Page MenuHomePhabricator

Global users should be Wikibase items on Meta
Open, Stalled, Needs TriagePublic

Description

Problem
There is no way to query existing publicly exposed data on users themselves.

A lot of users use the User_Info template and include data items (like IRC, image, full name, languages spoken, etc.). This data is already included on their User profile page and ought to be available through wikibase.

For instance Q23034479 has a a property P553 which has a value "Wikimedia project" and a scalar value of User:Katherine (WMF).

It would be helpful (I think) if publicly exposed properties (that are already on the User page) could be added to a item of User:Katherine (WMF).

Use-Cases

  • Query all of the users who speak at least level 3 Spanish.
  • Query all of the users who have a picture.
  • Query all of the users who have an IRC nick

Solution
Users should automatically become wikibase items in their own U namespace.

To do this we will need to enable wikibase on Meta and create a U namespace that can be distinguished from the existing main Q namespace on Wikidata. The Id of the user should be the same as the CentralAuth id.

For instance, my central auth id is 50584396 (as reported by the API), so my wikibase ID on Meta would be U50584396

This would also allow me to reference the property on my user page (just like referencing a wikidata property on an article).

Event Timeline

dbarratt created this task.Aug 11 2017, 8:56 PM
Restricted Application added subscribers: PokestarFan, Aklapper. · View Herald TranscriptAug 11 2017, 8:56 PM

Added three supporters and three opponents of a related property proposal. (I wasn't able to find Innocent bystander on Phabricator.)

This is orthogonal to the property proposal cited (which I submitted).

It seems that this is a proposal to extend, and indeed circumvent, the reach of the Wikidata notability policy; Phabricator is not the place to do this.

It seems that this is a proposal to extend, and indeed circumvent, the reach of the Wikidata notability policy; Phabricator is not the place to do this.

Does having a User page circumvent the Wikipedia notability policy?

Maybe something should distinguish between standard items and user items?

I expected to go to User:DBarratt_(WMF) and see a wikidata object rather than a user page (or maybe the data would be another tab or something).

It might be a good idea to change the prefix of User objects from Q to U or something like that so it is clear that it is not a standard item but a User item.

Izno added a subscriber: Izno.Aug 13 2017, 1:29 PM

We don't need items for this functionality--Wikidata should just do it. I can't find a task for it, though I'm pretty sure there already exists at least one. The work done for Wiktionary is probably suitable.

We don't need items for this functionality--Wikidata should just do it.

Agreed. Ideally the User ID would be the same as the global id (as in, centralauth id) rather than the local user id.

I totally disagree with this proposal. How long before these items would be used to doxx the real identity of users, and/or their religion/sexual orientation?

Hsarrazin added a comment.EditedAug 13 2017, 1:57 PM

Totally agree with Ash_Crow here... this is NOT a good idea :((

I disagree with a property creation, THIS would be even worse. Especially if the creation of item for each user was systematic and could not be refused by each user...

I totally disagree with this proposal. How long before these items would be used to doxx the real identity of users, and/or their religion/sexual orientation?

This is why I suggested

In T173145, @dbarratt wrote:

I'm not sure what the permissions would be like, perhaps only the user can edit their own item?

This seems consistent with User pages which can only be edited by the user. This would allow users to provide (or not provide) any data they would like.

While this data could be used to harass/doxx users, it could also be used to quickly mitigate harassment (especially with calculated fields).

It would also allow to search the database of users for any property.

Even with controlled data, this would be a serious breach in the privacy of users...

So this would be a totally new namespace? Like [[U123456789]] instead of [[Q123456789]]?

Envlh added a subscriber: Envlh.Aug 13 2017, 2:56 PM

I totally agree with Ash_Crow. Wikidata aims to be the sum of all knowledge which can be referenced, which Wikimedians data is not. A U namespace could be a solution, but only and only if only a Wikimedian can edit its own item.

It would also allow to search the database of users for any property.
Even with controlled data, this would be a serious breach in the privacy of users...

I don't understand how that would be a breach of privacy. You would be the only one that could edit your own item and any computed field (if that's something we even do) is already available in the Public API anyways, so this isn't exposing any additional information to anyone.

So this would be a totally new namespace? Like [[U123456789]] instead of [[Q123456789]]?

I mean I think that would be best.
For instance, my Global user info can be found here:
https://www.wikidata.org/w/api.php?action=query&format=json&meta=globaluserinfo&guiuser=DBarratt%20(WMF)
This reports that my global user id is 50584396. So I think my id on wikidata would ideally be U50584396. The separate namespace keeps user items separate from all other items (just like how user pages are separate from Article pages on Wikipedia)

A U namespace could be a solution, but only and only if only a Wikimedian can edit its own item.

Completely agree, that makes total sense to me and is consistent with user pages.

Oh I see, User pages can be edited by anyone, but I agree, User items in wikidata should only be editable by the user themselves, I don't see any reason why someone else would need to edit that.

Honestly, I don't know why User pages can be edited by anyone, that doesn't seem like a good idea.

Izno added a comment.Aug 13 2017, 5:02 PM

Found the task: T168792: Use Cognate to link user pages.

Honestly, I don't know why User pages can be edited by anyone, that doesn't seem like a good idea.

Wiki ethos. It's definitely a good idea, not least to help people who don't understand what they're doing.

You must understand that, to French people, the mere possibility to publicly cross-search personal data for people that are not "public" (i.e. famous... in a way or another) is very, very sensitive....

Sensitive to the point that the law obliges to anonymize jurisprudence databases...

which is why I'm so reluctant to the mere idea of a public database that can be publicly querieds with users data. For now, users global data are accessible, but not publicly searchable. It is not at all the same...

Found the task: T168792: Use Cognate to link user pages.

Honestly, I don't know why User pages can be edited by anyone, that doesn't seem like a good idea.

Wiki ethos. It's definitely a good idea, not least to help people who don't understand what they're doing.

I mean I understand the reasoning, but I can also understand that being a problem too.

You must understand that, to French people, the mere possibility to publicly cross-search personal data for people that are not "public" (i.e. famous... in a way or another) is very, very sensitive....
Sensitive to the point that the law obliges to anonymize jurisprudence databases...
which is why I'm so reluctant to the mere idea of a public database that can be publicly querieds with users data. For now, users global data are accessible, but not publicly searchable. It is not at all the same...

I understand the concern. Although, our publicly accessible data is absoltuley publically searchable already, as an example:
https://meta.wikimedia.org/w/index.php?search=Contact+me&title=Special:Search&profile=advanced&fulltext=1&ns2=1&ns12=1&searchToken=344piw8ul9b456bbe0rx3t3bz
or on a third party:
https://www.google.com/search?q=site:meta.wikimedia.org+%22Contact+Me%22+%22User:%22&ei=yJ6QWfXDJqnm0gLBg4OQBA&start=0&sa=N&biw=1164&bih=581

While it is much more difficult to parse, the data is available (albiet only the data that someone freely posts, which would be the same if it was in wikidata).

dbarratt updated the task description. (Show Details)Aug 13 2017, 6:55 PM
Esc3300 added a comment.EditedAug 14 2017, 7:17 AM

if entities are made, I don't think they necessarily need to be on www.wikidata.org .. Wikibase could easily run on Meta, as it does on Commons.

Lydia_Pintscher changed the task status from Open to Stalled.Aug 14 2017, 3:02 PM
Lydia_Pintscher moved this task from incoming to hold on the Wikidata board.
Lydia_Pintscher added a subscriber: Lydia_Pintscher.

At this point the drawbacks outweigh the benefits of doing this. The points raised in this ticket will be raised 100 times more in the larger community.

if entities are made, I don't think they necessarily need to be on www.wikidata.org .. Wikibase could easily run on Meta, as it does on Commons.

Fine with me, as long as you can reference a user item from wikidata, it doesn't really matter.

The points raised in this ticket will be raised 100 times more in the larger community.

I do not understand the validity of any of these arguments.

I do not understand the validity of any of these arguments.

Then perhaps you should interact with the Wikidata community more. Might I suggest dropping a line here? (For the record, I support creating the property that I linked to earlier but oppose the proposal which is the subject of this ticket.)

Izno added a comment.EditedAug 14 2017, 5:48 PM

The points raised in this ticket will be raised 100 times more in the larger community.

I do not understand the validity of any of these arguments.

  1. I think there is general opposition to having users manage items.
    1. I specifically oppose any item management. CentralAuth and T168792: Use Cognate to link user pages. Everything else can already be queried--you said it yourself.
    2. One might reasonably make meta.wikimedia.org a federated Wikibase installation a la structured data for Commons, so that users can define global attributes, potentially using data from Wikidata.
    3. Either way, I think the Wikidata community wants nothing to do with users managing data there.
    4. Fine with me, as long as you can reference a user item from wikidata, it doesn't really matter. is a really big caveat and probably will not be implemented, not least for the opposition in this task.
  2. There is specific opposition to having user information, which is seen as unimportant, mixed in to the Q main space items. You can review http://www.wikidata.org/wiki/WD:N for specifics.
    1. Also adds additional editorial burden for almost 0 gain for the Wikidata community.
  3. There is specific opposition to enabling users to out themselves.
    1. Adds additional administrative burden for almost 0 gain for the Wikidata community.

If you really think this is a good idea, you need to have a proposal ready to overcome all of those objections. Plus others that the community decides are worth objecting over. (Good luck.)

I do not understand the validity of any of these arguments.

Then perhaps you should interact with the Wikidata community more. Might I suggest dropping a line here? (For the record, I support creating the property that I linked to earlier but oppose the proposal which is the subject of this ticket.)

Are you implying that, to understand any of the arguments posted here, I must first acquire a 'secret knowledge' that can only be obtained by further participation in the community? Would you be willing to explain the context to me so I am no longer ignorant and am able to understand these arguments?

The points raised in this ticket will be raised 100 times more in the larger community.

I do not understand the validity of any of these arguments.

  1. I think there is general opposition to having users manage items.
    1. I specifically oppose any item management. CentralAuth and T168792: Use Cognate to link user pages. Everything else can already be queried--you said it yourself.

I see, so if T168792 is completed then the existing data in the API would be queryable (if I'm understanding this correctly)? I suppose that's a fine compromise. As much as I'd like the publicly exposed details to be available as well (i.e. not in Wikitext) I understand the objections (though, it is already in the wikitext, which is my point).

  1. One might reasonably make meta.wikimedia.org a federated Wikibase installation a la structured data for Commons, so that users can define global attributes, potentially using data from Wikidata.

I think that would be fantastic.

  1. Either way, I think the Wikidata community wants nothing to do with users managing data there.

Understandable.

    1. Fine with me, as long as you can reference a user item from wikidata, it doesn't really matter. is a really big caveat and probably will not be implemented, not least for the opposition in this task.
  1. There is specific opposition to having user information, which is seen as unimportant, mixed in to the Q main space items. You can review http://www.wikidata.org/wiki/WD:N for specifics.

This is why I suggested a different namespace (which is why we have a different namespace on Wiki's for User:). But also having them on meta makes the distinction even more clear.

    1. Also adds additional editorial burden for almost 0 gain for the Wikidata community.
  1. There is specific opposition to enabling users to out themselves.

I mean they are already outing themselves, it's just in a wikitext. This would give the data they are already publicly disclosing a structured format for research and analysis of different systems.

  1. Adds additional administrative burden for almost 0 gain for the Wikidata community.

Does having "User" on Meta increase the editorial burden? This seems like hosting on Meta might be the solution.

If you really think this is a good idea, you need to have a proposal ready to overcome all of those objections. Plus others that the community decides are worth objecting over. (Good luck.)

Well that's why I'm talking to all of you. I think everyone has made some great points.

Does it resolve most of the concerns if Users are in a U namespace and are hosted on Meta (and somehow, can be referenced from Wikidata as in the example)?

Are you implying that, to understand any of the arguments posted here, I must first acquire a 'secret knowledge' that can only be obtained by further participation in the community?

This proposal to expand Wikidata's scope was put forward without forewarning of the community in any other public forum—it is not guaranteed that frequent Wikidata contributors will all be members of the Wikidata project on Phabricator. I have added subscribers based on a related issue, who have continued and will continue to enlighten you on exposing structured user information, but trips through RFCs, project chat archives, and administrators' noticeboard archives will most likely show that this topic was fought over at length between occasional users and longtime contributors—it is not codified knowledge, but it is most certainly not secret and I am not trying to imply that it was. If you have in fact combed through these things (and discovered that the topic of user items has never been broached in the site's five years of existence) or if you have actually informed the community somewhere else other than Phabricator, then I sincerely apologize for misstating the facts.

Would you be willing to explain the context to me so I am no longer ignorant and am able to understand these arguments?

I will save a more thorough response for later in the week. Bear in mind, though, that I don't frequently take larger roles in the discussions linked to above and thus my own views are not nearly as developed and may not nearly be as representative of the overall community as others' might be. It is to your benefit to find sanction (or compel it) for this proposal among the community in venues other than Phabricator first as this is more than just a technical issue to many people.

I do not understand the validity of any of these arguments.

Then perhaps you should interact with the Wikidata community more.

Also, per WMF Policy, I use my personal account for contributions to Wikidata. Regardless, I do not believe that the validity of an argument is conditional on one's existing contributions, but ought to stand and on it's own validity.

Esc3300 removed a subscriber: Esc3300.Aug 16 2017, 6:03 PM

Just a note: Only allow the users themselves to edit items about them is impractical, as properties and items is managed by community, and sometimes properties get deprecated or items get merged. Probably we need a new user group to manage them.

dbarratt renamed this task from Global users should be wikidata items to Global users should be Wikibase items.Aug 18 2017, 2:18 PM
dbarratt updated the task description. (Show Details)

Just a note: Only allow the users themselves to edit items about them is impractical, as properties and items is managed by community, and sometimes properties get deprecated or items get merged. Probably we need a new user group to manage them.

There's a valid argument for allowing anyone to edit (since it is a wiki with the same details from the User page and that would be the same permissions the User page) or "admins" and the user themselves (which would prevent others from editing public details about yourself).

Now that I know User pages are editable by effectively anyone, I do not have a strong opinion either way.

dbarratt renamed this task from Global users should be Wikibase items to Global users on Meta should be Wikibase items.Aug 18 2017, 2:21 PM
dbarratt added a project: wikiba.se website.

This is unrelated to wikiba.se (the website that documents Wikibase). Instead we need a Meta-Wiki project.

This is unrelated to wikiba.se (the website that documents Wikibase). Instead we need a Meta-Wiki project.

Whoops... sorry 'bout that. Yeah I noticed there is no Meta-Wiki project. :/

Bugreporter renamed this task from Global users on Meta should be Wikibase items to Global users should be Wikibase items on Meta.Aug 18 2017, 6:54 PM
Legoktm added a subscriber: Legoktm.

-#Wikimedia-extension-setup as there's nothing here that needs setup yet.
-#mediawiki-extensions-centralauth as this has nothing do with CentralAuth AIUI.

Anyways, my two cents on this is that this seems like a solution in search of a problem. If you want to extract info out of {{user info}} or something, then consider adding some semantic markup that makes it easy to parse or have it generate vcards, etc.

dbarratt added a comment.EditedAug 18 2017, 9:47 PM

Anyways, my two cents on this is that this seems like a solution in search of a problem. If you want to extract info out of {{user info}} or something, then consider adding some semantic markup that makes it easy to parse or have it generate vcards, etc.

It's more of what I expected as a user. I expected to go to my user page and be given a wikidata item. If we are going to give data objects to pages, files (commons), etc. Why would user pages be excluded? (I am not saying they should be on wikidata, I agree that they should be on Meta).

Also, it's not just about semantic markup, it's about making that data queryable and relatable to other data (which is the whole point of a database).

You could also use your argument to say that Wikidata shouldn't exist at all. But here we are. :)

Folks, I think we can let this rest. It will require more product and development resources than I am willing to put into it anytime soon.

Well, I don't get this. Why do we want to flood Wikidata with items about users? Does not make sense to me...

Well, I don't get this. Why do we want to flood Wikidata with items about users? Does not make sense to me...

We already decided this would be on Meta, not Wikidata.

It'd still require decisions and work from the Wikidata team so please keep it on our board.

It'd still require decisions and work from the Wikidata team so please keep it on our board.

Apologies!

Stryn added a subscriber: Stryn.Sep 13 2017, 8:40 PM
dbarratt updated the task description. (Show Details)Jul 11 2018, 4:59 PM
daniel added a subscriber: daniel.Jul 11 2018, 5:29 PM

I don't quite get the rationale behind this. There seems to be an idea that Wikibase is the preferred mechanism for storing structured data in MediaWiki. That's a bit off.

Wikibase is a knowledge modeling tool. It's a good choice if you want to enable users to freely make statements of any kind about any subject, which can be sourced and can contradict. It's a great choice for a machine readable encyclopedia, or for scientific research knowledge bases.

It's NOT a good choice for storing plain facts, and it's not a good choice if tight control over some kind of schema is required, or if some of the information is not user editable. It's MUCH simpler to define a new content model for that, perhaps a JSON based one, instead of trying to lock down Wikibase for such a use case.

Sure, you then won't be able to re-use Wikibase's UI for editing statements. The good thing about this is that then, you won't have to re-use Wikibase's UI for editing statements, which is overly complicated for a use case like this one. From a UX perspective, it's much better to build specialized UI for specialized data.

Making structured data that does not use the Wikibase data model compatible with the Wikidata would of course be quite desirable. To do that, a) re-use Wikidata's (not Wikibase's) vocabulary, and use the DataValue representations and data types defined by Wikibase.

Much of that is already in libraries and can be re-used easily enough. Some of this may have to be pulled out of Wikibase, but that would not be too hard.

So, in summary: unless the idea is to collaboratively model knowledge about people, don't use Wikibase entities. Re-use the vocabulary, re-use data types and value representation, re-use widgets, but on't use full fledged knowledge modeling where it is not needed.

I'm certainly open to alternative solutions. It just felt like, of content models that we currently have (unstructured pages, form fields tied to database fields, wikibase, something else?) it was the best option I could think of.

dbarratt updated the task description. (Show Details)Jul 11 2018, 5:39 PM
dbarratt updated the task description. (Show Details)Jul 11 2018, 5:45 PM
dbarratt updated the task description. (Show Details)

Quiet a few things use JSON based content models to store structured data. Sadly, they tend to not have good UI. So that would need to be written. But that would also need to be written when basing this on Wikibase. Unless you want free form modeling of users.

Unless you want free form modeling of users.

I think that's fine. I'm assuming the properties though, would be unique to U on meta.

Ltrlg added a subscriber: Ltrlg.Jul 29 2018, 3:19 PM

I suppose now that we have MCR this could be rephrased as: "Adding structured data to global user pages" much like StructuredDataOnCommons adds structured data to global files.