Wikipedia:Village pump (technical)

 Policy Technical Proposals Idea lab WMF Miscellaneous 

• Table of contents
• First discussion
• End of page
• New post

The technical section of the village pump is used to discuss technical issues about Wikipedia. Bug reports and feature requests should be made in Phabricator (see how to report a bug). Bugs with security implications should be reported differently (see how to report security bugs).

Newcomers to the technical village pump are encouraged to read these guidelines prior to posting here. If you want to report a JavaScript error, please follow this guideline. Questions about MediaWiki in general should be posted at the MediaWiki support desk. Discussions are automatically archived after remaining inactive for five days.

« Archives, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217

dtenable testing

Hello, all. I'd like some folks to test this upcoming feature for me. Here's how it works:

1. Click on https://en.wikipedia.org/wiki/Wikipedia:Village_pump_(technical)?dtenable=1
2. Scroll all the way down to this section and find the [subscribe] button for ==this section==.
3. Click that.
4. Post a comment here.
5. Wait for another editor to do the same.
6. Tell me what you think.

You should get a note via Special:Notifications (Echo) about the new comment. You shouldn't get notified if someone is just fixing a typo (same rules as trying to ping someone; notifications require a new comment/line).

I understand that on the technical side, you're actually subscribing to the timestamp of the first comment, not the section itself. Consequently, changing the section heading, moving this message to another page, etc., will not prevent you from getting notified about new comments. It doesn't matter what editing tools the other comments are posted with.

BTW, that link will give you the [reply] tool as well; if you like it, then turn it on in Beta Features. Whatamidoing (WMF) (talk) 19:35, 27 August 2021 (UTC)

OK, leaving a comment to test. Elli (talk | contribs) 19:36, 27 August 2021 (UTC)

Thank you for testing. Whatamidoing (WMF) (talk) 19:41, 27 August 2021 (UTC)

Oh, that's pretty cool! Curious about how the timestamp thing works though - it's not uncommon for comments to share the same timestamp. Elli (talk | contribs) 19:42, 27 August 2021 (UTC)

There are some technical details in mw:Extension:DiscussionTools/How it works.

Maybe it includes the whole signature? But even then, someone could (with full-page editing) start several separate ==Sections== in the same edit, resulting in multiple identical signatures. Or we could copy it somewhere and have two copies of the same discussion. I'm not sure if it would track both, or neither, or pick one. @Matma Rex, do you have a prediction, or should we try to break it ourselves? Whatamidoing (WMF) (talk) 19:59, 27 August 2021 (UTC)

Subscribed and posting to test. TimTempleton ^{(talk) (cont)} 21:39, 27 August 2021 (UTC)

No discernible alert - is there a delay? TimTempleton ^{(talk) (cont)} 21:41, 27 August 2021 (UTC)

No one had replied yet... now you should get a notification? Elli (talk | contribs) 21:42, 27 August 2021 (UTC)

and another persion replied —TheDJ (talk • contribs) 21:48, 27 August 2021 (UTC)

Cool - I got the notifications. I thought it would alert me when any comments were added to the subscribed thread, mine or otherwise. The only thing I would have to think about is if it will overmessage people. Is there an easy way to unsubscribe if the discussion gets too busy? TimTempleton ^{(talk) (cont)} 21:50, 27 August 2021 (UTC)

Well, the unsubscribe button is right there, too, from the same link. I'd assume if the feature was turned on generally, it would always be there. Elli (talk | contribs) 21:52, 27 August 2021 (UTC)

Ah yes - you have to click the original link to see it - gotcha. Cool. This could be interesting, particularly when new users don't know to tag people. TimTempleton ^{(talk) (cont)} 21:54, 27 August 2021 (UTC)

You can also unsubscribe directly from the notification, without having to view the subscribed-to page. Click on the "..." and then "Unsubscribe". MANdARAX • XAЯAbИAM 21:52, 28 August 2021 (UTC)

Yes, I also saw that you can unsubscribe from the notifications drop-down. I think that's a really nice addition. ⁓ Pelagic ( messages ) 15:54, 15 September 2021 (UTC)

I'm assuming that people don't actually want to get a notification when they post in a discussion (because you just posted your comment, so you already know that you did it, right?). But if my assumptions are wrong, please let me know when that would be useful.. Whatamidoing (WMF) (talk) 15:47, 30 August 2021 (UTC)

Oooh, cool, you can subscribe to sections now? I did wonder what would happen or how they'd be distinguished if the same person posted multiple sections at once. And would you be auto-subscribed to any sections you create? Opabinia regalis (talk) 07:17, 28 August 2021 (UTC)

Somebody please reply to this :-) -- RoySmith (talk) 14:07, 28 August 2021 (UTC)

Never mind, I'll get one of my socks to do it. RoySmith-testing (talk) 14:13, 28 August 2021 (UTC)

Response for Roy. TimTempleton ^{(talk) (cont)} 14:22, 28 August 2021 (UTC)

I am not a sock! It's that stupid testing guy. He's always stalking me and calling me a sock. He should be blocked. I'm only here to improve wikipedia. RoySmith-Mobile (talk) 14:24, 28 August 2021 (UTC)

Wow, this is major-league cool. Being able to subscribe to a thread has long been at or near the top of my wiki-wishlist. Very nice.

Somewhat orthogonal to this, it would be nice if there were a published spec for how to add buttons to a section. The "Close" link in the attached screenshot is from User:DannyS712/DiscussionCloser.js (ping DannyS712). I assume with some CSS tweaks, the visual style of both links could be harmonized, and a published standard would make that happen. -- RoySmith (talk) 14:33, 28 August 2021 (UTC)

Coolest aspect is that it lets me subscribe to a section without adding the page to my watchlist. Schazjmd (talk) 14:36, 28 August 2021 (UTC)

Cool, it worked! Does this work on mobile too? Opabinia externa (talk) 17:33, 28 August 2021 (UTC)

Hey, sock, back in the drawer! Opabinia regalis (talk) 17:35, 28 August 2021 (UTC)

I'm a dummy, I forgot to click subscribe first. Opabinia externa (talk) 17:43, 28 August 2021 (UTC)

Try this one more time, because socks aren't very smart. (The link in the OP goes to the desktop site, will this also be enabled on mobile?) Opabinia regalis (talk) 17:44, 28 August 2021 (UTC)

Success! Opabinia externa (talk) 17:45, 28 August 2021 (UTC)

Testing, attention please! - Klein Muçi (talk) 21:40, 28 August 2021 (UTC)

@RoySmith @DannyS712 You can add the CSS class mw-editsection-like to any element that should look like a section edit link, but not behave like one (e.g. not open the visual editor when clicked). It's what we used for the subscribe links too (I'm one of the developers). Matma Rex talk 12:08, 30 August 2021 (UTC)

@RoySmith Re: this, I think User:Writ Keeper/Scripts/autoCloser.js aligns properly (I use it because it works on mobile devices way better). ― Qwerfjkltalk 12:48, 30 August 2021 (UTC)

@Elli @Whatamidoing (WMF) Sections you subscribe to are identified by the username and the timestamp of the oldest comment. If two sections have identical username and timestamp (even on different pages), and you subscribe to one of them, everything behaves as if you had subscribed to both – you'll get notifications for both of them. Unsubscribing from one also unsubscribes you from all others.

It is perhaps not the ideal behavior, but it allows for sections to be moved, renamed, or archived/unarchived, without losing the subscriptions. And it's more reliable and understandable than if we were trying to detect whether two sections in different pages/revisions are the same using some heuristics. Matma Rex talk 12:05, 30 August 2021 (UTC)

I don't see any issue with this. The chance of there being two threads that have the same (username, timestamp) pair seems so small as to be safely ignored. And the fact that the subscription survives archiving is a big win.

In fact, it would be awesome if this technology could be extended to links. People often link to threads on noticeboards. Those links soon go stale when the thread gets archived. Having a way to create a link which survives archiving would be huge. -- RoySmith (talk) 12:35, 30 August 2021 (UTC)

We've been thinking about permanent links (phab:T273341), and it should be possible to extend it this way, but it would be a somewhat large project – because right now we don't actually "remember" where each topic appears, we just generate notifications when we see it anywhere. Matma Rex talk 13:39, 30 August 2021 (UTC)

Volunteer-me says that this will come up when people start complex RFCs. The simplest solution might be a social one, however: just discourage people from posting multiple comments (at the top of a ==Section==) in the same edit. Whatamidoing (WMF) (talk) 15:51, 30 August 2021 (UTC)

Hi, Matma Rex, re. username and the timestamp of the oldest comment: does that mean the first comment after the heading, or the actual oldest? If someone top-posts will it break things? ⁓ Pelagic ( messages ) 17:24, 7 September 2021 (UTC)

The actual oldest, not the first. Top-posting should not cause trouble (in particular, we were thinking of templates like {{resolved}} and {{discussion top}} when implementing this). Matma Rex talk 17:58, 7 September 2021 (UTC)

Awesome, thanks for the answer! ⁓ Pelagic ( messages ) 10:53, 11 September 2021 (UTC)

(←) Clicking subscribe works even without having to post in the section as instructed. This is a Good Thing™.

Unscription becomes something of an adventure if the subscribed timestamp is edited away. I made the mistake of subscribing to Wikipedia:Usernames for administrator attention#User-reported, on the theory I'd see an update relatively quickly; I had to go searching back through old revisions until I found one that still had the original comment before the [ subscribe ] link turned back into [ unsubscribe ]. (I didn't notice until afterwards that there are also unsubscription buttons in the notifications themselves.) I didn't get notifications for edits made after the timestamp was removed, which I guess is probably the right thing to do for the sort of people who make reports to those kinds of sections, but it makes this feature useless for the sorts of people who monitor them. —Cryptic 22:26, 28 August 2021 (UTC)

Given that those aren't really talkpages, perhaps a special "subscribe to this particular section on this page" option would also work (maybe require an invisible keyword or template be placed on the page to prevent it from being used accidentally)? Elli (talk | contribs) 22:36, 28 August 2021 (UTC)

I don't think the distinction between talk and non-talk pages is useful. We have lots of pages in non-talk namespaces which support talk-like threaded conversations. Many (but not all) are in Wikipedia space. Tools like this should work equally well on those pages. -- RoySmith (talk) 22:40, 28 August 2021 (UTC)

I agree that it should work equally well in all namespaces. But a board like UAA (or AIV, RMT, etc) is not structured like a talk page. Therefore, I think a different subscription method for those would be reasonable to implement, instead of trying to pretend they're talkpages. Elli (talk | contribs) 22:43, 28 August 2021 (UTC)

Maybe a commented-out timestamp, or one in a display:none span. Though that'd probably confuse the poor bots. Not really the original usecase, anyway, just a missed chance for additional awesomeness. —Cryptic 22:41, 28 August 2021 (UTC)

I think it works in all namespaces, as long as there is a signature in the section. There was a brief bug that displayed the [subscribe] button on all ==Sections== (e.g., including in articles), but if there isn't a signature, you'll never get a notification.

Also, remember that it doesn't produce notifications for changes in sections. It notifies you only for new comments. This means less noise (especially if the editors you're talking to revise edit comments many times), but it wouldn't work for watching a section in an article to see if anyone changes it. You'll still need your watchlist for that. Whatamidoing (WMF) (talk) 15:54, 30 August 2021 (UTC)

How does it handle unsigned comments that get fixed up by SineBot (who, at more than 2 million edits, is one of our most prolific contributors). -- RoySmith (talk) 16:17, 30 August 2021 (UTC)

It doesn't send a notification (it checks that the username in the signature matches the user who saves the edit). It should notify in this case (and we support comments signed that way for the [reply] links), but if we simply removed that check, it'd result in notifications being sent when someone (or a bot) is copy-paste-archiving a discussion. Resolving this might need to wait until we can "remember" each comment that has existed (same thing we'd need for the permalinks that you mentioned earlier). Matma Rex talk 16:25, 30 August 2021 (UTC)

Roy, there should many be fewer unsigned comments once the Reply tool is deployed default-on for everyone (including new editors). That won't happen here (for at least weeks, maybe months), but it will help a lot with that problem. Whatamidoing (WMF) (talk) 19:35, 30 August 2021 (UTC)

This is a test. Please reply to me and I will let you know if I get a notif. BEEEP! Note: I'm not actually a robot Blaze The Wolf | Proud Furry and Wikipedia Editor (talk) (Stupidity by me) 19:29, 31 August 2021 (UTC)

Using the 'reply' link to leave a comment. Nick Moyes (talk) 21:09, 31 August 2021 (UTC)

This is possibly a comment. -- Asartea ^{Talk | Contribs} 04:59, 6 September 2021 (UTC)

I told you you should have blocked that sock! Now he's back, making a pest of himself again and stalking me all over. Imma not going to sign this because I don't want to even be associated with him.

I'm late to this party. Does someone need to reply to me with the reply tool for the ping to work, or does any update to the thread prompt a notification? CMD (talk) 12:39, 30 August 2021 (UTC)

Any update (I'm using CD). ― Qwerfjkltalk 12:45, 30 August 2021 (UTC)

Not sure what CD is, but I did get a notification, thanks. CMD (talk) 12:47, 30 August 2021 (UTC)

c:User:Jack who built the house/Convenient Discussions. ― Qwerfjkltalk 12:52, 30 August 2021 (UTC)

A comment here. — xaosflux ^Talk 13:29, 30 August 2021 (UTC)

It'll notify you regardless of what editing-method someone is using, so long as whatever edit is made looks like it's a new comment. This pretty much means "adds a new list item that ends with a signature" -- so any tool that doesn't do that is already probably being complained about for violating discussion norms. :D

There's a description of how it's working, if you're interested in more details. DLynch (WMF) (talk) 17:35, 30 August 2021 (UTC)

A comment. ― Qwerfjkltalk 19:32, 31 August 2021 (UTC)

Holy heck. That's actually rather cool. Didn't appear under my bell icon like I thought but under notices. Blaze The Wolf | Proud Furry and Wikipedia Editor (talk) (Stupidity by me) 19:36, 31 August 2021 (UTC)

WOOHOO! This is massive! Hats off to everyone who's been working on this. I can personally verify that this is a supremely technically difficult area to make progress in, and it's very, very cool that they've been able to get this far. Congratulations! Enterprisey (talk!) 05:04, 6 September 2021 (UTC)

Arbitrary subsection

One disadvantage of working with signatures rather than headers appears to be that the notification does not take you to the talkpage section, although this is perhaps a minor inconvenience. CMD (talk) 13:34, 30 August 2021 (UTC)

If you visit the page with ?dtenable=1, it actually does (for example, try this). Assuming this feature is turned on generally, it'll work. Elli (talk | contribs) 13:36, 30 August 2021 (UTC)

Subsection added. — xaosflux ^Talk 13:38, 30 August 2021 (UTC)

For some reason that particular link doesn't work, however I tested with other links and it goes right to the comment, which is great. CMD (talk) 13:42, 30 August 2021 (UTC)

It's been working for me. I especially like the blue background highlighting. I got a notification earlier this morning for multiple comments in the thread; not only were the multiple notifications collapsed into one (nice), but each comment was individually highlighted. Very nice. -- RoySmith (talk) 13:46, 30 August 2021 (UTC)

It will also work if you have the "Discussion tools" beta feature enabled in Special:Preferences#mw-prefsection-betafeatures (even though it doesn't enable topic subscriptions yet). I think we didn't consider this when testing the dtenable parameter, sorry! (I'm one of the developers) Matma Rex talk 13:43, 30 August 2021 (UTC)

Testing... --Yair rand (talk) 19:53, 1 September 2021 (UTC)

We're still subscribed, and still getting notifications.  ;-) Whatamidoing (WMF) (talk) 16:45, 3 September 2021 (UTC)

Flow?

Just curious, what's the relationship of this to WP:Flow? They both seem to fill much the same use case. Is this a successor to Flow? Will they both continue to be developed? -- RoySmith (talk) 22:38, 30 August 2021 (UTC)

Flow hasn't been actively developed since 2015. * Pppery * _{it has begun...} 23:51, 30 August 2021 (UTC)

Flow doesn't use normal wikitext talk pages, and this does.

If Flow had been built fully, it would be far more feature-fun. Imagine a world in which AFD didn't require scripts and bots, and the nomination pages would automatically file themselves in the proper list/category when they were ready to be closed. Or that ArbCom's clerks didn't have to manually count how many Arbs had voted which way, because the software did it for them. This work is a great improvement, but it's never going to be even close to what Flow could have been. Whatamidoing (WMF) (talk) 16:21, 31 August 2021 (UTC)

I'm of mixed-mind about Flow. I didn't particularly like the U/I, but I really don't understand why people hated on it so much. The idea of having structured data that you could reliably manipulate and navigate (as in the examples you gave) made so much sense. Tools like this are great, but trying to do automated things on any human-edited text is way more difficult and less reliable than if the structure of the conversation was rigidly enforced by software. No more conversations getting scrambled because somebody 600 lines up the page forgot a closing curly bracket. Or people using random combinations of ":" and "*" to indicate what's replying to what. But you knew that already. -- RoySmith (talk) 16:41, 31 August 2021 (UTC)

The interface that made it out was not what I had in mind. It's def. not what I designed, and the real Flow doesn't ... work like how the WMF made it work at all. Jorm (talk) 21:35, 31 August 2021 (UTC)

I see a typo in my comment above: "feature-fun" when I meant "feature-ful". But I kind of like it, so I'm not going to fix it. Whatamidoing (WMF) (talk) 19:18, 1 September 2021 (UTC)

Large number of notifications

I log in into Wikipedia today and I see 34 notifications, a number which I had never had the chance to see before in Wikipedia and which reminded me of the beginning times of Facebook notifications. Is that what we want from this function? (I'm not against per se.) Maybe we should do the same as Facebook and other social media did and start grouping some notifications together? Maybe you could have only 1 notification per specific subscription with the names of all the new commenters? Maybe you could have 1 for each commenter (not for each comment)? As I said, having a lot of notifications is not a problem for me really but I'm imagining things could get quickly out of hands in pretty dynamic discussions, especially if you haven't logged in for some hours, and there may be users complaining for this. - Klein Muçi (talk) 23:04, 30 August 2021 (UTC)

They are already grouped for me. Do you mean having the entire group display as just one new notification? CMD (talk) 01:25, 31 August 2021 (UTC)

@Chipmunkdavis, yes I suppose. And maybe not have a specific notification for each reply? I'm talking about having a general notification of this sort:

"X, Y, B and Z replied to [Specific Section]." This would remove the ability to immediately link to a specific reply and just show the discussion in general, maybe the new part of it, but maybe it would be better for some people? Or maybe we could have a preference tab for choosing between these 2 (or more) modes. - Klein Muçi (talk) 09:20, 31 August 2021 (UTC)

I wonder if you're using the no-JS interface of notifications? In the JS interface, they are grouped exactly like you're proposing – you only see one notification per section, and it says something like 34 new replies in "‪dtenable testing"., and it highlights all of the replies in the topic when clicked. But it looks like this feature (called "bundling") has never been implemented in the no-JS version. Matma Rex talk 14:28, 31 August 2021 (UTC)

@Klein Muçi, are you concerned about the number on the bell/inbox icons? Or about the number of lines shown when you click on them? Whatamidoing (WMF) (talk) 17:55, 31 August 2021 (UTC)

@Matma Rex, @Whatamidoing (WMF), I didn't spend too much time to look at the details when I wrote this because, as I said, it didn't concern me much personally. But after your comments now, I was examining it more carefully. I got 5 notifications (the number on the inbox icon) and when clicked, they opened as 1 notification which could be expanded in individual notifications in regard to this post. I guess this is similar to what I proposed, as you mention. My idea was to get 1 notification icon and 1 notification in total per post but I guess this can be fine as well. - Klein Muçi (talk) 19:40, 31 August 2021 (UTC)

I've been using this feature to subscribe to lots of sections and am loving it so far! I think it'd be best to have only one bell/inbox icon but to keep the behavior the same after you actually open up notifications. {{u|Sdkb}} ^talk 05:31, 11 September 2021 (UTC)

I've only subscribed to a few, but can say that the notifications and their grouping is working well for me on both Timeless and mobile/Minerva. (The font for the [subscribe] link in Timeless is wrong; I've delayed logging a ticket, will hopefully do that soon.) I like the idea of having alerts separated from notices so that I can prioritise pings, etc. but, on the other hand, having them all together in Minerva doesn't seem unnatural.
⁓ Pelagic ( messages ) 15:34, 15 September 2021 (UTC)

Master list

I've been continuing to enjoy this feature. One thing I'd like to see is a way to access a list of all the discussions to which you've subscribed, similar to how it's possible to see a list of all the pages you've watchlisted. {{u|Sdkb}} ^talk 20:26, 12 September 2021 (UTC)

@Sdkb I'm glad you posted about the prospect of being able to see all of the topics you have subscribed to. In fact, we have a prototype ready for this very thing that I'm keen to hear what you, and others here, think about it.

Although, before sharing that with you, are you able to share what inspired you to request seeing, "...a way to access a list of all the discussions to which you've subscribed." ? What would you value a page like this helping you to do and/or see? PPelberg (WMF) (talk) 19:06, 13 September 2021 (UTC)

@PPelberg (WMF) My own thought would be that being able to see what discussions one has subscribed too, and being able to unsubscribe, allows one to reduce the number of notifications one is liable to get. So quite a valuable tool, I'd have felt. Nick Moyes (talk) 19:39, 13 September 2021 (UTC)

Backup, mainly; that's the only thing I ever access my list of watchlisted pages for, too (since I'm terrified I'll accidentally click the "clear watchlist" button one day). In this case, my thought process was "I'm already starting to rely on this a lot, so I hope they don't wipe the list of subscribed sections when they move this out of beta. If I had a way to access a list I'd make a backup. Oh, that's something everyone should probably be able to see, so I'll request it". {{u|Sdkb}} ^talk 19:45, 13 September 2021 (UTC)

(wrong topic!)

• @Whatamidoing (WMF): did I read technews correctly that force enabling this feature for every user is imminent? — xaosflux ^Talk 18:40, 20 September 2021 (UTC)
  • @Xaosflux: Tech News says that they're enabling the reply tool, the bit that adds a "reply" link under comments, not the topic subscription feature. 192.76.8.74 (talk) 18:56, 20 September 2021 (UTC)
  • Oops, so many things going on - I should only read one of these updates at a time! — xaosflux ^Talk 19:04, 20 September 2021 (UTC)

    Aww, you'd take all the fun out of it.

    Also, "imminent" is not the right word, especially here at the English Wikipedia. I'm still talking to Peter about that. Tech/News will have the same announcement next week, and then it will go out to all the small-to-mid-sized Wikipedias and sister projects. I think that the Turkish Wikipedia will be the largest affected one next week. After that (hopefully soon after that), it'll be out for everyone at Wikidata and Commons (and a few similar multi-lingual projects), and the last batch is the remaining heavyweights: English, German, French, and Russian Wikipedias. I hope that will still qualify as "soon", but it's not quite "imminent", and there is no specific date set for that batch yet. Whatamidoing (WMF) (talk) 21:27, 20 September 2021 (UTC)

The Image Sizes

The image length cannot be changed without affecting the image width, request the ability to change the image length without affecting the image width as well as the ability to change the image width without affecting the image length.Mohmad Abdul sahib 08:53, 12 September 2021 (UTC)

It sounds like you want to display only part of an image that already exists on Wikipedia. {{CSS image crop}} or {{Annotated image}} might work for you. – Jonesey95 (talk) 15:52, 13 September 2021 (UTC)

Jonesey95

No, for example, I want to change the width of the image without forcibly changing the length of the image, leave notifications for me next time.Mohmad Abdul sahib — Preceding undated comment added 16:34, 13 September 2021

@Mohmad Abdul sahib: Nothing to do with your problem, but please read through WP:SIGNATURE. I've just made two fixes to your signatures in this little section. (Also, you can easily add the pages you edit to your watchlist: Help:Watchlist.) Thanks, — JohnFromPinckney (talk / edits) 16:44, 13 September 2021 (UTC)

@Mohmad Abdul sahib: And about your problem: aren't you worried about stretching or squishing the image? It will be distorted if you change only one dimension. — JohnFromPinckney (talk / edits) 16:45, 13 September 2021 (UTC)

JohnFromPinckney I know about this, this problem was in old versions of Adobe Photoshop and the problem of this distortion has been resolved in newer versions, so you can use the same technique.Mohmad Abdul sahib 19:57, 13 September 2021 (UTC)

That's not how the wiki software works and it will not work that way any time soon. Aspect ratio is important. If you think a free image would look better with a different aspect ratio, you may crop them and subsequently contribute them to Commons. (Don't modify non-free images.)

Please do not amend the timestamps of previous posts. Izno (talk) 06:23, 14 September 2021 (UTC)

Izno The image dimensions are distorted if the display is changed excessively, but if the size is slightly changed, the image will not be distorted.Mohmad Abdul sahib 10:48, 14 September 2021 (UTC)

No, it still distorts the image.

I will not be responding further since you don't seem to be interested in listening. Izno (talk) 16:07, 14 September 2021 (UTC)

Izno With the presence of this algorithm (Seam carving), doing this kind of resizing the image does not cause it to be distorted, so I ask to be able to resize the image with this algorithm.Mohmad Abdul sahib 13:22, 15 September 2021 (UTC)

As noted above, there are lots of ways to change the aspect ratio of an image, all of which distort it in some way. Using the example image in Seam carving, measure the angle between the top of the castle, the man, and the base of the castle. Do that in both the original and the seam carved version. Those angles will be different. Is that important? It might be, it might not. Only a human can determine that.

If you want to generate a seam carved variant, download the image (assuming it's appropriately licensed), perform whatever manipulations you want, and upload the new image. I often do this for images that could be improved. Cropping, perspective correction, rotation, and many other types of image manipulations are all fair game. -- RoySmith (talk) 22:04, 20 September 2021 (UTC)

New tool to identify sockpuppets based on writing style

Checkusers on the English Wikipedia will soon have access to a new tool aimed at identifying misuse of multiple accounts based on a person's writing style. masz, developed by Ladsgroup, uses natural language processing to create an individual 'fingerprint' of a user based on the way they use language on talk pages. Checkusers can log into a web interface to compare the fingerprints of two accounts or list accounts with similar fingerprints. The tool is already live on several projects and is expected to start running on enwiki after phab:T290793 is resolved. – Joe (talk) 07:15, 13 September 2021 (UTC)

• Comment: I'm not involved in the development of this tool, so I can't answer any questions about how it works beyond this. – Joe (talk) 07:18, 13 September 2021 (UTC)
• Sounds like a GDPR violation but okay. (article 22 on automated decision-making and profiling) Stifle (talk) 08:33, 13 September 2021 (UTC)

  I'm quite certain that it is not a violation. Art. 22 par. 1 GDPR provides: "The data subject shall have the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning him or her or similarly significantly affects him or her." In our use case, the decision to block is not "based solely on automated processing", because it is made by a human checkuser considering the fingerprint together with other evidence. Also, a block from editing Wikipedia is not a "legal effect" or something of similar significance. In any case, the GDPR does not apply in the US, where the Wikimedia Foundation is located. Sandstein 08:43, 13 September 2021 (UTC)

  Besides while not blocking accounts, we are already blocking and tagging edits based solely automated processing, including profiling so if the GDPR applies to actions we take against editors, we already have problems. Definitely some of those affected e.g. the editor complaining about being libelled recently on ANI because their edit was tagged as a possible vandalism or BLP violation and some of those trying to add nonsense to Adam's Bridge or complaining about the Ahmadiyya Caliphate seem to think they've been significantly affected by our automated processing. Nil Einne (talk) 11:54, 13 September 2021 (UTC)

  Uh, how could that possibly apply? "The data subject shall have the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects' concerning him or her or similarly significantly affects him or her." The tool does not automatically ban people; an administrator has to ultimately review and apply the ban based on all available evidence. Therefore it will never be the sole determinant. And in any case being banned from Wikipedia, while some people may not like it, is not a legal effect or anything of remotely comparable impact. --Aquillion (talk) 14:32, 13 September 2021 (UTC)

I have no doubt that this is legal (per Sandstein, although whether the WMF is bound by the GDPR is murky; many US-based newspapers stopped serving EU-geolocated IP addresses when it went into effect, I suppose because of a possible argument that the content and ads being read at a EU location creates EU jurisdiction).

We can certainly ask whether it is wise, though. False positives are bound to happen and I do not look forward to bans based on unclear behavioural evidence because computer says so. Also, it is well-known that such systems are subject to bias laundering; for instance, if the training set contains few examples of the Cheshire dialect, then two users using that dialect becomes irresistible evidence of socking. Tigraan^{Click here for my talk page ("private" contact)} 09:04, 13 September 2021 (UTC)

The highest EU court (CJEU) has argued that GDPR applies to non EU companies, when it involved EU subjects in Google v CNIL (C-507/17). The territoriality of that has deep implications, but I have no doubt that the WMF legal has examined these. ~ Shushugah (he/him • talk) 13:35, 13 September 2021 (UTC)

References

Given these fingerprints are based solely on public information(correct me if I am wrong), is there a good reason to limit it to checkusers? Other edit analysis tools are not limits. I would be interested in checking a few hunches.

Regarding concerns of people being blocked based on computer evidence, I am assuming this will be used as an investigation tool and not used as the sole evidence of a block. We already point out similar writing styles when making a sock case, this would only help us find them not decide what weight we give them. HighInBC ^{Need help? Just ask.} 09:37, 13 September 2021 (UTC)

• *Whistles and admires slippery slope* I might have less of an issue with the idea of something like this, if it didn't sound so much like giving CSI-style tools to untrained admins. The suggestion that this be expanded to all users (rather than just checkusers) makes me even more leery. I'm no expert, but my job means I deal with dozens of writing samples every day. It's hard for me - with two pieces of writing side-by-side - to tell the difference sometimes. It's also my experience that cultural factors play a huge part in determining why someone might write like someone else. Were they educated in the same school system? Did they have the same English teacher? Did they participate in the same ICU chat-groups as a teenager? Did they use the same app to learn English? Did they learn English from the same base language? I see those dozens of writing samples every day and I could find you similarities between any two selected at random, simply because of context. Alcohol, recreational and prescription drugs, stress, emotion, fatigue, and fluency of language are all factors in making a piece of writing similar to another, or different. You can't change your IP address based on how many drinks you've had. The same person can write two different things under different conditions and have them be very different. But two different people can write about similar things under similar conditions and the writing they produce will be very similar. St★lwart¹¹¹ 10:52, 13 September 2021 (UTC)

• Do not trust algorithms to pick up on the nuances of language usage and language evolution among disparate individuals. Random example: in the pre-2010s, I was using the word programme a lot (was doing a lot of UK'ish content work then), but in the post-2010s I pretty much exclusively use program. Is that something this computer programmmmme will not want? I'm Skynet'ing this. El_C 11:14, 13 September 2021 (UTC)

• El_C, exactly. And what does it tell us if your use of the word programme is similar to someone else's use of the word programme? Is that confirmation they are the same person, or is it simply confirmation that they fall within the same demographic category. Does falling within several of the same demographic categories make the pair more or less likely to be the same person? Is that confirmation they are the same person, or simply confirmation they are in a similar demographic category and so similarities should be dismissed? St★lwart¹¹¹ 01:14, 14 September 2021 (UTC)

• Stylometry is difficult – people have spent decades trying to improve AI methods for authorship identification/attribution (with, albeit, some promising results), and professional forensic linguists have dedicated whole careers to end up producing what can sometimes be low confidence or ultimately inadmissible evidence in court regarding authorship analysis. The method Ladsgroup is using, looks to be extremely simple from the displayed graphs (counting the relative frequency of commonly used words) though his code is private so who knows. I would be extremely cautious in using this tool to infer two accounts are likely sockpuppets due to language similarities, and extremely cautious in using this tool to infer two accounts are unlikely to be sockpuppets due to a lack of similarity in their language. "Use this tool but do not generally trust its output" sounds like the perfect way to bolster your confirmation bias—"They're similar? That's proof they're sockpuppets! They're dissimilar? Well, this tool is imperfect..."—so I heavily question the worth this program has for SPIs. ‑‑Volteer1 (talk) 12:09, 13 September 2021 (UTC)
  • Well said. The usage of the Editor interaction utility at SPI and beyond just confounds me. As I've said elsewhere recently (AN diff in response to its misuse, aside from being a novelty, the thing is pretty much useless. El_C 12:40, 13 September 2021 (UTC)
    • I think it does have uses, but only with caution. I find it useful as a starting point when I suspect two accounts are sockpuppets - a way to find places to look where they may have interacted in ways that support them being sockpuppets (eg. a user overtly restoring the exact edit of a banned user a year later, or one sock glaringly hyping up or defending another across multiple unrelated pages.) For example, in a recent case I noticed an account with very few edits defending another account; glancing at the interaction utility's results let me easily see that they had done this across multiple pages and helped me build a case. I could have done this without the tool, but it would have been more time-consuming. You said in your post that there's no shortcut to diff evidence, but the Editor Overlap Tool is in fact exactly such a shortcut, in that it makes it easier to find relevant diffs that can be used as actual evidence. When I suspect that two editors are sockpuppets of each other it's also useful as an at-a-glance way to help me figure out if it's worth the time to try and build a case. I'd never just point to it in a SPI case, since that's barely more useful than pointing at an editor's entire edit history - going over the interactions and finding relevant diffs (or if they even exist) is fairly time-consuming - but it is useful in the early stages of an investigation. I would expect the "list accounts with similar fingerprints" thing would be similar to eg. taking an account who is behaving in ways that make you suspect a sockpuppet, and checking what blocked users (who were blocked before they began editing) have edited pages they edit - useful as one possible starting point for an investigation, no more. Actual evidence would still be required to go from there. --Aquillion (talk) 14:48, 13 September 2021 (UTC)
      • Fair point. I suppose my point is that I see a ton of its misuse, often resulting in distortion that leads to conflict for naught. The example highlighted in that AN diff I cited above illustrates this pretty well, I think — I'd also point to the preceding comment by Softlavender (on the misuse of the interaction tool as suffered by GS) to which my note was in response to. Anyway, returning to Volteer1's point: inaccurate reading accompanied by confirmation bias is a paramount concern when it comes to analyses that are in part AI-derived. Would hammering these caveats be enough to make adoption of this tool worthwhile? I have my doubts, but I suppose time will tell. El_C 15:21, 13 September 2021 (UTC)
        • I guess useful as an at-a-glance way to help me figure out if it's worth the time to try and build a case. I'd never just point to it in a SPI case, since that's barely more useful than pointing at an editor's entire edit history... Actual evidence would still be required to go from there is a fairly benign use for such a tool that wouldn't exactly be what I was worrying about. I don't know enough about SPI to know if that is the extent of how it would be used, but if it's just a pointer to see if gathering actual evidence is more likely to be worth your time it's probably not that problematic (though you may want to empirically test its usefulness first). ‑‑Volteer1 (talk) 16:00, 13 September 2021 (UTC)

          That's definitely how I use the Editor Interaction Analyzer. Sure, anyone can run it on two editors, wave their hand at the amount of overlap, and say "See‽", but that really only plays well among conspiracy theorists on Wikipediocracy. At SPI, it's little more than a starting point. In some cases where an overlap is particularly obvious (say, a dozen AfDs where two low-edit-count accounts commented a minute apart), it might "speak for itself", just as an editor's contribs might. But no one's getting indeffed because they happen to have some overlap with some other editor. In general, it's just a good way to find diffs to look at. For instance, I filed an SPI that had massive overlap among the accounts, but that wasn't in itself surprising because they all edited in the same topic area; you would have found lots of overlap comparing them to their opponents in the area too. Instead I had to go through each page and each diff to figure out whether the overlap looked innocent or whether there were signs of sock/meatpuppetry. Ultimately there will never be a substitute for that; just tools that help us aggregate suspicious diffs better. -- Tamzin^{[cetacean needed]} (she/they) 18:39, 13 September 2021 (UTC)

• I may not use the tool myself but I don't see any concern, at least not any that doesn't also apply to the checkuser tool itself. Checkuser already reveals when one person is using multiple accounts, and then it's up to the human operator to determine if the use constitutes a violation of the multiple accounts policy, which is what blocks are based on. Neither checkuser nor this new tool actually create blocks. I am concerned though with making the results of this program available to the public, it would definitely be used for abuse, like to out editors with valid undisclosed privacy accounts. If access is limited to checkusers (who are selected for proficiency in handling sensitive information) then it's less of a concern, but I guess the fact that Ladsgroup created this means that anyone else could create their own if they really wanted to. Ivanvector's squirrel (^trees/_nuts) 13:45, 13 September 2021 (UTC)

  It is probable that such a tool could be easily created, or already has been. I believe Ladsbroke is absolutely entitled legally and morally to use it (assuming it does not break API query rate limits), and share the code to the world if they so wish (security by obscurity is not a solution). Whether "we" (the community? the WMF? the developers?) should approve its use for official CU-like purposes is another issue entirely.

You are supposed to only use CU when you have sufficient evidence already. In contrast, such a tool would be used to create such evidence (otherwise, one would go straight to CU), and therefore would need to be run against everyone (or everyone who edited such-and-such article, or meets certain criteria that would be insufficient for a CU).

Checkuser creates log which can, at least in theory, be audited; "fishing expeditions" initiated without behavioural evidence would bring unwanted attention to their author. In contrast, I fail to envision a use policy for that tool that would meaningfully restrict its use. (If such a policy is produced, I will gladly withdraw my objection to its use by checkusers or admins or potentially anyone.) Tigraan^{Click here for my talk page ("private" contact)} 16:13, 15 September 2021 (UTC)

• I think it sounds useful, but it is important to train people properly in its use (and, more importantly, its limitations), the same way we do with existing SPI tools. It is just one flawed and limited indicator... but since all our indicators are, generally, flawed and limited in some way, it is useful to have if used properly and in concert with them. --Aquillion (talk) 14:35, 13 September 2021 (UTC)
• When the tool is available for enwiki, someone should run it on, say, 100 closed sockpuppet cases (to see how many false negatives it produces), and on 100 definitely-not-sockpuppetcases (both sock cases which turned out to be false, and e.g. between 2 different CUs) to check how mony false positives you get. Only then can some decision be made whether it can be trusted somewhat or not at all. Fram (talk) 15:40, 13 September 2021 (UTC)

  I don't know if I will individually get up to 100 of each but a version of this is definitely my plan once it's available on enwiki. Best, Barkeep49 (talk) 15:42, 13 September 2021 (UTC)

• Thank you Joe for bringing this up. Regardless of its accuracy, I have several issues with this tool.
  Firstly, and perhaps the most disconcerting, this tool appears to build a digital fingerprint of every single enwiki editor with more than a minor number of edits. (I believe the threshold is 100, but obviously it could be raised or lowered depending on the number of edits required for the tool to build the fingerprint.) Currently, checkusers and other editors comparing writing styles can only compare suspect accounts—it is an impossibility to compare a suspect sock against every other Wikipedia editor. This keeps investigations focused.
  Secondly, this fingerprint is perpetual, which is in direct contrast to the checkuser tool, which only stores IP and user agent information for 90 days. I don't know how this interacts with GDPR, since I am neither a lawyer nor European.
  Thirdly, the information that can be gained from a digital fingerprint of your writing style is significantly more profiling than even most checkuser data. Checkuser data tells checkusers who view it your rough geographic location and the web browser that you use. In rare cases, it can also tell us what company you work for or school you go to, if your company or school owns their own subnet. Other than that case, checkuser data tells us nothing about your age, gender, and other personally identifiable information. A digital fingerprint of your writing style and articles edited can potentially be used to build a significantly more detailed profile of you. (See what Google and Facebook do as examples.) Your writing style alone gives a strong indication of your gender and age. It also tells what your native language likely is, where and how you learned English, and potentially more information about you as a person. A technique very similar to this is what advertising companies do to display ads relevant to you on websites. Obviously, this particular tool does not attempt to profile editors, and I am in no way attempting to claim that checkusers are building shadow profiles of every user. I am simply pointing out what information could be gleaned from a fingerprint.
  Sorry, but as I mentioned several months ago when this idea was first proposed on the checkuser mailing list, I don't think it's a good idea. I do not think that most users on here were aware of the potential power of machine learning when they made their first edits. Yes, I am aware that all edits are public. No, I don't think that we should do something just because we can. I am open to being convinced otherwise if the community does not mind the use of tool(s) like this. Reaper Eternal (talk) 16:46, 13 September 2021 (UTC)
• Am I missing a link to a page with more information on how the tool works, or is that too limited to checkusers? meta:User:Ladsgroup/masz presents two sample graphs, which do not present a clear threshold for distinction (there are differences between the socks, and close similarities between the non-socks). That suggests a lot of interpretation would still be needed, going back to diff comparisons or similar. I do feel, similar to as has been noted above, that there is great potential use for such a tool in indicating possible other socks, similar to the way CU checks currently do. A tool which pointed out for example where a sock diff had been fully/partially reinstated for example would save a lot of time diff searching. However, if the use is just to point out areas that would be great for analysis, this is not too helpful if limited just to CUs, who don't seem to have the time to efficiently cover the existing SPI process where diffs are expected to be already found. Given what sounds like limited benefit, and the potential drawbacks covered above, I am wary of adding further to CU/admin burden. CMD (talk) 17:16, 13 September 2021 (UTC)
• I support improved tooling for SPI and CheckUser. Although, given the potential of state-of-the-art NLP, I think this should be subject to usage auditing comparable to that of the rest of CheckUser technical evidence. MarioGom (talk) 18:46, 13 September 2021 (UTC)
• I know I had comments about this idea when it was first initially proposed, but I don't really want to trudge through (probably mailing list) archives to find it. While this tool may be useful, I'm concerned about two things: false positives and fishing. CheckUser is a technical tool requiring human interpretation to be useful. This tool is not. We've had problems in the past with people seeing large numbers in toolforge:copyvios and thinking that means something, without actually checking. This tool, as I understand it, allows for that same big number bias. I also don't want to see this tool being used as a pretext for CheckUser checks that would otherwise be called fishing. I don't think this tool should be used by any CUs until and unless it has been established that it is sensitive enough and does not introduce new bias and there is policy guidance for its use. --AntiCompositeNumber (talk) 02:25, 14 September 2021 (UTC)

  Those comments were in this wikitech-l thread. --AntiCompositeNumber (talk) 02:30, 14 September 2021 (UTC)

• There is definitely an ethical element to this debate. On the one hand, we have public information, which can be manipulated and analysed. It's fair to say that our users (checkusers, yes, but also other editors) do analyse edits, compare stylistic and linguistic differences - which is roughly what I understand this tool to do. What's more, it's also fair to say that AI has been a massive boon to Wikipedia over the years. I have considered writing a similar program myself, but never had the time or energy to put something like it together.
  On the other hand, there's the question of "should we do something just because we can?". This tool will mass profile our editors. Call it a digital fingerprint if you like, but it is an identifier available. I don't expect our Checkuser team to abuse such information, but I can think of many abuses of it - As a member of the Arbitration Committee, I've seen many disclosures of Legitimate Alternate Accounts, where individuals would be targeted in real life if their editing was known. I can certainly imagine situations where having that sort of information could be used by unscrupulous individuals for harassment. What's more, depending on the output of this fingerprint, it could be used beyond Wikipedia - to identify individuals who have written blog posts and twitter posts.
  What's more, Wikipedia has a unique setup - it's history cannot be deleted easily, every page is a culmination of every edit that happened in the history. As such, we don't offer a real ability to remove yourself from Wikipedia - we don't have the option to have all your edits deleted. Combined with the fact that edits made years ago would not have come close to expecting this digital fingerprint, we are saying that 1) we're going to profile you and 2) there's nothing you can do about it. But it's ok, because we won't use it to do anything besides check if you've been abusing our Sockpuppetry policies - and only be given to a select number of individuals.
  Perhaps the cat is out of the bag - AI is progressing at an exponential rate, and if it wasn't this tool, it would be someone else's. But let's be clear, there are things we should be considering here. I, for one, am extremely uncomfortable. Worm^TT(talk) 09:16, 14 September 2021 (UTC)
• I'm a bit concerned that there hasn't been any published information on the effectiveness of the tool or on a plan to validate its effectiveness. New tools for checkusers can be good, but we ought to ensure the tools are evaluated (with a feedback loop for improvements), and proper guidance given to checkusers on how to take best advantage of the tools. isaacl (talk) 00:46, 15 September 2021 (UTC)

• I remember attending a presentation at Wikimania a couple years ago on the subject of algorithm-assisted sockpuppetry detection. It left an impression on me more than most of the sessions I was in. It strikes me as extremely important to improve our sock puppet detection capabilities, as one of our most glaring weaknesses and, if I may be a little dramatic, one of the greatest threats to the integrity of the project amid declining good faith participation and shrinking admin pool. I had reservations at the Wikimania session largely due to the premise, which if I recall correctly was about automatic detection that would then be checked. Having such software always running and detecting matches would cause any problems with the algorithms to cause damage at a large scale, and errs on the side of outing legit socks. (Tangent: I still think we don't have clear enough processes documented about legit socks, rules for use, limits, and protections thereof. It's one of the only reasons I don't advocate for more liberal usage of checkuser in general.) That said, the way I'm reading this, it's just another tool in the checkuser toolbox, employed as needed when a suspicion arises. More information is good. Like others, I'm curious to see this in action, but love that it's happening. — Rhododendrites ^talk \\ 01:30, 16 September 2021 (UTC)
• I take what's being said about our need to identify socks. Is this tool reliable in doing so or will it find too many false positives? There are so many opportunities. Most obviously, editors with a (non-English) first language in common will often write English with similarly distinctive vocabulary, grammar and sentence structure carried over from their first language, as well as often having similar national concerns. Many of us switch registers to communicate and fit in; our Wikipedia "voice" may be only one of our voices and it may subtly conform with others. The tool should be tested on editors well-understood to be different but editing in the same areas, as well as on the more exciting cases of sockmasters who made a special effort to write differently (eg).
  If the tool works well, could it dox? Suppose an editor (not me) used to edit under their real name, took a long break and came back with an ordinary pseudonym. If the tool runs on all history, would it link those accounts? NebY (talk) 21:26, 16 September 2021 (UTC)

  It would not be useful for fishing out random connections as the tool would not be that precise. It would not be able to distinguish 100,000s of different editors. So it could connect edits, but not very precisely say that this is the same person, as there could be hundreds that write the same way. If there is additional suspicion, eg edits to the same article then it would be much more likely that it was the same person if the editing was in the same style. Graeme Bartlett (talk) 23:12, 16 September 2021 (UTC)

  Currently all it does is give you a list of users with similar language patterns, and depending on what areas of the project they are active on it is of limited usefulness. There's still a substantial degree of human judgement involved, and I would never base a check or block solely on it. TonyBallioni (talk) 02:52, 17 September 2021 (UTC)

• I'll try to summarize my various thoughts on this from the list here: all of this information is public, and there were academics studying how to do this before Ladsgroup started it. A significant number of the edits we make are already put under analysis by a machine either automatically or at the discretion of anyone with an internet connection (see: ORES, ClueBot, Editor Interaction Utility, Interaction Timeline, and I'm sure there are others.)
  While I get there are different views on the ethics of using data, everything on Wikipedia is by it's very definition public, and we can assume that users who have been here long enough (for whom the privacy concerns mainly apply), are aware that everything they post is public. Every time they push the edit button, they agree to make their contributions public and be used in any way they want. I personally don't see an issue with that, especially when there's a human at the other end making the ultimate judgement call. Ultimately, there's really not all that much the community can do on this: while Ladsgroup would likely be willing to disable it on en.wiki if asked, eventually someone is going to create something like this and make it available for anti-abuse purposes, and they won't be willing to turn it off, nor would there be anyway to require them to, and efforts to restrict people from using it would be just as ethically dubious: telling people what they can or can't do on their home computers with data that is public and where the actions don't directly impact others really isn't something that as a community we should be trying to do, in my view.
  So, yes, I think letting people know this is a capability that someone's developed is fine. It's also been a capability others have been testing for a while. There's not really much new here except Ladsgroup has made it easy to do. There's also really not much that can be done about it from a policy angle, and I personally don't see any ethical issues with analyzing publicly available data by people who knew that their posts would be public. TonyBallioni (talk) 02:52, 17 September 2021 (UTC)

  I agree that checkusers (or any editor) can use any tool they want off-wiki, and there isn't a lot of scope to try to limit this. I think it would be helpful nonetheless for there to be guidance on the effectiveness (or ineffectiveness) of popular tools, as well as guidance on best practices in using them. isaacl (talk) 03:22, 17 September 2021 (UTC)

  Yeah, I'd agree. I think that's something we could talk to Ladsgroup about/develop as it is used. Still relatively new globally, and hasn't been rolled out on en.wiki. TonyBallioni (talk) 03:29, 17 September 2021 (UTC)

My first thought is being reminded of some bits of that business in 2019 (1, 2, 3... I wasn't around then and maybe I shouldn't bring the matter up. I don't know...). I don't know what to think about this present tool, but I suppose I lean Worm-wards. —2d37 (talk) 09:19, 17 September 2021 (UTC)

• I personally don't see why this needs to be restricted to checkusers. CUs have access to private data; the data here isn't private, in the sense that anybody can download the raw data from the WMF data servers and do whatever analysis they want on it. Well, at least I'm assuming there's no non-private data involved; correct me if I'm wrong about that.

As far as blindly using the tool, it's just a tool. And like all tools, has to be used properly, which means applying human judgement. In my own work as an SPI clerk, I use most of the tools mentioned here (and have written some of my own). But it's all just guidance and assistance to sift through the morass of raw data. It would be a terrible mistake for anybody to use any tool and block somebody because the tool returned a value greater than 0.5, or whatever. At the same time, I fully understand that doing things in scale changes them in fundamental ways. I have no problem with the fact that when I walk down the street, people can see me. I do have a problem with there being cameras on every street corner recording every move I make. On the other hand, when my car was broken into a while ago, I really wished there were cameras on every street corner near my house so we could have caught the person who did it.

I am befuddled by this statement that "The code for it is licensed GPLv3 but it's not public". I don't understand how something that's GPLv3 can be "not public". And the fact that the code's not public bothers me a lot, because it precludes broad analysis of the algorithms. -- RoySmith (talk) 06:59, 21 September 2021 (UTC)

@RoySmith these were discussed on wikitech (see para 5 of OP for why code is private). – SD0001 (talk) 13:32, 21 September 2021 (UTC)

Is viewing IP edit history on mobile removed now?

Tracked in Phabricator
Task T291233

I’ve tried clicking the edit history of IPs including mine but they only show the invalid user error and I’m wondering if the IP edit history has been removed on mobile. 2600:1003:B8DE:FBCA:B8C1:C57:CEF0:F1EC (talk) 00:40, 17 September 2021 (UTC)

• It's a bug, standby for a tracking number; WP:ITSTHURSDAY. — xaosflux ^Talk 01:06, 17 September 2021 (UTC)
• phab:T291233 opened on this bug. — xaosflux ^Talk 01:09, 17 September 2021 (UTC)
  • For temporary workaround, use the desktop view at the bottom of the page first. — xaosflux ^Talk 01:10, 17 September 2021 (UTC)
  • For those with access to the options panel, selecting "advanced mode" also is a work around. — xaosflux ^Talk 01:30, 17 September 2021 (UTC)
• FYI: Bug has been confirmed, pending patch acceptance and deployment. — xaosflux ^Talk 13:08, 17 September 2021 (UTC)

Can't see IP contributions in mobile view

[[1]] leads to a bad username error message

[[2]] resolves without issue.

Bug? New Feature? Slywriter (talk) 17:01, 22 September 2021 (UTC)

@Slywriter: see above. — xaosflux ^Talk 17:04, 22 September 2021 (UTC)

Thanks. Was about to strike my post as I noticed it was staring right at me. Slywriter (talk) 17:07, 22 September 2021 (UTC)

Toward a standard for editing tools vis-a-vis signatures.

We've got lots of tools that let you type stuff into a box and add it to a page. Some of them automatically add your signature, some don't. Some of them give you a live preview. Some offer the ability to generate a preview manually. Some have no way to preview. The bottom line is I often think a tool will add my signature, then discover it doesn't, and have to go back and add it later.

We should standardize how this works. I think the ideal case is what I'm using now (Reply Tool) does; give you a live preview, and add your sig automatically. But, at the very least, there should be some visual indication whether you signature is going to be added automatically or if you need to do the squiggle thing. That would at least give tool developers guidance on how they should be building their tools. -- RoySmith (talk) 15:07, 19 September 2021 (UTC)

(For those who don't know, I note that Reply Tool (and New Discussion Tool) gives some visual indication that one's signature will be added, namely that the signature is shown, translucent, at the end of the live preview.) —2d37 (talk) 20:59, 19 September 2021 (UTC)

The live preview is only in the source mode. Whatamidoing (WMF) (talk) 00:04, 21 September 2021 (UTC)

Javascript error

Hello, I went on a script installation spree earlier then I started getting an error message on the right side of my screen. The error says "Failed to set referrer policy: The value 'origin-when-crossorigin' is not one of 'no-referrer', 'no-referrer-when-downgrade', 'same-origin', 'origin', 'strict-origin', 'origin-when-cross-origin', 'strict-origin-when-cross-origin' or 'unsafe-url'.

resolveStubbornly — load.php:8:543"

I have tried bypassing my browser cache.

I use a Safari browser Version 14.1 and macOS Big Sur Version 11.3

Many thanks for the help. Princess of Ara 22:49, 19 September 2021 (UTC)

@Princess of Ara: turn off all those scripts, then try them one at a time. — xaosflux ^Talk 01:09, 20 September 2021 (UTC)

You probably also shouldn't install foreign project user scripts if you don't review them first, and ensure you trust the editor (e.g. you are now loading w:ru:Участник:Vlsergey/wef.js). — xaosflux ^Talk 01:12, 20 September 2021 (UTC)

You can safely ignore that warning, MediaWiki emits two referrer policies in the HTML of each page. origin-when-crossorigin is non-standardized but used to support older browsers. T248526 and T180921 have some more details. Legoktm (talk) 08:01, 20 September 2021 (UTC)

Thanks for the response. I'm not exactly technically inclined but I gather that the problem has to do with Safari itself. To save myself the headache, I've made chrome my default browser and not experiencing any issues so far! Im grateful. Princess of Ara 08:25, 20 September 2021 (UTC)

@Princess of Ara, no there's nothing wrong with your browser or Safari. In order to support older browsers we have to add the extra code, which causes a harmless warning on some newer ones - but you can just ignore it. Legoktm (talk) 19:11, 22 September 2021 (UTC)

"Limited Access" redirect error.

Resolved

Firstly, apologies if this is the wrong place to report this, I'm unfamiliar with the inner workings of Wikipedia. If you click on the likes of the free access symbol (), you are redirected to a page discussing the meaning of free access in regards to written material. However, if you click on the limited access symbol (. you are redirected to a page about limited access roads. I don't know how to correct this, and hoping somebody here will. Cheers! Xx78900 (talk) 13:38, 20 September 2021 (UTC)

@Xx78900: the target for that comes from the link=Limited access part on Template:Limited access, change the part after "link=" to the title of the article you want this to go to instead. — xaosflux ^Talk 13:47, 20 September 2021 (UTC)

@Xaosflux: Thanks, I've changed it now! :) Xx78900 (talk) 13:54, 20 September 2021 (UTC)

And I've removed the link to Freemium because that is only one possible meaning of 'limited access' which can also mean a cap on daily views, a restriction to certain day or night times, or providing the contents only to certain IP ranges/locales on behalf of the provider of the source, etc.

—Trappist the monk (talk) 14:13, 20 September 2021 (UTC)

Tech News: 2021-38

Latest tech news from the Wikimedia technical community. Please tell other users about these changes. Not all changes will affect you. Translations are available.

Recent changes

• Growth features are now deployed to almost all Wikipedias. For the majority of small Wikipedias, the features are only available for experienced users, to test the features and configure them. Features will be available for newcomers starting on 20 September 2021.
• MediaWiki had a feature that would highlight local links to short articles in a different style. Each user could pick the size at which "stubs" would be highlighted. This feature was very bad for performance, and following a consultation, has been removed. [3]
• A technical change was made to the MonoBook skin to allow for easier maintenance and upkeep. This has resulted in some minor changes to HTML that make MonoBook's HTML consistent with other skins. Efforts have been made to minimize the impact on editors, but please ping Jon (WMF) on wiki or in phabricator if any problems are reported.

Problems

• There was a problem with search last week. Many search requests did not work for 2 hours because of an accidental restart of the search servers. [4]

Changes later this week

• The new version of MediaWiki will be on test wikis and MediaWiki.org from 21 September. It will be on non-Wikipedia wikis and some Wikipedias from 22 September. It will be on all wikis from 23 September (calendar).
• The meta=proofreadpage API has changed. The piprop parameter has been renamed to prpiprop. API users should update their code to avoid unrecognized parameter warnings. Pywikibot users should upgrade to 6.6.0. [5]

Future changes

• The Reply tool will be deployed to the remaining wikis in the coming weeks. It is currently part of "Discussion tools" in Beta features at most wikis. You will be able to turn it off in Editing Preferences. [6]
• The previously announced change to how you obtain tokens from the API has been delayed to September 21 because of an incompatibility with Pywikibot. Bot operators using Pywikibot can follow T291202 for progress on a fix, and should plan to upgrade to 6.6.1 when it is released.

Tech news prepared by Tech News writers and posted by bot • Contribute • Translate • Get help • Give feedback • Subscribe or unsubscribe.

18:29, 20 September 2021 (UTC)

Delete redirect

Can someone please delete Kongsberg Defence & Aerospaces redirect to Kongsberg Defence Systems?--Znuddel (talk) 19:43, 20 September 2021 (UTC)

@Znuddel This belongs at WP:RFD. (Unless it can be speedily deleted.) ― Qwerfjkltalk 19:50, 20 September 2021 (UTC)

You will have to give a good reason. It's a former name and the redirect is helpful. PrimeHunter (talk) 19:52, 20 September 2021 (UTC)

Linking to special page, block log, etc.

Hi. I want to create an internal link to my block log for my bio page. How can I render this as a blue link? Piotr Jr. (talk) 21:03, 20 September 2021 (UTC)

• URL query strings are not supported in wikilinks directly. You can style the external link several ways, see Help:Link#Links_containing_URL_query_strings. — xaosflux ^Talk 21:14, 20 September 2021 (UTC)

Thank you. Piotr Jr. (talk) 03:19, 21 September 2021 (UTC)

@Piotr Jr.: If etc means you want other links then many special pages can be wikilinked with a parameter. Special:Log can be wikilinked by type+performer but not by target as far as I know. Special:Log/block/Piotr Jr. is blocks performed by you (none since you aren't an admin). Some others like Special:Log/create/Piotr Jr. and Special:Log/move/Piotr Jr. have entries. Special:Log/Piotr Jr. is most logs by you. It says "All public logs" but that's inaccurate. PrimeHunter (talk) 11:16, 21 September 2021 (UTC)

@PrimeHunter Is this documented anywhere? ― Qwerfjkltalk 19:41, 21 September 2021 (UTC)

It's in the MediaWiki documentation under Help:Log, at the bottom. William Avery (talk) 20:04, 21 September 2021 (UTC)

And partially at Help:Log#Viewing logs which is linked on "Help" at the top right of Special:Log. MediaWiki documentation is scattered and incomplete. Both pages are probably mainly written by volunteer editors with what they know. Some of the MediaWiki page was copied from the English Wikipedia. PrimeHunter (talk) 20:19, 21 September 2021 (UTC)

Template question

Hello, folks,

I was deleting an expired draft and when I looked at the page creator's contributions, I found they also started a template for this wanna be artist that was blank (Template:Infobox artist Carla Pricop). I regularly look at database reports and reports compiled by bots that reported orphaned talk pages, expiring drafts, empty categories, broken redirects, etc. but I've never come across a report that would have reported a template page like this, blank & uncategorized, so that it could either be nominated at TFD or tagged for speedy deletion. Is anyone here aware of a bot that handles this or reports that are generated on things like blank template pages? Coming across this page just made me wonder if we might have a lot of unused, misplaced junk lying around in the Template space.

I've also found that some new editors who can't start a new article in the main space, start a template page instead and put content on it like they should be putting on a Draft or User page.

I looked at Template talk pages where I might post this query but the ones I came across don't get a lot of traffic so I'm posting this here. Thank you. Liz ^{Read! Talk!} 23:12, 21 September 2021 (UTC)

There is Wikipedia:Database reports/Unused templates. {{Infobox artist Carla Pricop}} is only transcluded on Draft:Carla Pricop so it would have been reported if the draft was deleted first. PrimeHunter (talk) 23:21, 21 September 2021 (UTC)

This discussion about unused templates may be of interest. – Jonesey95 (talk) 02:18, 22 September 2021 (UTC)

Math puzzle: figure out the equation for cropping image collages

Lately I've been working on a bunch of image collages that use {{Multiple image}}. However, one of the trickiest parts is cropping so that images align with each other, as at rows 2/3 of Beijing where the vertical line goes straight down. I've been using a trial and error approach where I make a crop file on Commons and then keep overwriting it until I find the right proportion to get the lines to align, but this takes a lot of time, clutters my edit history, uses up more server space, and doesn't always reach a perfectly aligned result. Would there be a system of equations where I could plug in the aspect ratios of all but one of the images I'd like to use and it'll give me the ratio for the final one? {{u|Sdkb}} ^talk 08:28, 22 September 2021 (UTC)

@Sdkb: If I understand the problem correctly it should be as simple as summing the vertical and horizontal pixels and finding out what the effective size of the gap is to use as your aspect ratio. Taking an example where we have just 3 images in a 2x2 grid and we're trying to find the aspect ratio of the 4th:

5x4	3x9
2x10	?

• Sum horizontal pixels: 5+3 = 8. Then subtract the horizontal pixels from the row you're placing the image in (2): 8-2 = 6
• Sum vertical pixels: 4+10 = 14. Then subtract the vertical pixels from the row you're placing the image in (9): 14-9 = 5
• The aspect ratio of the final image should be 6:5.
• I guess in one equation this would be (sum of horizontal pixels from other rows - sum of horizontal pixels already used in this row):(sum of vertical pixels from other rows - sum of vertical pixels already used in this row)

Is that what you were after? Sam Walton (talk) 09:10, 22 September 2021 (UTC)

It gets a little trickier since the images are all scaled in relation to each other (and there's a—I think—4px buffer between them, which comes into play especially if there's 3 images on a row rather than two). But I think that basic approach ought to work; I'll continue working on it. {{u|Sdkb}} ^talk 22:20, 22 September 2021 (UTC)

Xena: Warrior Princess (comics)

Resolved

 – Error in template was fixed. — xaosflux ^Talk 23:52, 22 September 2021 (UTC)

A couple of formatting problems on Xena: Warrior Princess (comics) that I haven't been able to solve.

1. At the very top of the article, ahead of the introduction there's a stray "Xena: Warrior Princess (comics) ]]", but I can't find where it's coming from; it does not appear to be getting transcluded by the navbox template, and I can't find that string as a broken category declaration in the article text. So my best guess would be that it appears to be getting artificially generated by the infobox, but I can't see where or how.
2. At the very bottom of the article, there's a redlinked "Greco-Roman mythology in comics" category, pursuant to a recent renaming of that category to Category:Classical mythology in comics — but again, I can't find where that category is coming from in order to fix it either. Again, most likely being artificially generated by the infobox, but again I can't solve where.

Can somebody solve this, please? Bearcat (talk) 18:09, 22 September 2021 (UTC)

The part at the top is from |Adaptation=TV. MB 18:47, 22 September 2021 (UTC)

All of this is certainly coming from the monster Template:Infobox comics meta series. Will have to come back to try to unravel all of that. — xaosflux ^Talk 18:48, 22 September 2021 (UTC)

 Fixed both (editing Template:Comics infobox sec/genrecat) * Pppery * _{it has begun...} 19:07, 22 September 2021 (UTC)

Thanks! Bearcat (talk) 20:16, 22 September 2021 (UTC)

New archive provider: ghostarchive.org

https://ghostarchive.org is a new web archive provider, similar to Conifer ie. Webrecorder.io without the login or usage cap requirements. It uses the Open Source Webrecorder technology (which is state of the art IMO). It would be suitable for YouTube and other video formats for example, although there are size caps on video file saves. It's great for advanced JavaScript pages, such as scrolling content, social media sites. Anything that has trouble saving at Wayback or archive.today might do better here. There are a couple usage steps to be aware of, documented at Help:Archiving_a_source#Ghostarchive.org. -- GreenC 19:49, 22 September 2021 (UTC)

Excessive whitespace

I keep seeing deep whitespace at the top of Talk pages where templates have been added to new user (example: User talk:Chrashley). Is this required and 'normal'? On one such I manually reduced it, so I want to clarify if I'm in the wrong here (example). Thank you.--Rocknrollmancer (talk) 20:55, 22 September 2021 (UTC)

@Rocknrollmancer: a quick tag search for NewPages by RedWarn suggests this is at least one of the culprits. I've reported this as a bug at Wikipedia talk:RedWarn. — xaosflux ^Talk 23:49, 22 September 2021 (UTC)

Follow up at Wikipedia_talk:RedWarn#Whitespace_Bug as needed. — xaosflux ^Talk 23:52, 22 September 2021 (UTC)

@Xaosflux: - I've had a quick skim through and can see it's a Beta. It makes sense that - being a minimal click utility launching a pop-up window - the users would not be looking at the end result. I have a cat who likes to walk on the keyboard so if I leave an editing pane open and walk away I have to check what changes have been made! rgds,--Rocknrollmancer (talk) 08:43, 23 September 2021 (UTC)

@Rocknrollmancer: it looks like they fixed it, here is a page created before the fix: User talk:148.77.89.202, and one after: User talk:27.110.174.204. — xaosflux ^Talk 11:00, 23 September 2021 (UTC)

Now, other scripts and tools could also be introducing similar issues - for example on User talk:Chrashley - the whitespace was introduced in this edit Special:PermaLink/1045846338 by User:ClueBot NG; when you see these being made by someone or some tool - following up with that maintainer is the next step (in that case: User talk:ClueBot Commons). — xaosflux ^Talk 11:04, 23 September 2021 (UTC)

Thanks @Xaosflux: - I've seen another but that was redwarn - I'll be more aware for any others.--Rocknrollmancer (talk) 21:10, 23 September 2021 (UTC)

Includeonly with category

Hello! I'm in the process of deprecating parameters at Template:Infobox Playboy Playmate and have set up the following:

<includeonly>{{#if:{{{bust|}}}{{{waist|}}}{{{hips|}}}|[[Category:Pages using infobox Playboy Playmate with deprecated parameters]]}}</includeonly>

However, the template is still categorised into the tracking category. What's going on? Did I do something wrong, or is this a bug? Tol (talk | contribs) @ 00:33, 23 September 2021 (UTC)

Removing the parameters from the template documentation removed the template and the documentation from the category. Also added {{Category handler}} to remove it from unwanted namespaces. Nardog (talk) 00:50, 23 September 2021 (UTC)

@Nardog: Ahh! Thank you so much for catching that. Tol (talk | contribs) @ 01:32, 23 September 2021 (UTC)

The usual way to do this is to wrap the if statement in {{main other}}. – Jonesey95 (talk) 03:36, 23 September 2021 (UTC)

Strange search results (gold (III) cloride)

When I search for "gold (III) cloride" I get very odd results. It fails to find Gold(III) chloride (which would correct my single-letter typo), suggests the much more hamming-distant "Did you mean: gould iii clotilde", and says "There were no results matching the query." On the other hand, it gives me "Results from sister projects" which include wikisource:1911 Encyclopædia Britannica/Photography. Searching for "gold(III) cloride" (i.e. without the space after "gold"), gives equally perplexing results. Can anybody explain what's going on here. -- RoySmith (talk) 13:48, 23 September 2021 (UTC)

FWIW, when I type "gold (III" or "gold (III) cloride" in the search box, I am shown helpful suggestions, including the desired result. It is only after clicking the search button that things go wrong. – Jonesey95 (talk) 14:43, 23 September 2021 (UTC)

The unspaced gold(III) is a rare construction outside chemistry. It doesn't surprise me if search isn't coded to discover it when you both have spaced gold (III) and another mismatch. Autocompletion and search use different algorithms so different results should sometimes be expected. PrimeHunter (talk) 16:33, 23 September 2021 (UTC)

Suggests to me that we should have redirects for these various chemical names for the "incorrected" spaced names to simply help searching. This may even be a bot-able task, based on categories and titles. (list of metals followed immediately by an oxidation state). --Masem (t) 16:48, 23 September 2021 (UTC)

@EBernhardson (WMF), does your team know about this? Whatamidoing (WMF) (talk) 17:31, 24 September 2021 (UTC)

List of most frequently used URLs in sources?

I recall at one point seeing a list of the most frequently used URLs in citations on Wikipedia. I'd like to reference it, but I can't find it now. Does anyone here know where to find (or how to create) a list of the most frequently used websites in external links within reference tags? {{u|Sdkb}} ^talk 20:04, 23 September 2021 (UTC)

@Sdkb: Wikipedia:WikiProject Fact and Reference Check/Most cited domains has a top-15 from 2015 and links phab:P587 with a long list. PrimeHunter (talk) 23:37, 23 September 2021 (UTC)

142429 references to IMDb? Yikes... Elli (talk | contribs) 23:40, 23 September 2021 (UTC)

I think it includes external links sections. {{IMDb title}} is currently transcluded in 173,406 articles, and {{IMDb name}} in 139,593. PrimeHunter (talk) 23:51, 23 September 2021 (UTC)

Where's the Search Button?

Resolved

 – Fixed in T291272

Tracked in Phabricator
Task T291272
Resolved

As soon as I type something into the search bar, the [Search] button vanishes and is replaced exclusively by the [Go] button. What a huge impediment. -- Veggies (talk) 23:09, 23 September 2021 (UTC)

@Veggies: this is currently broken, see Wikipedia:Village_pump_(technical)/Archive_192#Search_box and phab:T291272. — xaosflux ^Talk 23:16, 23 September 2021 (UTC)

You can start a search with tilde ~ or click "Search for pages containing" below the dropdown search suggestions. PrimeHunter (talk) 23:39, 23 September 2021 (UTC)

@Veggies: it should be fixed now, please let me know if you're still running into issues. Legoktm (talk) 01:24, 24 September 2021 (UTC)

Extremely long timestamps in Special:NewPagesFeed

Tracked in Phabricator
Task T291675
Resolved

In Special:NewPagesFeed, I get now extremely, irritatingly long and repetitive timestamps (I don't remember seeing this before). On the left side, I get "Created by Kailash29792 (talk | contribs) · 179016 edits since Fri Apr 22 2011 09:51:21 GMT+0200 (Midden-Europese zomertijd)" or "Created by Djflem (talk | contribs) · 92486 edits since Fri Dec 08 2006 10:22:01 GMT+0100 (Midden-Europese standaardtijd)", while at the right side I get "Fri Sep 24 2021 10:55:36 GMT+0200 (Midden-Europese zomertijd)" and "Fri Sep 24 2021 10:54:10 GMT+0200 (Midden-Europese zomertijd)". I use Legacy Vector, time formatting "10:46, 24 September 2021", and no timezone, and language "en - English" (my Google account as well has English as language). Is this some new unwanted feature in the feed, or is this somehow Chrome interfering with how these are displayed? I don't have this issue on e.g. the watchlist or a page history. Fram (talk) 10:52, 24 September 2021 (UTC)

I'm using Firefox and I get Created by Paruart (talk | contribs) · 10 edits since Thu Sep 09 2021 20:12:15 GMT+0100 (British Summer Time). I'm using the legacy vector skin too. Dreamy Jazz ^{talk to me | my contributions} 10:59, 24 September 2021 (UTC)

Disabling legacy vector does not change the behavior for me. Dreamy Jazz ^{talk to me | my contributions} 11:01, 24 September 2021 (UTC)

This change does not seem to be because of a skin change. Dreamy Jazz ^{talk to me | my contributions} 11:02, 24 September 2021 (UTC)

It's not also caused by any of my userscripts. I tested this by enabling safemode. Dreamy Jazz ^{talk to me | my contributions} 11:04, 24 September 2021 (UTC)

Thanks. So probably some Thursday WMF magic? Fram (talk) 11:06, 24 September 2021 (UTC)

• WP:ITSTHURSDAY - I've opened phab:T291685. — xaosflux ^Talk 11:09, 24 September 2021 (UTC)
  • Which is now phab:T291675. — xaosflux ^Talk 11:23, 24 September 2021 (UTC)
    • Thanks. Fram (talk) 11:57, 24 September 2021 (UTC)

This should be fixed now. Legoktm (talk) 19:25, 24 September 2021 (UTC)

Paalika Keka

Tracked in Phabricator
Task T291717

When looking at the View history tab, the words for "newest" and "oldest" have been replaced by "Paalika" and "Keka". I think this is very recent. Any idea what the cause might be? CMD (talk) 13:16, 24 September 2021 (UTC)

In you preferences is you language set to EN-GB? There have been some issues with the wrong date being uploaded to the translation tables and what you are seeing looks like it may an example of this. Try changing language to just EN. Nthep (talk) 13:31, 24 September 2021 (UTC)

Yup, MediaWiki:Histfirst and MediaWiki:Histlast are wrong for en-gb. Really, no one should use en-gb, it is a monstrosity. Looks like these were screwed up at translatewiki - they were recently reverted but you will have to wait for the next translation sync. — xaosflux ^Talk 13:37, 24 September 2021 (UTC)

I see, thanks. What an odd mistake, and google translate won't give me anything for Keka. I have no recollection of setting a language, although if no-one should use it I don't understand why it's there. CMD (talk) 14:19, 24 September 2021 (UTC)

I think that the language concerned may be Frafra (Gurenɛ), something we don't have either as a selectable language at Preferences or as a Frafra-language Wikipedia - it would be gur:. --Redrose64 🌹 (talk) 15:22, 24 September 2021 (UTC)

• I put temporary local overrides in for these for now. — xaosflux ^Talk 17:00, 24 September 2021 (UTC)

translatewiki:Special:Contributions/Akobire Amiyine shows many other en-gb errors from 14 September, e.g. "Binŋɛ" on the preferences save button.[7] PrimeHunter (talk) 20:15, 24 September 2021 (UTC)

@PrimeHunter: yup - one of the reason users using English variants in the interface here is a bad idea and we warn about it on the pref page! — xaosflux ^Talk 20:18, 24 September 2021 (UTC)

It's a tame warning in MediaWiki:Preferences-summary/en-gb:

Your language setting of "British English" means that you may miss some local customisations.

I originally made a stronger "not recommended" warning but it was opposed. PrimeHunter (talk) 20:29, 24 September 2021 (UTC)

• Wikipedia:Village_pump_(proposals)#Discourage_en-xx_UI_variants proposed. — xaosflux ^Talk 20:43, 24 September 2021 (UTC)
• I filed phab:T291717, deploying changes to i18n messages is a little more involved so it'll probably need to wait until Monday. Legoktm (talk) 01:35, 25 September 2021 (UTC)

Is <math> broken?

Getting broken rendering at Equivalent air depth#Derivation of the formulas. Cheers, · · · Peter Southwood ^(talk): 13:23, 24 September 2021 (UTC)

It looks OK to me. Please be more specific. What is your Math setting at the bottom of Special:Preferences#mw-prefsection-rendering? PrimeHunter (talk) 13:46, 24 September 2021 (UTC)

MathML with SVG or PNG fallback (recommended for modern browsers and accessibility tools) · · · Peter Southwood ^(talk): 14:15, 24 September 2021 (UTC)

Works fine on tablet with Chrome. Problem is on desktop with Firefox. Problem is probably Firefox, as it also gives problems with upload wizard on Commons. Says browser is incompatible. · · · Peter Southwood ^(talk): 14:35, 24 September 2021 (UTC)

You still haven't said what "broken rendering" means but it looks fine to me in Firefox 92.0 with the same Math setting. Which Firefox version do you have? PrimeHunter (talk) 15:49, 24 September 2021 (UTC)

Looks fine to me, unless I'm missing something subtle. Firefox 91 desktop for Mac OS, "MathML with SVG or PNG fallback" selected. – Jonesey95 (talk) 00:35, 25 September 2021 (UTC)

Problem with mapframe with 7 lines from OSM via Wikidata

In the page Draft:High-voltage transmission line Kassø-Tjele I have tried to make a mapframe fetching 7 lines from OSM via Wikidata.

It is 5 OSM ways which should be displayed as orange lines:

• d:Q63184980: https://www.openstreetmap.org/way/220924125
• d:Q63208922: https://www.openstreetmap.org/way/287543094
• d:Q63209004: https://www.openstreetmap.org/way/287543100
• d:Q63208969: https://www.openstreetmap.org/way/287543103
• d:Q63208445: https://www.openstreetmap.org/way/319088854

and 2 OSM relations of type route which should be displayed as black lines:

• d:Q62924256: https://www.openstreetmap.org/relation/5207262
• d:Q62922138: https://www.openstreetmap.org/relation/6698731

The 5 OSM ways are displayed in the mapframe in preview mode and when you maximize the mapframe, but not when the mapframe isn't maximized. The 2 OSM route relations are not displayed at all in the mapframe.

I have tried both defining the mapframe tag code directly in revision https://en.wikipedia.org/enwiki/w/index.php?title=Draft:High-voltage_transmission_line_Kass%C3%B8-Tjele&oldid=1046272983 and by using the {{maplink}} template in revision https://en.wikipedia.org/enwiki/w/index.php?title=Draft:High-voltage_transmission_line_Kass%C3%B8-Tjele&oldid=1046279152 with no difference in the result. I am not sure if this is a bug in the Kartographer extension or if I am doing something wrong. Any insight into this will be appreciated. Thank you. --Dipsacus fullonum (talk) 21:49, 24 September 2021 (UTC)

MediaWiki message delivery to Wikipedia:Tech news no longer produces timestamped section ids

Template:Latest tech news has been producing empty output for a while. The template uses Labeled section transclusion, which relies on timestamped section IDs on the page Wikipedia:Tech news. However, the page no longer has timestamped section IDs.

The format of posts by User:MediaWiki message delivery to Wikipedia:Tech news has changed between weeks 12 and 13 of 2021. Wikipedia:Tech news/Archive 8#Tech News: 2021-12 has the timestamped technews-2021-W12 but Wikipedia:Tech news/Archive 8#Tech News: 2021-12 has tech-newsletter-content. However, the corresponding pages on Meta don't seem to correspond to that. Both meta:Tech/News/2021/12 and meta:Tech/News/2021/13 have the non-timestamped section ID tech-newsletter-content Even in meta:Tech/News/2020/01 from January of 2020 the section ID is not timestamped. So I assume that something changed with the delivery of these posts to Wikipedia:Tech news.

Pinging User:Johan (WMF) and User:Quiddity (WMF) – editors of recent Tech News on Meta, User:Guillom and User:wbm1058 who implemented the Template:Latest tech news, User:Quiddity and User:Evad37 who edited the preamble of Wikipedia:Tech news. —⁠andrybak (talk) 11:40, 25 September 2021 (UTC)

⁠andrybak: Ah, yes, the MassMessage delivery system was changed earlier this year, mainly to allow the use of translated fallback languages. This should be the culprit. /Johan (WMF) (talk) 11:44, 25 September 2021 (UTC)