Page MenuHomePhabricator

Use site-specific tags for cross-wiki-upload
Closed, DeclinedPublic

Description

I want wiki-spcific tags from which wiki the images was uploaded from using Cross-wiki-upload, as with "cross-wiki-upload-1", but for e.g. -svwiki, -enwiki, -eswiki and -thwiki (reason for thwiki), etc.

22:57:03 <Josve05a> Can we create new "Tags" for Crosswiki-uploads, based on sourcewikis? We did have -1, -2, -3 and -4 based on the different buckets, but if would be nice to have -svwiki, -enwiki etc as well
22:57:29 <Josve05a> Yeah, but then I could perhaps ask someone else if you're not here right now
23:01:17 <Josve05a> do you understand what I'm asking for/want?
23:06:06 <MarkTraceur> Josve05a: I get the idea, but that's a *lot* of tags, and we have to make them manually for each wiki
23:06:39 <MarkTraceur> Josve05a: MatmaRex knows this better, but maybe we could help y'all with log filtering based on the comment field?
23:06:55 <MarkTraceur> Then you could search for ones with the comment "Cross-wiki upload from sv.wikipedia.org" etc.
23:06:57 <Josve05a> hmm....howabout prhpas the biggest 5 wikis get each, and the smalles 50 has one together...or something..
23:07:16 <Josve05a> "Cross-wiki upload from sv.wikipedia.org" etc. << that would be nice
23:07:29 <MarkTraceur> OK let me ask someone who knows more about log filtering than I do, sec
23:07:34 <Josve05a> :)
23:25:16 <Josve05a> I'm guessing, not possible...
23:45:41 <MarkTraceur> Josve05a: OK, RoanKattouw seems to think doing the per-wiki tags is maybe best

Event Timeline

Assigning MatmaRex so he can look at this when he gets back.

Is it in scope (or technically possible) for the tags to also indicate the original file page on the original wiki?

Sorry for late reply.

I want wiki-spcific tags from which wiki the images was uploaded from using Cross-wiki-upload, as with "cross-wiki-upload-1", but for e.g. -svwiki, -enwiki, -eswiki and -thwiki (reason for thwiki), etc.

This would indeed be technically possible, but I'm not convinced it's a good idea.

  • I don't want tags to turn into a generic metadata storage system, we already have a few. Honestly, I would have preferred if they had stayed limited to AbuseFilter tags. But I guess this ship sailed when VisualEditor and then MobileFrontend started using them to mark the tool used to make the action, so it seemed appropriate to mark the cross-wiki upload tool actions with them too. But this is where I think we should draw the line for now.
  • We'd need to somehow sync the list of active tags with the list of active wikis, every time a wiki is created or closed. We'd need to do some development to display the wiki name (or at least the domain), rather than the database name (which is not very user-friendly and can be misleading when it differs from domain name – be_x_oldwiki is a good example of both).
  • All of the hundreds possible tags would be listed on https://commons.wikimedia.org/wiki/Special:Tags (and anywhere else where we'd want to display the list of available tags).
  • I don't think a use case for this comes up often enough for existing tools not to be sufficient.

So, to summarize, I think this could leave a future maintainer with technical debt to resolve, and it would also require some development work from us right now.

Instead, I'd try:

  • Setting up an AbuseFilter rule if you have junk uploads coming in from specific wikis. (This was mentioned in the Commons discussion you linked to, but I couldn't find any filter for this, perhaps it's private.) I haven't tested it now and my filter skills are rusty, but you should just need a filter like this: action == "upload" & summary == "Cross-wiki upload from xx.wikipedia.org".
  • Doing Quarry queries if you just want to review uploads from a specific wiki. For example: http://quarry.wmflabs.org/query/8142 displays the latest 100 cross-wiki uploads from sv.wp.

I'll leave this task open (perhaps somebody has stronger arguments than mine, or perhaps something changes in the future), but I don't think we're going to be working on this.

Is it in scope (or technically possible) for the tags to also indicate the original file page on the original wiki?

This is not possible. Ignoring any other practical issues with arbitrary tags, tag length is limited to 255 bytes, same as page title length – and we'd also need to record the namespace, source wiki and add some prefix to the tag name.

matmarex triaged this task as Lowest priority.
matmarex removed a project: JavaScript.

Is it in scope (or technically possible) for the tags to also indicate the original file page on the original wiki?

This is not possible. Ignoring any other practical issues with arbitrary tags, tag length is limited to 255 bytes, same as page title length – and we'd also need to record the namespace, source wiki and add some prefix to the tag name.

Well, all pages have a page-ID, which is most likely less than that, and then normal math (e.g. 2¹¹ for 2048) could compress the number even further...but...out of scope for this task.

Right, I should have said not feasible, very few things are actually impossible. Although note that there isn't really a page ID if the page gets deleted ;) – or if it is never saved.

Jdforrester-WMF subscribed.

I agree with @matmarex. There isn't a problem statement, and this is a bad solution in search of a problem when we've got actual issues to support.

I'll note that with T139848 resolved, you can create an AbuseFilter for this. It would match action='upload' and the summary specific to the site/sites you want to review uploads from. You can set it to tag them and use the tag, or to just do nothing and use Special:AbuseLog to watch incoming uploads.