Wikipedia talk:Wikidata interwiki RFC

Support or Oppose

 * Support as proposer. (Noting that I am an active admin here and on Wikidata). --Rschen7754 09:26, 17 January 2013 (UTC)
 * Support, exactly like this. --Stryn (talk) 09:51, 17 January 2013 (UTC)
 * Oppose This proposal is very unclear, of no benefit and there is no need to restrict what people do in this area. A very large proportion of people will not be aware of this "policy" and will be in breach.  So in order to cause good relations between enforcers and gnomes we can do without it. If we are going to have such a statment it can be an essay, an opinion of a person. Graeme Bartlett (talk) 10:38, 17 January 2013 (UTC)
 * Well, then I have to ask why we had two entire arbitration cases largely dealing with gnome edits... also, what's not clear about it?--Rschen7754 10:43, 17 January 2013 (UTC)
 * Assuming AN(I) is sensible enough to treat "considerable leeway" with considerable leeway, and not be too enthusiastic to jump on people, then support - this seems a logical enough position. Andrew Gray (talk) 11:05, 17 January 2013 (UTC)
 * "Considerable leeway" should include at leaat two unambiguous warnings. Andy Mabbett ( Pigsonthewing ); Talk to Andy; Andy's edits 16:08, 17 January 2013 (UTC)
 * And also the recognition that some editors may not understand English very well as they are global users. --Rschen7754 19:19, 17 January 2013 (UTC)
 * Oppose For the bigger articles, some of these gnomish edits can be removing 4 or 5 KB of content from what is an already overly long article, which is a good thing and not something we should be stopping. KTC (talk) 14:43, 17 January 2013 (UTC)
 * Agree; this needs a size limit; and a bot should run for larger articles with many such links. We should also ensure that our articles are not promoted to GA/FA, unless the interwikis have been replaced. Andy Mabbett ( Pigsonthewing ); Talk to Andy; Andy's edits 16:08, 17 January 2013 (UTC)
 * While I think all GA/FA should be converted, that opinion's not grounded in the GA/FA criteria. --Rschen7754 17:22, 17 January 2013 (UTC)
 * Then the community should update those criteria; and not let the tail wag the dog. Andy Mabbett ( Pigsonthewing ); Talk to Andy; Andy's edits 18:03, 17 January 2013 (UTC)
 * You're welcome to propose it. --Rschen7754 18:54, 17 January 2013 (UTC)
 * This is a centralised RfC; I just did. Andy Mabbett ( Pigsonthewing ); Talk to Andy; Andy's edits 18:56, 17 January 2013 (UTC)
 * Comment per Graeme Bartlett; this is very badly worded. It appears to start with a rationale, and then segue into a proposal, rather than the more sensible vice versa. It must be made understandable to people with no prior familiarity. Andy Mabbett ( Pigsonthewing ); Talk to Andy; Andy's edits 16:08, 17 January 2013 (UTC)
 * Speedy close the discussion and rewrite the proposal completely until it's made clear what is exactly being proposed (I only vaguely understand that something should not happen with Wikipedia edits related to Wikidata), who will be affected, what restrictions it will impose on those affected, and why this proposal was made (no, a link to a news report at the top labeled "Background" is not enough). The proposal seems to have deep implications, but the consequences of either passing it or rejecting it are clear as mud. Diego (talk) 16:55, 17 January 2013 (UTC)
 * Speedy close per Diego Moya; good call. Andy Mabbett ( Pigsonthewing ); Talk to Andy; Andy's edits 18:03, 17 January 2013 (UTC)
 * Oppose. Drama alert: more divisive discussions over Wikidata will be coming later on; Wikipedians tend not to like people inserting things in Wikipedia that can't be dealt with through the people and processes we have here, and so far, I'm not seeing anyone tackling that issue, here or at Wikidata. So ... when possible, let's try to get clarity and transparency into RfCs involving Wikidata. No objection to the general goal of following established WPian norms on editing by humans; no objection to supporting the development of intelligent scripts to help with the workload; no objection to many of the goals of Wikidatans (Wikidatians?) - Dank (push to talk) 18:28, 17 January 2013 (UTC)
 * It's a problem that Wikidata is seen as "outside" Wikipedia, rather than an adjunct to it; I see the same in Wikivoyage, in the other direction. Sad. Andy Mabbett ( Pigsonthewing ); Talk to Andy; Andy's edits 18:56, 17 January 2013 (UTC)
 * It's not necessary that the communities be separate, but so far, they are, largely. The fun part comes when they start talking with each other ... then we'll have a lot of "ZOMG what were you thinking?" conversations. I'm not pessimistic though, I think there's a lot of intelligence and a lot of good faith in both communities. - Dank (push to talk) 19:15, 17 January 2013 (UTC)
 * The problem is not "what were you thinking?" as much as "wtf are you talking about?" A discussion addressing both communities should not assume that people have a deep understanding of what is going on, and should try to summarize and explain the background and implications of the discussed matter. Diego (talk) 07:12, 18 January 2013 (UTC)
 * I can't help it if people don't read the linked pages before giving their opinion on the RFC. --Rschen7754 07:16, 18 January 2013 (UTC)
 * You're asking us to support your proposal. The least you could do is try to minimize the amount of work you request from other editors, by pointing out what part of the linked pages (which are pretty long, and of unclear relation to the proposal) is relevant and explain why. Otherwise, it's legitimate to react to it as "unclear proposal, will have unknown effects, I oppose changing the status quo" that you're getting. Diego (talk) 13:18, 18 January 2013 (UTC)
 * On the contrary, I see no reason to try and explain myself the Wikidata concept when others have done so, and even better than I could. --Rschen7754
 * If, after all the oppose !votes, you still can't see the need to submit a clear proposal with reasonable context and background, I cant't do anything more than what I did to try to enlighten you. Diego (talk) 10:58, 20 January 2013 (UTC)
 * You could say the same thing about Commons... --Rschen7754 18:52, 17 January 2013 (UTC)


 * Comment As currently written, this proposal seems likely to cause the opposite of what it is trying to achieve. We dislike edits that have no visible effect because they usually only tend to flood everyone's watchlists, recent changes and the page histories without much positive benefit (besides satisfying someone’s obsessive-compulsive desires). In this case, however, not making one single edit to remove the interwiki links might result in multiple edits over time by the interwiki bots.
 * A better thought out plan is needed (for example, the interwiki bots must replace the wikitext-based interwiki links with wikidata-based interwiki links if and only if they would otherwise have updated the interwiki links.) —Ruud 16:41, 18 January 2013 (UTC)


 * Oppose in favor of a policy of neither encouraging nor discouraging the removal of non-FA/GA interwiki links. There's no harm in letting people remove them whenever they want. I very much doubt that after the first few days there will be many widespread campaigns of interwiki link removal. Most importantly, the main problem is that a policy like this would just be a paper tiger: Would we actually warn or block users for "unnecessarily removing interwiki links"? Common sense should take care of this. Additionally, in the spirit of writing the best proposal, whether or not I agree with it, you should probably also include the resolution of interwiki conflicts (e.g. we list a French article as a translation, and that French article lists an Arabic one as a translation, and that Arabic one lists a different English article), which I've come across several times on Wikidata, and are the kind of issues that Wikidata was created to avoid. — PinkAmpers   &#38;   ( Je vous invite à me parler )  23:25, 18 January 2013 (UTC)
 * Speedy close in order to give a revised proposal a chance to be considered before Wikidata is enabled. —WFC— FL wishlist 19:15, 21 January 2013 (UTC)

Discussion
Could you add a short summary to the top that explains what is going to happen? If I understand, they will load all the interwiki data into wikidata, at which point we can safely remove all interwiki links from articles. And the proposal is to do that slowly, rather than trying to have a bot edit every article on the wiki. &mdash; Carl (CBM · talk) 11:41, 17 January 2013 (UTC)
 * My understanding is that the intention is to have this done alongside other edits (perhaps recruiting the developers of AWB for this?), where it won't create "new" revisions taking up extra server space and connection time unnecessarily. עוד מישהו Od Mishehu 13:00, 17 January 2013 (UTC)
 * I don't understand this. Surely using Wikidata capability is an improvement to any article and not a null edit, right? Isn't that the whole purpose of Wikidata? Why would we discourage this? Rmhermen (talk) 16:11, 17 January 2013 (UTC)
 * This. --MZMcBride (talk) 17:03, 17 January 2013 (UTC)
 * Because at the beginning it will not provide any visible improvement to the article. --Rschen7754 17:17, 17 January 2013 (UTC)
 * So? That's also true of hidden categories, and I haven't seen any proposal to limit then. What is exactly the problem that you're trying to avoid, and why this RfC was created? Some concrete example of what kind of edit is affected will also help. Diego (talk) 18:20, 17 January 2013 (UTC)

Neither the proposal nor the background links explain what we are replacing interwiki links with. If I create a new en:Friend and want to connect it to the existing zh:朋友 and es:Amigo, how do I do that? I don't think it's possible to comment intelligently on this proposal without that information. – Philosopher Let us reason together. 17:15, 17 January 2013 (UTC)
 * You have to create an item on Wikidata, which is on another project and outside the scope of this RFC. --Rschen7754 17:17, 17 January 2013 (UTC)
 * I assume these are being created automatically from the current interwiki links, right? This is the sort of thing that a short summary could clear up. &mdash; Carl (CBM · talk) 23:05, 17 January 2013 (UTC)
 * There are bots working on the Wikidata site, but with 7-8 bots and millions of items, it's taking a while. --Rschen7754 23:12, 17 January 2013 (UTC)
 * I'm not sure then how a user would know when it is safe to remove the interwikis. The front side of this page implies that at the time of rollout, it will be safe to remove the interwikis as long as it is part of a larger edit... &mdash; Carl (CBM · talk) 12:32, 18 January 2013 (UTC)
 * Also, the blog post seems to claim the data has already been collected. I have started a new section below about that. &mdash; Carl (CBM · talk) 14:13, 18 January 2013 (UTC)
 * A lot of Wikidata users have a script installed, as I've indicated below. --Rschen7754 18:41, 18 January 2013 (UTC)


 * I believe if we want to proceed in a civilized way we need to do the following:
 * Write down the policy. People usually tend to be lazy, they will not click on links. A compact, one paragraph proposal will do the job;
 * Discuss the exceptions, pretty much in the same way it was discussed in Wikidata: if the proposal gets implemented, what happens if a new iw link gets added to Wikidata? Does it show up in the en.wp watchlist? In the history? In the list of recent edits? What happens to the articles under PC, do edits get marked as unprotected? How do interwiki conflicts get handled?
 * Make a roadmap: what needs to be done before we can switch this on. (For example, enable the edit history modification).
 * I am afraid without this the RfC has a chance for snow closure, and then it will be difficult to discuss this really serious issue anytime soon.--Ymblanter (talk) 15:55, 18 January 2013 (UTC)
 * On the contrary, if this gets SNOW closed, then I will tell all the Wikidata editors that it's open season in terms of removing interwikis, and wash my hands of any complaints that we get. --Rschen7754 18:49, 18 January 2013 (UTC)
 * You perfectly know that Wikidata is not in a position of telling any other project what to do, and if the thing escalates we can get pages like "Boykott Wikidata" or just a full prohibition of interwiki link removals. And the problems I outline can not be addressed in this kind of setting. It is still better to de-escalate than to escalate things.--Ymblanter (talk) 19:01, 18 January 2013 (UTC)
 * If there is no prohibition against such removal, then editors are free to do as they wish. I'm just trying to have things a little more orderly than they were for the rollout at hu.wp earlier this week, multiplied by the chaoticness of the English Wikipedia. --Rschen7754 19:40, 18 January 2013 (UTC)
 * What do you mean by no prohibition? Currently removal of interwiki links is considered to be a sort of vandalism and can be rolled back.--Ymblanter (talk) 19:45, 18 January 2013 (UTC)
 * .... what? Good-faith edits are never vandalism. --Rschen7754 19:48, 18 January 2013 (UTC)

From the dev corner
The articles will be a lot smaller after the change, and after a few revisions the saving will be much larger than one additional revision would give, and after some time its most likely there will be an additional revision anyhow.

In addition this policy would enforce continuous operation of iw-bots for prolonged time as we can't rely on working sitelinks for working iw-links. If there are langlinks there they will take presedence. So if the links get out of sync things gets weird.

As this isn't enough templates that are using properties will not work. The parser function for properties could although be set up to use sitelinks anyhow and just disregard  any langlinks.

I would say let the bots do their job, it will take weeks, probably months, to switch from langlinks to sitelinks. Perhaps the discussion should rather be whether English should be an earrly bird and rather wait. -- jeblad on his cellphone 82.113.121.29 (talk) 21:01, 17 January 2013 (UTC)


 * Just to make sure: this is not an opinion of the dev team, but of a single developer, who totally have the right to speak their own mind :) But please do not read this as "the Wikidata team says". Officially we do not have an opinion on this stance. We will answer all technical questions that you have, but we leave any policy decision to the communities of a given wiki. If there is a policy decision which might raise issues, and we are made aware of it, we will comment on it to point this out, but in general we regard the Wikipedia communities as autonomous. Cheers! --Denny Vrandečić (WMDE) (talk) 10:39, 18 January 2013 (UTC)


 * What does "it will take months to switch" mean?
 * It will take months to load the data into Wikidata
 * or
 * It will take months to remove all the remaining links from enwiki after the data is loaded into wikidata
 * I assume that the data will be loaded using the info already in the enwiki database's langlinks table, right? &mdash; Carl (CBM · talk) 12:37, 18 January 2013 (UTC)
 * I doubt it will be that easy, there are a lot of times where interwiki links are a mess. What makes enwiki more correct than say dewiki in these cases? Mass conversion will probably work for 90% of cases but the remainder will be a disaster to straighten out. Werieth (talk) 13:29, 18 January 2013 (UTC)
 * I read through the two links provided on the other side of this page, and the proposal. What I took away was: after a certain data, interwiki links on enwiki will be converted to use Wikidata. At that point, we can begin to blindly remove all interwiki links from the source code of pages on enwiki, because the data will already be in Wikidata. The proposal here states that the removal should only be done as part of another edit, rather than having people simply go from article to article removing interwiki links. If that is not what the proposal is saying (and it seems like it is not, from your comment) then the proposal needs to be significantly clarified. (I realize you did not write the proposal; I am not trying to assign any blame, just to clarify the situation that I don't understand.) In particular, in what circumstances can I blindly remove the interwiki links from a page on enwiki? &mdash; Carl (CBM · talk) 14:07, 18 January 2013 (UTC)
 * You can never blindly remove links from a wiki page after implementation of Wikidata. For example, let's say that the Wikipage for John Doe matches a page named "John Doe" in four or five language projects, but on 4 or 5 other ones it links up to "John Doe Company, Ltd", because those languages put more emphasis on John Doe's company than on his biography. Eventually, there will be two articles on all Wikipedias (if John Doe and his company are notable enough in all languages). The trick is sorting and checking the interwikis on Wikidata. Everyone who is checking interwikis will have to make this manual check on Wikidata, rather than the source code of the Wiki page. This will have to be manual, and therefore it will go slowly. Jane (talk) 17:29, 18 January 2013 (UTC)
 * Thanks for the info. If this is the case, I am not sure (and of course this is not your fault): why does the RFC talk about "Edits that solely remove the interwiki links should not be performed," if it is not possible to make sure edits blindly in the first place? &mdash; Carl (CBM · talk) 17:47, 18 January 2013 (UTC)
 * If someone has not yet played around with Wikidata but just reads all the headlines about it, it would be easy to assume that during any casual edit that all interwikis can be blindly removed from any article, which of course is perfectly possible to do. However, per the above, Wikipedians should *not* blindly remove the manual interwiki links without checking them against the Wikidata links first. In my example above, let's say that after Wikidata implementation, one of the languages that previously had just one article on John Doe creates an article on John Doe's company. The person creating the article on the company needs to untangle the interwiki links on the John Doe page as well as on Wikidata for both articles. Jane (talk) 11:05, 19 January 2013 (UTC)
 * I have a script installed that tells me if there is an item on Wikidata for the current article. --Rschen7754 18:25, 18 January 2013 (UTC)


 * @CBM: that is exactly what the proposal is saying. --Rschen7754 19:06, 18 January 2013 (UTC)


 * @Rschen7754: the proposal does not really say anything clearly about the process. So far on this talk page I have seen "You can never blindly remove links from a wiki page after implementation of Wikidata" and elsewhere I have seen someone claim that manual attention is required to examine each Wikidata item. But you are talking as if someone might start doing several articles per minute. Someone else said it's not even the case that all the links are loaded into Wikidata. I'm not planning to comment here additionally, because I don't think I can say anything sensible. &mdash; Carl (CBM · talk) 19:26, 18 January 2013 (UTC)


 * That's basically what I was getting at earlier - this proposal assumes that contributors are already wikidata experts. At this point, it's not even fair to assume that people know what it is, let alone that they understand how it works/the background/etc.  I, similarly, have nothing else to offer here. – Philosopher Let us reason together. 06:54, 20 January 2013 (UTC)
 * On the contrary, I see no reason to include the facts about how Wikidata works, which are not up for discussion because that's how the developers programmed the site. Neither do I see a reason to include information that may become outdated at any point, as the developers are still changing the site around enough to break many bots. This would be like explaining how to use Commons in our image policies. --Rschen7754 06:58, 20 January 2013 (UTC)

Is the data in Wikidata?
According to the blog post,
 * Now that the editors (with the help of a few bots) have collected all these links to articles on the various Wikipedias it is time to make use of them.

I read that as saying
 * Editors and bots have collected into Wikidata as much data as can be automatically collected from all the Wikipedia projects, including enwiki.

Is that not what it is trying to say? &mdash; Carl (CBM · talk) 14:11, 18 January 2013 (UTC)
 * They've collected a massive amount of data and are continuing to do so. It's not complete yet however. --Lydia Pintscher (WMDE) (talk) 14:45, 18 January 2013 (UTC)
 * Someone calculated that there are 17 Million different topics i.e. possible Wikidata Items i.e. groups of languagelinks of that 3 Million have been created yet and each 15 day about another million is created. But only a few bots are working on it yet once it goes live on all wikipedias all the interwikibots could work on wikidata instead and create the rest in no time.--Saehrimnir (talk) 22:24, 18 January 2013 (UTC)

Common sense
My thought is that this RFC needs common sense behind it. If an article has substantial savings to removing the interwiki links, that's a good thing. But what we don't want is untrained hordes of people racing around to every article trying to scrub the interwiki links out. Among other things, doing it from regular accounts will flood recent changes and make patrolling more difficult. Also, people could screw up and take out GA/FA data. And some interwiki data apparently hasn't been ported to Wikidata yet. If it's done, it should be done by a limited number of bot accounts that are set to target articles with significant numbers of interwiki links (say more than 10 or 20). It should also be integrated into AWB and .py framework so that as other bots and users go through and do other changes, it will be done along with them (like how we did the Image: -> File: conversion).

I assume a wikidata dev (or MZMcBride or Legoktm) could give us a nice chart of the number of articles with at least 1, 5, 10, 20, 50, or 100 interwiki links to give us some actual impact data to work with. Could one of you do that, please? Thanks.  MBisanz  talk 19:32, 19 January 2013 (UTC)
 * Common sense sounds good! After a day or two of contributing to Wikidata I found the odd set of data that was missing interwikilinks. Makes sense to keep them on EN until we are sure that all data is extracted (if that is possible). I think, when the time comes, the task should be dealt with by bots on the whole and if an editor feel inclined to remove interwikilinks while making another edit then they should be able to do so, But not just to load up something such as AWB go through every article looking for interwiki links.  ·Add§hore·  T alk T o M e ! 19:44, 19 January 2013 (UTC)


 * Per request posted at Wikipedia_talk:Database_reports:
 * {| class="wikitable"

! Has at least this many wikilinks ! Number of articles
 * 1
 * 2650578
 * 5
 * 837048
 * 10
 * 411428
 * 20
 * 166797
 * 50
 * 20987
 * 100
 * 3833
 * } - TB (talk) 20:10, 19 January 2013 (UTC)
 * Based on that data, I think it would be wise for a bot to run through those articles with more than 50 interwiki links and leave articles below that threshold to the normal editing of bots, AWB, and what not.  MBisanz  talk 20:17, 19 January 2013 (UTC)
 * I like this idea, there's probably a lot of overlap with watchlists in those with lots of interwiki links due to them likely being high profile articles. Get the most watched pages out of the way quickly and painlessly. James086 Talk  12:07, 20 January 2013 (UTC)
 * One of the things AWB is very good at is these "edit side-effects" - that is, a user loads a page to do some other task (such as stub sorting, semi-automated category renaming, etc) and these edits are suggested automaticly. I definitely think that AWB should implement this - and make sure they know what they're doing; if they do this - we will probably have lots of editors who will gradually remove these while not taking them off the wrong articles. עוד מישהו Od Mishehu 06:24, 21 January 2013 (UTC)
 * It would be great if in AWB you could somehow link to the wikidata item for checking. I also agree that 50+ interwikis will be a safe bet for auto-removal of interwiki links. In my example above I was thinking of the far more common examples with 4-10 interwiki links. Jane (talk) 13:32, 21 January 2013 (UTC)
 * Well, if bots are going to be removing the links from here, they should first be double-checking to make sure that Wikidata has all of them. For highly-interwikied articles, it's actually quite common for us to have a handful of links that Wikidata doesn't have. — PinkAmpers  &#38;  ( Je vous invite à me parler )  06:01, 2 February 2013 (UTC)
 * Does it matter if wikidata has all of them? Surly you just need to check if the language link you are going to remove exists on wikidata, and then remove any others that also appear linked to your article..?  ·Add§hore·  Talk To Me! 18:46, 13 February 2013 (UTC)
 * Based on that data, I think it would be wise for a bot to run through those articles with more than 50 interwiki links and leave articles below that threshold to the normal editing of bots, AWB, and what not.  MBisanz  talk 20:17, 19 January 2013 (UTC)
 * I like this idea, there's probably a lot of overlap with watchlists in those with lots of interwiki links due to them likely being high profile articles. Get the most watched pages out of the way quickly and painlessly. James086 Talk  12:07, 20 January 2013 (UTC)
 * One of the things AWB is very good at is these "edit side-effects" - that is, a user loads a page to do some other task (such as stub sorting, semi-automated category renaming, etc) and these edits are suggested automaticly. I definitely think that AWB should implement this - and make sure they know what they're doing; if they do this - we will probably have lots of editors who will gradually remove these while not taking them off the wrong articles. עוד מישהו Od Mishehu 06:24, 21 January 2013 (UTC)
 * It would be great if in AWB you could somehow link to the wikidata item for checking. I also agree that 50+ interwikis will be a safe bet for auto-removal of interwiki links. In my example above I was thinking of the far more common examples with 4-10 interwiki links. Jane (talk) 13:32, 21 January 2013 (UTC)
 * Well, if bots are going to be removing the links from here, they should first be double-checking to make sure that Wikidata has all of them. For highly-interwikied articles, it's actually quite common for us to have a handful of links that Wikidata doesn't have. — PinkAmpers  &#38;  ( Je vous invite à me parler )  06:01, 2 February 2013 (UTC)
 * Does it matter if wikidata has all of them? Surly you just need to check if the language link you are going to remove exists on wikidata, and then remove any others that also appear linked to your article..?  ·Add§hore·  Talk To Me! 18:46, 13 February 2013 (UTC)

Interwiki links and G/FAs
Just a thought, if there is a G/FA in another language and therefore link FA or link GA exists on a page, once the extension is enabled would that language be listed twice? Edgepedia (talk) 18:30, 13 February 2013 (UTC)
 * No, because those templates do not generate the links. --Rschen7754 18:35, 13 February 2013 (UTC)
 * I just checked, if interwiki links get removed but the templates stay in the article, everything looks perfect.--Ymblanter (talk) 21:28, 13 February 2013 (UTC)

Wikidata deployed
Wikidata Phase I (interwiki links) was deployed on English Wikipedia about an hour ago.--Ymblanter (talk) 21:29, 13 February 2013 (UTC)
 * Where is this documented, with Wikipedia editors in mind? Andy Mabbett ( Pigsonthewing ); Talk to Andy; Andy's edits 22:58, 13 February 2013 (UTC)
 * Village pump (miscellaneous) contains the necessary links, including the blog post and the FAQ on Meta.--Ymblanter (talk) 23:03, 13 February 2013 (UTC)

Shouldn't this RfC be closed, now? Andy Mabbett ( Pigsonthewing ); Talk to Andy; Andy's edits 11:52, 18 February 2013 (UTC)