. ~ Rob₁₃^Talk 06:56, 19 January 2017 (UTC)

I agree that it's better than nothing to partially enforce the restriction I proposed if desysopping isn't done. I just find it unusual and open to gaming. He has altered edit summaries before for concealment. Ramaksoud2000 ^{(Talk to me)} 07:08, 19 January 2017 (UTC)

I see no reason to think that desysoping would even help any more than the edit filter would. Keep in mind that just as easy as Magio could reprogram it to not use the "using [[Project:AWB|AWB]]" (I got it from examining this AWB edit), he could also program it to bypass the account check. And ArbCom can only ban users from edits to English Wikipedia, not to modifications of external opn-source software (although using that software to edit Wikipedia is an action which a user can be banned from). עוד מישהו Od Mishehu 13:51, 19 January 2017 (UTC)

I mean, from a semantical point of view, we can decide to exclude editors from Wikipedia for any reason, including actions outside of Wikipedia. This has happened before in serious situations (harassment). There is community consensus that AWB access requires administrator approval at WP:PERM, so modifying the program to allow edits without approval would be highly improper. If he used this to get around an actual restriction on semi-automated editing while still an administrator, that restriction violation would warrant sanction. In any event, my point is that the precedent is always that if we must remove an aspect of the administrative toolkit which cannot be restricted otherwise, we desysop. I'm saying an edit filter would restrict access to this aspect of the administrative toolkit to the extent that desysopping would. I find it odd that some editors arguing that Magioladitis is acting in good faith and will abide by community consensus going forward are now arguing that he's so far gone that he'd bypass an edit filter. ~ Rob₁₃^Talk 19:26, 19 January 2017 (UTC)

You should find it odd, because no one is arguing that. I'm surprised that you find the argument difficult to follow:

If we decide to tell Magioladitis to stop using AWB either:

A) he will. (Good Magioladitis scenario.)

Or

B) he will try to continue. (Evil Magioladitis scenario.)

Neither the edit filter nor de-sysopping help in the Good scenario. In the Evil scenario neither desysopping, removing his name from the permissions list, nor the edit filter have any effect.

All the best: Rich Farmbrough, 22:38, 19 January 2017 (UTC).

You're missing the possibility of "Not-Evil-But-Apparently-Unable-To-Follow-Consensus Magioladitis", which is precisely the version of Magioladitis I see in this case. Desysopping, as an aside, is wholly unrelated to preventing the semi-auto issues. It's in response to conduct unbecoming an administrator and abuse of the administrative tools. The Committee has never taken the approach of telling an administrator that they may continue holding the mop but may not use certain aspects of it, to my knowledge. Generally, when you get to that point, you just desysop. This restriction is entirely intended to act as a justification to not desysop and instead remove one aspect of the mop via technical means of doing so, if the Committee decides to go that route. In any event, I know the Committee has already seen this and is considering it on its merits, so I'm going to stop debating this. ~ Rob₁₃^Talk 00:09, 20 January 2017 (UTC)

Proposals by User:CBM

Proposed principles

Bot policy lays out high standards

1) The bot policy explains that the community expects high standards for bots and their operators. Operation of unapproved bots, or use of approved bots in unapproved ways outside their conditions of operation, is prohibited. The contributions of a bot account remain the responsibility of its operator. Bot accounts should not be used for contributions that do not fall within the scope of the bot's designated tasks.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Excerpts from the bot policy. A general principle about the conditions under which bots can be operated. — Carl (CBM · talk) 14:31, 19 January 2017 (UTC)

Proposed findings of fact

Yobot has performed unauthorized edits

1) Yobot has been operated in a way that made it likely to make edits - especially stylistic fixes that are often viewed as cosmetic edits - that were not specifically authorized by any of its bot approvals. Many editors have informed Magioladitis about the problem between 2010 and 2016.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Lack of effective action

2) Magioladitis has not made the kind of changes to Yobot that would end the unapproved edits permanently. He has described the situation primarily as an issue of correcting the lists of pages that Yobot should edit, rather than an issue of changing the way that Yobot edits those pages.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Patently either fix would work. Both would be better. I don't think you can describe hi actions as ineffective without solid statistics to back that up. All the best: Rich Farmbrough, 22:50, 19 January 2017 (UTC).

Given that Magioladitis has been trying to fix the problem for years by just making the lists more accurate, I do think we can describe that method as ineffective. A real solution has to look at the actual edits being made, and reject those that don't do the intended task. Otherwise, we have a nice example of a race condition. THe way to eliminate that is to eliminate any dependence on the list being correct; if the bot was bug free the task could be run on any list of pages without causing errors. — Carl (CBM · talk) 13:44, 21 January 2017 (UTC)

Existing BRFAs lack clarity

3) The bot approvals (BRFAs) for Yobot are not sufficiently clear about additional changes, such as stylistic changes and general fixes, which go beyond the changes directly relevant to the approved task.

Comment by Arbitrators:

Comment by parties:

Comment by others:

This was highlighted by a recent set of edits Magioladitis made on his main account, which were discussed [34]. See [35] in particular. The issue here is that the BRFA was never specific enough, and could indeed be treated as an open-ended authorization for any kind of edit. This finding is intended to support a remedy to vacate the BRFA so it can be refiled, allowing the community to establish a clearer consensus. — Carl (CBM · talk) 13:35, 21 January 2017 (UTC)

Edit summaries lack clarity

4) Many edit summaries by Yobot have lacked clarity about which bot task is being carried out, or which problem with a page is being addressed.

Comment by Arbitrators:

Comment by parties:

I agree that this is true. This got better very recently by adding a link to the BRFA used each time. Still not perfect. -- Magioladitis (talk) 22:33, 19 January 2017 (UTC)

Comment by others:

Proposed remedies

Note: All remedies that refer to a period of time, for example to a ban of X months or a revert parole of Y months, are to run concurrently unless otherwise stated.

General fixes disabled

1) Yobot is not permitted to operate with the "general fixes" of AWB enabled. All changes that Yobot makes must be coded specifically for an approved bot task at hand, rather than via the "general fixes" module of AWB.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Magioladitis has explained that the underlying source of some issues is that, rather than using task-specific code, Yobot simply uses AWB "general fixes" code for some tasks, even though general fixes were not coded to be used in this way, making it nearly impossible to prevent unauthorized edits while general fixes are enabled. AWB does allow operators to code custom changes, and so it would be straightforward to code each desired task directly. — Carl (CBM · talk) 14:31, 19 January 2017 (UTC)

Too general. It is standard practice for bot tasks to be reverse engineered into AWB general fixes. Removing the ability of one or more bots to run the GFs undermines the whole philosophy of GFs.

Given that we have no measure of what is an acceptable rate of false positives in these CHECKWIKI scenarios is, it is impossible to derive a solution guaranteed to be satisfactory, however there are various steps that could be taken:

1. Ensure that the lists are freshly generated when the task is run.
2. Add more granular skip functions (as already requested by Magioladitis).
3. Code the particular fix using "Normal Settings" or "Advanced Settings" in such a way that the edit is skipped if the fix ins't done.

All the best: Rich Farmbrough, 15:58, 27 January 2017 (UTC).

My point is that "roll them into general fixes" is exactly the wrong way to use AWB as a bot. Bot tasks need to be specific and objectively described; lumping together an approved tasks with a lot of other, unapproved changes can only lead to the sort of problems seen here. The solution is to code the specific bot job on its own, and leave the general fixes module for manual editors who can review each edit. — Carl (CBM · talk) 16:21, 27 January 2017 (UTC)

This is complicated, and it probably needs its own community discussion. There's no good answer: either bots have the ability to do 50 tasks at once, or we run 50 bots, and perhaps have 50 edits to the same page (see also: all the complaints about "but I still have bot edits hidden in my watchlist..." – times 50).

Also, I'm not sure that "unapproved changes" is the clearest way to put this. "Not specifically, individually approved for this particular bot to do at the same time as this particular task" is probably true, but I understand that the main point of general fixes is supposed to be that these general fixes should be done always, so it's kind of weird to make rules that discourage people from getting things like [[foo||bar]] to [[foo|bar]] fixed. WhatamIdoing (talk) 04:12, 4 February 2017 (UTC)

BRFAs voided

2) All existing bot approvals (BRFAs) for Yobot are voided. For each task that is desired for Yobot, a new BRFA must be filed.

Comment by Arbitrators:

Comment by parties:

This is already done by BAG. All taske were voided and new BRFAs started being filled. -- Magioladitis (talk)

Comment by others:

This would allow the community to review the additional changes Yobot makes, and either specifically authorize them or specifically forbid them in the new BRFA. The existing BRFAs did not receive sufficient scrutiny. — Carl (CBM · talk) 14:31, 19 January 2017 (UTC)

It's worth noting that there's probably consensus to revoke the tasks already here. I had hoped that thread would be closed before this got kicked off so the committee could deal with one less thing, but it's not too surprising that no BAG members wanted to step into this mess. The Committee may want to take it upon themselves to do these revocations if they see community consensus in that thread to do so (or perhaps no community consensus to continue allowing the tasks - there's no clear bar for revocations because they're rare). ~ Rob₁₃^Talk 06:12, 21 January 2017 (UTC)

Unfortunately, I think that some BAG members might be unsure of their authority, or might not want to have the authority, to actually tell a bot operator to stop doing something. I had always viewed that as clearly part of their charter. In any case, the remedy here is completely independent of whether BAG might also revoke the BRFAs. — Carl (CBM · talk) 13:32, 21 January 2017 (UTC)

Again too sweeping. If anyone can demonstrate changed consensus for a task that's fine. A bulk revocation just wastes the time of all concerned.

All the best: Rich Farmbrough, 16:00, 27 January 2017 (UTC).

Edit summaries

3) All subsequent edits by Yobot must include a reference to the approved BRFA that authorizes them. For edits related to the CHECKWIKI project, the specific CHECKWIKI issues that is being addressed must also be referenced. For clarity, separate CHECKWIKI issues must be addressed in separate bot runs.

Comment by Arbitrators:

Comment by parties:

This is an interesting suggestion. The only reason I am not doing it exactly to save runs and save some of my time loading each error separately. -- Magioladitis (talk) 22:32, 19 January 2017 (UTC)

Comment by others:

This will allow the community to review edits more effectively and to revisit BRFAs or CHECKWIKI lists that cause any issues. — Carl (CBM · talk) 14:31, 19 January 2017 (UTC)

I agree that this would be a better state of affairs. However it is not practical with AWB as it stands. If CBM would like to reprogram that part, then I'm sure his patch would be accepted by the devs.

All the best: Rich Farmbrough, 16:02, 27 January 2017 (UTC).

Proposals by User:Stevietheman

Proposed principles

Community rejects wholly nonsubstantive edits by bots/meatbots

1) Saves of wholly nonsubstantive (called 'cosmetic' by many) edits by bots or meatbots are generally rejected by the community, which bases its concerns on policy (AWB Rules of Use #4, COSMETICBOT) and the site's common practices and understandings.

Comment by Arbitrators:

Comment by parties:

Comment by others:

This applies no matter how the policy came into effect or why it's necessary, or whether an editor understands or agrees with all nuances of it. If the community is regularly informing an editor (especially via venues like ANI) that a particular kind of edit goes against policy, then the community's view should be respected. This applies to the community's view that, for instance, a bot only bypassing template redirects is not substantive, or is cosmetic. Stevie is the man! ^{Talk • Work} 23:41, 19 January 2017 (UTC)

AWB rules of use are not policy.

All the best: Rich Farmbrough, 16:04, 27 January 2017 (UTC).

Bots don't "hide vandalism"

2) Bot or meatbot edits don't hold any responsibility for hiding vandalism; a watchlist bug does.

Comment by Arbitrators:

Comment by parties:

Comment by others:

No matter the quality of a bot edit, whether widely welcomed or not, it's the longstanding watchlist behavior that is making it difficult for vandal hunters. Yobot highlights this problem because of how pervasive it is. Stevie is the man! ^{Talk • Work} 23:41, 19 January 2017 (UTC)

Yes, but bots doing non urgent edits could be put on a delay such that all likely watchlisters have had the opportunity to check a non bot edit before there is a bot edit. ϢereSpielChequers 09:23, 23 January 2017 (UTC)

This would certainly require extra programmer work - I doubt that thisis in the current abilities of AWB. עוד מישהו Od Mishehu 14:34, 23 January 2017 (UTC)

Wholly nonsubstantive edits by bots are pernicious nevertheless

3) An article history filling up with just cosmetic tidying from bots (which could sometimes even make dueling format edits) becomes an obvious disruption to non-bot editors trying to keep up with the development of an article. This is much less likely if bots are required to do something substantive along with any tidying.

Comment by Arbitrators:

Comment by parties:

Comment by others:

COSMETICBOT when respected naturally reduces disruption to the editing process. I have devised this principle from common sense. Do we really want fluff edits and tidying wars piling up on many articles? Stevie is the man! ^{Talk • Work} 23:41, 19 January 2017 (UTC)

"Tidying wars" are extremely rare, and usually quickly resolved.

All the best: Rich Farmbrough, 16:06, 27 January 2017 (UTC).

Proposed findings of fact

AWB's design contributed to problems

1) AWB's design (up to this point) per its use with fixing CHECKWIKI issues has made it somewhat difficult for any bot/user, let alone Yobot and Magioladitis, to comply with the strong community position against making wholly nonsubstantive edits in fast succession as a bot or meatbot.

Comment by Arbitrators:

Comment by parties:

Comment by others:

Per evidence on AWB's past/present limitations and how Yobot is using AWB (i.e., running GenFixes edits whether or not intended CHECKWIKI error in article exists, and running against stale article lists), an "error rate" of wholly nonsubstantive edits is inevitable. They are also effectively unstoppable while running as a true bot. Stevie is the man! ^{Talk • Work} 19:39, 19 January 2017 (UTC)

I think you are correct that an "error rate" is inevitable. But an important question is, "What is the error rate?" No-one has made the effort to establish that. I would think it would be worth forgiving a very substantial number of them [inadvertently "cosmetic only" edits] before bringing a case like this that costs hundreds of volunteer hours, even without taking into account the benefits of those that achieve their stated goal.

All the best: Rich Farmbrough, 21:01, 19 January 2017 (UTC).

I think this is true, but remember the adage that "a bad carpenter blames the tools that were used". If AWB is not designed well for a particular task, the operator should take this into account, or use a different tool. Responding to Rich's comment, I don't think that the numerical error rate is the key fact here. The important point is that number of attempts by different users to remedy the situation, over a period of years, which is on its own a sign that the error rate is too high. When a bot is blocked numerous times, it is hard to claim that it is operating under best practices. In principle a bot that makes only 100 edits per month might have a much higher error rate than one that makes 100,000, but the latter one might still be viewed as more problematic if the absolute number of errors leads to too many complaints. — Carl (CBM · talk) 16:16, 27 January 2017 (UTC)

But AWB is not responsible (its users are)

2) AWB software is not ultimately responsible for the ongoing violations of AWB Rule #4 and COSMETICBOT, but instead its bot operator or non-bot user (Yobot/Magioladitis).

Comment by Arbitrators:

Comment by parties:

Comment by others:

Consider that Magioladitis is an AWB developer himself and thus should be in the position of understanding the full ramifications of #1 above. Evidence shows he knew of various pertinent technical issues well ahead of many periods of controversial series of wholly nonsubstantive edits. At the same, as a fellow AWB user, I understand that AWB up to this point hasn't particularly aided and arguably gets in the way of "using it correctly" (satisfying AWB Rule #4 and COSMETICBOT). But then this leads to a circular conundrum, as again, Magioladitis is an AWB developer and should have fixed (or coordinated the fixes for) the problems that have Yobot making wholly nonsubstantive edits so often. Also consider the evidence of Magioladitis running AWB as a meatbot, while knowing the technical issues, still having the responsibility to review all edits before saving, and is therefore frequently deciding to save edits that don't comply with AWB Rule #4 and COSMETICBOT. AWB does not force a user to save. Given the many years involved with these controversial edits, Magioladitis is ultimately responsible for all the edits in question, and the degree of responsibility has only increased over time due to his intimate knowledge of the software and its limitations, and strong community blowback that is based on policy. Stevie is the man! ^{Talk • Work} 19:39, 19 January 2017 (UTC)

Another aspect here is that AWB doesn't have to be used per #1 when run as a bot. There are other ways to set it up for use as a bot, such as writing module code or doing find/replace's, both of which can be more tightly controlled to not save unless something (presumably substantive) has been changed. AWB as software is not really the "bad guy" in any of this, although certainly it has issues that need to be addressed toward improving its use per #1. Stevie is the man! ^{Talk • Work} 19:58, 19 January 2017 (UTC)

Proposed remedies

Note: All remedies that refer to a period of time, for example to a ban of X months or a revert parole of Y months, are to run concurrently unless otherwise stated.

Punt (except for adminship concerns)

1) Outside of considering adminship of Magioladitis, ArbCom should punt all the various technical, policy and process concerns back to the English Wikipedia community.

Comment by Arbitrators:

Comment by parties:

Re Carl: I don't know what other evidence to bring in order to prove that I am dealing with a problem hard for me, that I tried to fix it and I keep asking other for solutions. I have offered almost all the tasks to more experienced bot owners, I reported all the bugs and complains and I asked other programmers to fix the remaining bugs. -- Magioladitis (talk) 17:24, 27 January 2017 (UTC)

Comment by others:

First, I offer this remedy in all seriousness. There are seemingly a multitude of technical, policy and process concerns brought to the fore in this case that need hammering out, and I see no hard reason why the English Wikipedia community cannot do this in its own good time. For this punting, ArbCom should will the community to pursue solutions to as many concerns brought up in this case as possible. Stevie is the man! ^{Talk • Work} 00:26, 21 January 2017 (UTC)

@Stevietheman: We've had a couple bot operators interested in making cosmetic-only edits and a former administrator who was previously desysopped and sanctioned for making automated edits against consensus pop out of the woodwork to try to get ArbCom to change policy by fiat, but if you look at related discussions at community venues, the community itself has had little difficulty determining what the policies currently are and what they mean. There's been a concerted effort to convince the Arbitration Committee that they should invalidate existing policy, which they cannot do, but please do not be taken in by a few editors pushing their view of what the policy should be. What the community has failed to do is apply sanctions to Magioladitis which have addressed the issues. There have been blocks, which have been lifted repeatedly, either by Magioladitis or by another admin after Magioladitis has promised to change their editing patterns. This hasn't panned out in a long-term sense. This is a classic "break the back of the dispute" situation. ~ Rob₁₃^Talk 00:39, 21 January 2017 (UTC)

If I'm not mistaken, I think you're making the same point I am. I think ArbCom should only consider Magioladitis' adminship, and not spend time working through all the complexities necessary to do what the community itself is responsible for. In my estimation, it's just too much. For instance, if COSMETICBOT needs review, then the enwiki community can have a big fat RfC about it. Or, if the community needs to rein in CHECKWIKI or Yobot or GenFixes, there are already ways to move in that direction if only there's a will to do so. What ultimately brought this case is an individual's behavior that is suggested to not be in line with community standards, no matter what policies are (and whether 'legit' or not), what the technical issues are, or how various processes work or are supposed to work. Stevie is the man! ^{Talk • Work} 01:42, 21 January 2017 (UTC)

@Stevietheman: No, it's the Arbitration Committee's role to resolve behavioral issues that the community is unable to. The community has plainly been unable to solve these issues in a lasting way. See the 25+ related discussions in the original case request that led us here. See WP:A/G, which makes plain that the Arbitration Committee is intended to "break the back" of disputes. I fully agree on what brought us here, but since the community (a) recognized a problem exist, and (b) was unable to fix the problem, we've been led here. If this is punted back to the community, I can tell you exactly what will happen: Issues will continue, we'll have an ANI which won't go anywhere, and then we'll be back here for another whole month-long case. ~ Rob₁₃^Talk 02:10, 21 January 2017 (UTC)

I am confused by your position. I am saying that ArbCom should deal with the behavioral issues of the individual this case is about. Otherwise, punt. (also, no need to ping me) Stevie is the man! ^{Talk • Work} 02:17, 21 January 2017 (UTC)

The failure to abide by community consensus on how they may make AWB and bot edits is a behavioral issue, and considering adminship alone doesn't resolve it. Is that where we're having a disconnect, or am I misunderstanding you? ~ Rob₁₃^Talk 03:03, 21 January 2017 (UTC)

I think the community should deal with the non-admin side of it because too-high-up decisions could end up negatively affecting other editors who aren't causing issues. At any rate, this is my last contribution to this case. I have spent enough time on this already. Stevie is the man! ^{Talk • Work} 12:02, 21 January 2017 (UTC)

I have the opposite opinion, to some extent. I don't see the adminship issues as very serious, but the other issues form the core of the case. From the evidence, this is a longrunning issue that ordinary methods have failed to resolve. There is room for a well crafted decision to move things forward. — Carl (CBM · talk) 13:40, 21 January 2017 (UTC)

In which case the community can deal with it, by working together, rather than making demands of Magioladitis. All the best: Rich Farmbrough, 16:12, 27 January 2017 (UTC).

I think the point of this arbitration case is that the community has attempted to deal with it, but Magioladitis has not been willing to resolve the issue, over a period of years. One role of the arbitration committee is to step in when normal community processes have not been successful. — Carl (CBM · talk) 16:19, 27 January 2017 (UTC)

Proposal by User:Headbomb

I'm going to jump in the fray here, not bothering to read anything else here because there's little point in going over the minutia. I'm going to make this pretty straightforward, drama free (HA!) and simple. A bunch of us from WP:BAG (alongside with the community) are going over a few things at the much-lower drama WT:BOT and related pages as a result of this mess, and a few productive things are happening, including a much better and expanded "dealing with issue" section.

I think what I'll propose should be able to defuse the situation, in as much as it will focus on "what do we do now" rather than focus on a redress for past grievances/punishment. This should prevent User:Magioladitis and his bots from annoying people while at the same time allowing for him to keep contributing to WP:CHECKWIKI and WP:AWB improvements. This is important for a few reasons. On one hand the community is exhausted by these repeated violations of WP:COSMETICBOT, but banning/denying Magiodilitis his ability to do CHECKWIKI-related work and fixes would be a massive loss of expertise (as well as the loss of one of Wikipedia's most prolific WP:GNOME). This involves some work by WP:BAG, some work by the WP:AWB team, some work by the WP:CHECKWIKI people, and some work by the community at large.

Here is what I propose:

1. Put a temporary moratorium on Magiodilitis on making AWB genfixes/checkwiki edits with AWB, either directly or through a bot account.
2. WP:BOTPOL needs to have better guidance on what constitutes a cosmetic edit. This is currently worked on at User:Anomie/Sandbox2 and will be submitted to a community-wide RFC.
3. WP:BOTPOL needs to have better guidance on self-unblocks when it comes to bots. This is currently worked on at WT:BOTPOL and will be submitted to a community-wide RFC.
4. WP:CHECKWIKI needs a reviewing process to go over each checkwiki fixes. Each individual CHECKWIKI fixes gets marked as 'major' or 'minor' or 'not suitable for bots/AWB'. This is possibly already done at Wikipedia:WikiProject Check Wikipedia/List of errors, but the updated guidance on cosmetic edits may change the categorization of some of those.
5. WP:AWB needs to get a dedicated WP:CHECKWIKI module/settings tab/whatever (similar to the AWB genfixes, or typo fixes). The exact details can be hammered out by the AWB team, but the idea would be that there is a "Enable checkwiki fixes" option with a "Skip if only minor checkwiki fixes" option, as a mirror of the current "Enable genfixes" / "Skip if only minor genfixes" options. If checkwiki fixes are done, the relevant checkwiki # are explicitly listed in the edit summary, with an explicit link inviting users to report mistakes and improvements at Wikipedia talk:WikiProject Check Wikipedia.
6. Once #2-#5 are done (or something along that spirit since I don't like the idea of ARBCOM mandating specific implementations), the moratorium is lifted (via motion by ARBCOM) and Magioladitis can resume making AWB genfixes/checkwiki edits with these new modules, subject to WP:BRFAs.
7. Indefinite ARBCOM restriction: Magioladitis must always have the 'skip if only minor genfixes' and 'skip if only minor checkwiki fixes' options enabled when editing with AWB, either directly or through a bot account.
8. Indefinite ARBCOM restriction: Magioladitis may not unblock or take other administrative action concerning his bots or himself for any reason. (If haven't read if a desysop is proposed, if that's the case those restriction would apply if Magio is desysopped and regains adminship later.)

Since Magioladitis is a great coder with loads of AWB/CHECKWIKI expertise, he should be more than able to help with #4 and #5 and design this framework. The temporary moratorium gives everyone a breather, gives Magioladitis a path forward, AWB/CHECKWIKI development continues and gives the community with much better tools. I'll also take the opportunity to advertise T157100 which may or may not be relevant to the proposed AWB/CHECKWIKI framework.

Feel free to call this the 'Headbomb option' for the general idea, although feel free to disagree with/propose refinements to individual items on this list.

Headbomb {talk / contribs / physics / books} 13:17, 3 February 2017 (UTC)

Comment by Arbitrators:

Comment by parties:

For me, it's very satisfactory that it is recognised that a BOTPOL is required for two core issues I am involved the last 10 years. I al also satisfied that hopefully more people will get involved in CHECKWIKI. I have a question, if anyone really checked the "self-unblocks" and if there were really self-unblocks but anyway. This is very minor at the moment. One way or another, it is recognised that the big problem occurs from the fact that the rules were/are not clear. I would like also if a comment whether the complains are/were all reasonable. Because, in my perspective: Not all complains are for the same reason (and I do not mean in the details), not all wanted the same thing. Moreover, I am not satisfied by the fact none proved that any of edits ever hidden any vandalism. -- Magioladitis (talk) 14:24, 3 February 2017 (UTC)

Headbomb your phab ticket was already in my to-do list and was already on Phabricator. I add you in the subscribers list. (You could also watch the Thesswiki Hackathon video where I explicitly ask the students fo the Informatics Department of the University of Thessaloniki to do this... as exercise but it's in Greek. Lol. ) -- Magioladitis (talk) 14:35, 3 February 2017 (UTC)

Headbomb (talk · contribs) Some of the edits should not have happened independently of the fact there were cosmetic(?) or not. But in a span of 10+ years and after a series of 4 million edits I consider the number of pages that the bot actually failed is very small. Still I make a lot of effort to reduce these numbers and I have provided proof for that while other bot owners have left more criticla errors unsolved for years. Still no motion was against bots that actually broke the wiki syntax while there is an open Arbcom for a bot that never broke wiki syntax and on the other hand fixes it.. it is also worth to say that in fact no actual problems were reported but only potential problems (i.e. it may have hidden vandalism while I proved that in contrary CHECKWIKI works in finding vandalism and none bothered to bring an example that Yobot hid a vandalism). -- Magioladitis (talk) 15:20, 3 February 2017 (UTC)

xaosflux (talk · contribs) If you read the discussion you 'll see I unblocked my bot account after discussion with the blocking admin. Where is the wrong to that? For instance, Did you read the first block and the discussion I had with the blocking admin? -- Magioladitis (talk) 23:28, 3 February 2017 (UTC)

Comment by others:

Magioladitis (talk · contribs) I want to make it clear here that that I do not believe the core issues occur because the "rules were unclear". The core issue is that you and your bot clearly have been making cosmetic edits when they shouldn't have. While many of the complaints may have originated from unclear rules, and from a lack of understand about what truly consists of a cosmetic or non-cosmetic edit, several of your edits were clearly cosmetic under any definition of the term. Regardless of whatever technical or non-technical reasons caused them, the need for clearer guidance on this issue is to give you (and the community) tools to address the core issue, and allow for a clearer categorization of AWB genfixes/CHECKWIKI fixes into "This is OK on its own"/"OK to do as part of a more substantial edit"/"Not OK" and deny you the opportunity to screw up despite yourself (or despite the software).

Now, to get back on solving these issues, you said you're happy with #2/#3 an increased involvement from the community with checkwiki fixes. The question is now are you willing to help with the framework of #4/#5 so #6/7 can be done? Headbomb {talk / contribs / physics / books} 14:57, 3 February 2017 (UTC)

Components such as changing community policies are generally beyond the scope of an arbcom case, focusing only #8 above related to administrative actions: a restriction on Magioladitis from unblocking himself or any alternate account of his (including bot accounts), excepting reversal of block actions he places himself seems reasonable If there is a finding that prior actions were against policy or standards (ins: 16:04, 3 February 2017 (UTC)), if coupled with an desysop enforcement condition. Magioladitis does positively contribute administratively in general. — xaosflux ^Talk 15:17, 3 February 2017 (UTC)

This isn't ARBCOM changing policy, but would rather be endorsement of a general path forward which might be preferable to the usual "block/ban/restrict" done in vacuum. Or if you prefer, recognition that efforts are being made elsewhere in the community that should help with the situation. These two RFCs will happen regardless of ARBCOM, but if you prefer putting this into a "findings of fact" format, I suppose you could have something that reads like "The existing bot policy does not strictly define what a cosmetic edit is" / "What exactly constitutes a cosmetic edits has no strict definition". Headbomb {talk / contribs / physics / books} 15:29, 3 February 2017 (UTC)

@Magioladitis: I am not referring to your use of AWB in anyway above, only administrative actions. I think that in general your administrative actions are helpful and supported by the community, with a notable exception unblocking of your own account(s). As for AWB use, you could almost as trivially fork AWB to one that doesn't use a permissions list. If the community thinks you are making disruptive edits, they can block you just like anyone else - the interface you use to make the edits is far secondary to determining if the edits made are or are not disruptive. — xaosflux ^Talk 15:41, 3 February 2017 (UTC)

Not sure why we are having this discussion again as I clarified above the only thing I'm supporting here is that if there is a finding of administrative abuse related to the unblock tool that I think the remedy should be a restriction on future use, as opposed to a desysop. — xaosflux ^Talk 00:50, 4 February 2017 (UTC)

Proposals by User:Example

Proposed principles

Template

1) {text of Proposed principle}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) {text of Proposed principle}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposed findings of fact

Template

1) {text of proposed finding of fact}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) {text of proposed finding of fact}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposed remedies

Note: All remedies that refer to a period of time, for example to a ban of X months or a revert parole of Y months, are to run concurrently unless otherwise stated.

Template

1) {text of proposed remedy}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) {text of proposed remedy}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposed enforcement

Template

1) {text of proposed enforcement}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) {text of proposed enforcement}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Analysis of evidence

Place here items of evidence (with diffs) and detailed analysis

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

General discussion

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposals by Iazyges

Proposed principles

Watchlist intention

1) The intention of a watchlist is for the visibility of changes to users who watch them, so that they may respond to, add to, or remove the changes. This abilities purpose is to foster collaboration and aid quality control. Iazyges Consermonor Opus meum 21:24, 4 February 2017 (UTC)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Purpose of Bots

2) Bots are intended to perform edits that are at the same time worthwhile enough to be made, but also not to make edits that are complex enough that only human editors can correctly perform them. Iazyges Consermonor Opus meum 21:24, 4 February 2017 (UTC)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Purpose of Tools

2) The purpose of all tools on Wikipedia is to perform more efficient actions, or in some way make an action more efficient for human editors. Iazyges Consermonor Opus meum 21:24, 4 February 2017 (UTC)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposed findings of fact

Cosmetic edits

1) Yobot has performed numerous cosmetic edits, under the current classification of cosmetic edits, such as one digit byte change edits not intended to fix issues. ([36]). Iazyges Consermonor Opus meum 21:24, 4 February 2017 (UTC)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) Magioladitis has unblocked the bot numerous times, against the wishes of the blocking administrator, without being certain that the problem has indeed been fixed. ([37])

Comment by Arbitrators:

Comment by parties:

The evidence presented by Ramaksoud2000 were just wrong. I gave the accurate status of the unblocks. -- Magioladitis (talk) 22:57, 4 February 2017 (UTC)

Comment by others:

Proposed remedies

Note: All remedies that refer to a period of time, for example to a ban of X months or a revert parole of Y months, are to run concurrently unless otherwise stated.

Bot unbanning

1) Any bot that has been blocked for malfunctioning, and making incorrect edits, must be unbanned by Bot Approvals Group consensus, unless it is of such critical importance that Wikipedia as a whole would suffer from its lack of use (Such as Cluebot NG). All consensus made should be by process of an appeal to the BAG page, where parties external to the situation are allowed to comment, but are not considered part of the consensus. Iazyges Consermonor Opus meum 21:24, 4 February 2017 (UTC)

Comment by Arbitrators:

We can encourage the community, or BAG, to consider certain things. Doug Weller talk 19:11, 5 February 2017 (UTC)

Comment by parties:

Comment by others:

This actually seems to be an excellent idea - although ArbCom generally doesn't dictate policy. עוד מישהו Od Mishehu 12:35, 5 February 2017 (UTC)

@Od Mishehu: Where would you suggest I bring it up? Blocking policy, bot policy or the BAG talk page seem like the best places. -- Iazyges Consermonor Opus meum 18:46, 5 February 2017 (UTC)

Probably any one of the above, and mentioning it on the other 2. עוד מישהו Od Mishehu 19:02, 5 February 2017 (UTC)

Usage of automatic and semi-automatic tools

2) Any usage of automated tools, or semi-automated tools performing actions involving advanced permissions (those that require ARBCOM or community consensus to grant), must be both declared and approved. AWB, Twinkle, and any other well developed and commonly used vandal fighting tool (broadly construed) is not party to this, as their policies to its have already been developed. Iazyges Consermonor Opus meum 21:24, 4 February 2017 (UTC)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposed enforcement

No-consensus unbans

1) Any admin who unbans their own bot (broadly construed to include bots they co-own or write) without permission granted by consensus of the BAG, will be strongly admonished. Should the administrator repeat said action, they may be stripped of their advanced privileges (as defined in the Usage of auto and semi-auto tools section), and may be banned for a period of up to a month. Iazyges Consermonor Opus meum 21:24, 4 February 2017 (UTC)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) Any user that uses either automated tools, or semi-automated tools that perform actions requiring advanced permission (as defined in the Usage of auto and semi-auto tools section), without declaration shall be admonished and given a period of a week to declare in a place easily seen in their userspace (assuming user is still active). If they fail to declare the tool usage, or use the tools without permission to begin with, they shall receive a one week block (if both are true, the bans do not run concurrently). Should they do either, or both, again, they shall receive a one month block should (which also does not run concurrently). A third violation of either or both will result in an indef block (which is run concurrently for obvious reasons). Iazyges Consermonor Opus meum 21:24, 4 February 2017 (UTC)

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposals by User:Example 5

Proposed principles

Template

1) {text of Proposed principle}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) {text of Proposed principle}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposed findings of fact

Template

1) {text of proposed finding of fact}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) {text of proposed finding of fact}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposed remedies

Note: All remedies that refer to a period of time, for example to a ban of X months or a revert parole of Y months, are to run concurrently unless otherwise stated.

Template

1) {text of proposed remedy}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) {text of proposed remedy}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Proposed enforcement

Template

1) {text of proposed enforcement}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

2) {text of proposed enforcement}

Comment by Arbitrators:

Comment by parties:

Comment by others:

Analysis of evidence

Place here items of evidence (with diffs) and detailed analysis

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

Template

Comment by Arbitrators:

Comment by parties:

Comment by others:

General discussion

Comment by Arbitrators:

Comment by parties:

Comment by others: