Wikipedia talk:Hard category redirects


 * See also: commons:Commons:Requests for comment/Hard category redirects REDIRECT.

Please discuss the proposal (mine) to switch to hard category redirects.--Kotniski (talk) 17:46, 23 October 2009 (UTC)

Hotcat
I suppose as an advantage of hard redirects you could mention that if a user is using hotcat and he adds the redirected category, hotcat will automatically apply the redirected-to category. Hotcat doesn't do this with softredirects. Not everyone uses hotcat so it's really only a minor benefit. Good Ol’factory (talk) 00:05, 24 October 2009 (UTC)
 * Mentioned it on the page.--Kotniski (talk) 08:43, 26 October 2009 (UTC)
 * Actually, it should fix both soft and hard redirects. I recently fixed that after someone changed the structure of the soft redirect categories. If you find a specific case of a soft redirect that is not automatically resolved by HotCat, then please let me know. —Th e DJ (talk • contribs) 10:31, 31 October 2009 (UTC)

FYI, HotCat on Commons automatically resolves soft redirects. It has for a long time. (I don't know about hard ones, though.) Rocket000 (talk) 19:28, 31 October 2009 (UTC)

Proposal
It is proposed to switch from soft to hard redirects for categories, for the reasons described here.--Kotniski (talk) 08:43, 26 October 2009 (UTC)


 * Fully support. Will also have the added benefit of reducing the number of categories mistakenly (re)created by users unaware of the relevant prevailing naming conventions.  --  X damr  talk 12:59, 26 October 2009 (UTC)
 * Support, after much consideration. Hard redirects for categories is not a perfect solution, but in my opinion we have reached the stage where hard redirects are preferable to soft redirects with categories. Good Ol’factory (talk) 22:53, 26 October 2009 (UTC)


 * Oppose. A hard redirect in such a case is inherently confusing; whereas a soft redirect is only "ugly" and "confusing" to the extent that our template sucks. The template can be fixed. Also, this claim is fallacious:
 * "The situation described in (B) is possible only during a short period (a few hours or days) for a given page, while it waits for the next bot run, while the main benefit described in (A) is a permanent one."
 * On the contrary, benefit (A) and drawback (B) come into being simultaneously, the moment someone links to or categorises into a redirect category; and disappear simultaneously the moment a bot fixes these links. Hesperian 00:11, 27 October 2009 (UTC)
 * No, benefit A (at least, the main part of it) remains for the lifetime of Wikipedia. And I don't know what you mean by saying hard redirects are inherently confusing. Are you suggesting that all redirects be made soft? If not, why use soft ones for categories specifically?--Kotniski (talk) 10:08, 27 October 2009 (UTC)
 * Support. Makes good sense, especially if my understanding is correct. From the historical start of discussion it sounds to me like the use of soft redirects was a work around for a problem with Software. Will this change be a continuation of an interim solution until the Software is fixed? Pknkly (talk) 18:17, 27 October 2009 (UTC)
 * Well, yes and no - the software fix we're still waiting for is something that would automatically put pages from a redirected category into the target category. The interim workaround for that problem is the bot. The question of whether we use hard or soft redirects is independent of that (though if such a fix was made, then it would be even more obvious that hard redirects are preferable).--Kotniski (talk) 13:54, 28 October 2009 (UTC)
 * In case anyone is interested, the relevant bug is 3311. Also, registered Bugzilla users can "vote" if they want this bug to have a higher priority. --R'n'B (call me Russ) 14:02, 28 October 2009 (UTC)
 * Support, Seems very practical to me. Jeepday (talk) 22:02, 27 October 2009 (UTC)
 * Comment: for an earlier discussion of a similar proposal, please see Template_talk:Category_redirect. --R'n'B (call me Russ) 14:02, 28 October 2009 (UTC)
 * Strong Opppose - Unless something has happened in the last few months since the last several times that this has been proposed. There have been some very good comments concerning this in previous discussions, and I don't see much here justifying trying again, except to just "try again". - jc37 15:33, 29 October 2009 (UTC)
 * What do you mean? What comments are you referring to? I'm not aware of any arguments against other than the weak ones presented on this page (but if you know of some, please tell us).--Kotniski (talk) 19:47, 29 October 2009 (UTC)


 * Strong oppose HotCat should be fixed to recognize soft redirects. Why should Wikipedia be changed to conform to how HotCat (or any optional tool) functions? A hard redirect will not inform users that they are adding the wrong categories to articles, if they bother to navigate into the categories, since they will not see it, unless they bother to look at the small notice about being redirected... so these editors will forever be wondering why they can't categorize things, since they don't show up in the category when they click on it, especially if the bot were to fail for an extended period of time, which all bots are known to have from time to time. Further, they don't hurt navigation, since you can see articles in the category they are placed in, and when you click on the template, you can see the rest. You wouldn't see these articles if you were redirected. 70.29.209.91 (talk) 19:55, 29 October 2009 (UTC)
 * But the same applies the other way round - you can navigate in one direction or the other (in either case there should be an explanation of what's going on). But you seem to ignore the whole point of redirects - we shouldn't configure them for the very rare situations where some editor is adding an article to the wrong category, we should be thinking about readers who all the time might enter "Category:American Xs" when the category is called "Category: US Xs" (or whatever) and ought to be taken to the right page without any fuss, as they would with any other redirected title. --Kotniski (talk) 09:00, 30 October 2009 (UTC)
 * We should be thinking about being able to access everything categorized, no? If so, then the contents of the redirected category are "virtually" invisible, since users will not know about the fact that there may be articles there. Since bots are known to not be 100% uptime, and there are delays in moving from one category to another by bot, these articles will be inaccessible to the majority of readers of Wikipedia, unless they know about arcane aspects of Wikipedia like how hard-redirected categories function. 70.29.209.91 (talk) 03:54, 31 October 2009 (UTC)


 * I'm confused about your statement "Why should Wikipedia be changed to conform to how HotCat (or any optional tool) functions?". Somewhere above I mentioned it looked like Wikipedia had a bug that caused hard redirects for categories to not function as it was designed.  So, somehow, it seems to me, a soft redirect was implemented to attempt to emulate or come close to emulating the hard category redirect functions.  This current proposal sounds like it wants to improve that "near emulation".  Your statement makes it sound like Wikipedia was meant to use soft category redirects. Perhaps I'm misunderstanding your statement or don't know the technical aspect of this proposal. What does "changing Wikipedia" mean?  Does the proposal intend to change the Wikipedia program so that a bot will work? Pknkly (talk) 18:08, 30 October 2009 (UTC)
 * No, not at all; it's for the benefit of readers. (That's another misunderstanding in the IP's argument; this isn't being proposed because of HotCat - the fact that we're told HotCat would work better is just a minor incidental argument in favour of the proposal.)--Kotniski (talk) 18:21, 30 October 2009 (UTC)
 * The fact that it works better with HotCat should not be mentioned at all. It makes it appear that the proposal is for the benefit of HotCat users. 70.29.209.91 (talk) 03:54, 31 October 2009 (UTC)
 * The current proposal does not change the fact that the bug still exists, it merely reduces navigability, since a basic user cannot see the articles categorized in the "redirected" category, as they will automagically be transported to the correctly named category. It also reduces basic editor usability, since users who add categories to articles will not see the article show up, since it's in the hard-redirected category and not the correctly named one. Since not all editors use HotCat, the change from soft to hard reduces editor usability for those who do not use HotCat, and are not overly familiar with redirects. Until Wikimedia software is fixed so that redirected categories will automagically dump their contents to the target, soft redirects should remain. When that bug is fixed, the soft redirects can be converted to hard ones. 70.29.209.91 (talk) 03:54, 31 October 2009 (UTC)
 * I understand what you're saying, but I think all this is addressed in the "motivation" section of the proposal. You're giving a lot of weight to a rare, transitory occurrence where it's far from obvious that soft redirects make things any clearer anyway (though I agree with the above comment that the soft-redirect template could be better designed), while ignoring the big and permanent advantage to readers (and editors) of using hard redirects to take them straight to the right page, like they do in any other namespace. Remember that there's already a statement on category lists that the list may not be up-to-date, with a link to an explanation why, so confusion is minimal even compared with a possible improved soft-redirect message.--Kotniski (talk) 11:45, 31 October 2009 (UTC)
 * See also my statement regarding HotCat here:


 * Support just like a year ago when we first realized the problem could be solved this way. I'm not sure why we didn't push it trough back then, but I guess we were all busy. —Th e DJ (talk • contribs) 10:34, 31 October 2009 (UTC)
 * That was this discussion I believe: Template_talk:Category_redirect —Th e DJ (talk • contribs) 10:36, 31 October 2009 (UTC)


 * Comment Ever since the bot started working on these, I thought it was time to do away with the soft redirect system. But there's one and only one thing that gives me concern (and trust me, I've considered everything when creating the current system over at Commons where the bot also runs). And that is security. When you have a bot that will recategorize potentially 100s of pages simply because someone redirected a category, well, that can make quite a big mess. You're essentially letting everyone control a mass-category-moving bot. So we have cool-down period, right? The redirect has time to be reverted if it's unwanted. But this leads to the problem with hard redirects. The reader is redirected but the content hasn't been moved, so they skip over some or all of it probably without realizing it. With soft redirects, they get a chance to see anything in the redirected category before moving on to the rest. Encountering them coming from articles is very rare in my experience and one once in awhile is not really an inconvenience. Rocket000 (talk) 19:57, 31 October 2009 (UTC)


 * That does raise another issue... should redirecting a category be a restricted function? We could have editwars over category names, with the bot recategorizing potentially thousands of articles, the same articles, every few hours... (ie... categories can only be redirected by decision at CfD, unless the category did not already exist (ie. new page creation still will redirect) ... ) 76.66.203.102 (talk) 07:07, 1 November 2009 (UTC)

Well, we seem to be 6-3 in favour. I propose leaving it up to R'n'B to decide whether that constitutes a consensus - after all, it's he who'd have to do the work reprogramming the bot. In the meantime, perhaps we could think about improving the soft redirect templates to make the situation clear to users (I'm not sure that applies to the category one so much as the general soft redirect, in fact).--Kotniski (talk) 12:39, 11 November 2009 (UTC)


 * Strong oppose. None of the problems with hard redirects have been fixed since the previous times this has been discussed. Hard redirects are still broken. It was a long time since I studied these issues, but here's what I remember:
 * Using hard redirects only partly works. It has a number of problems:
 * 1: It depends on that the redirect bot is running. But a bot is not available forever.
 * 2: Pages in a hard redirected category is not visible in the category they redirect to. Thus they are invisible and very hard to find as long as they are not fixed. Thus without the bot we can't fix them.
 * 3: Even after a bot has corrected the categorisation of a page it can take several days before the category lists are updated, if the servers are busy. (Category list updating is a very low priority task in the MediaWiki servers.) Since hard redirected categories are not visible, then those pages are not visible anywhere during that time.
 * 4: Sometimes the category name is added by some template code and thus the bot won't be able to fix it. If we can't see that the category name points to a redirected category, then we can't manually fix it.
 * It is not true that soft redirects are problematic:
 * 5: As long as the bot is running most pages will quickly be fixed. Thus the "click through twice" is not a problem, since it is just visible for a short time. Thus users will only very rarely see a soft redirected category.
 * 6: When the bot is not running it is good to land in the soft redirected category, so we can see that the pages need fixing.
 * If/when the bot is not available we need some other method to find hard redirected categories that have pages in them, so we can manually fix those cases. I asked the devs long ago to make a Special: page that listed redirected categories with pages in, sorted with the one with the most pages on top. But they still have not implemented it. Thus we still have no good way to find pages that needs fixing when the bot is not available. Unless we use soft redirects.
 * And a side note: Smaller Wikipedias (other language editions) tend to copy what we do. But they might not have a bot, so if they use hard redirects they have some serious problems.
 * --David Göthberg (talk) 08:24, 17 November 2009 (UTC)


 * This all seems to be based on extreme scenarios - bot stopping working, people misconstructing templates and no-one realizing it, category lists not being updated (I've never known category lists to take more than a minute or two to update after a direct edit to the page). But the point is that under perfectly normal everyday usage conditions, people who enter the name of a category which happens not to be the right name ought to be taken transparently to the target page as they would with any other redirect. (Maybe they don't do this very often - I would have thought they don't - but when the use of non-keyboard characters in category names are discussed we get people making a big fuss about this very scenario, so for some users it clearly matters, and I don't think it should be held up just on the basis that something scary might happen one day.)--Kotniski (talk) 10:18, 17 November 2009 (UTC)


 * Bots stopping and leaving Wikipedia happens all the time, that's not an extreme scenario. And well designed templates that use categories correctly become outdated all the time because people decide to change the name of a category, so certainly not an extreme scenario.
 * If you have never seen category updates take more than some minutes, then it sounds like you haven't done much category handling around here. Every now and then some servers get the hiccups and have to be taken off-line, then the rest of the servers get overloaded and then it usually gets very noticeable what priorities different tasks have. Category updates seems to have lower priority than most other things. And different kinds of category updates take different time: Removal from a category takes longer time than adding to a category. And updating subcategories in a category takes longer time than updating pages in a category. So removing a subcategory from a category takes the longest time. When the servers are very busy that sometimes takes several days.
 * --David Göthberg (talk) 12:14, 17 November 2009 (UTC)
 * So adding a page to a category, which is what concerns us here, takes the shortest time, presumably? It's always been instant whenever I've done it. And if this bot stopped, another would take over (or someone else would take over the running of it). And templates shouldn't be being used for the kind of category we have in mind here (if they are, then someone ought to know about it and update the template if the category gets renamed - or better, take the category off the template). --Kotniski (talk) 12:30, 17 November 2009 (UTC)


 * Just to be clear, I'm not prepared to make a commitment that RussBot will be able to run as long as Wikipedia exists. However, I do think #3 above is an urban legend; the only time I have seen a significant delay in category updating is when a category is linked through a template, and the template is changed.  When the category links appear directly in the article's text, they always appear to be updated immediately upon saving, apart from any database replication lag (which affects all edits/changes, not just categories).  --R'n'B (call me Russ) 17:13, 17 November 2009 (UTC)


 * I hope I am wrong and that hard category redirects work better than I think. And I know that R'n'B is very skilled, and your bot has been very reliable and have been running for a long time. And yes, if RussBot leaves, then someone else will probably make a new bot in some weeks. Especially since as far as I remember you have published the source code of your bot. So I am actually not that worried. Still, I can't see that my and others concerns above have been handled.
 * Kotniski: Are you saying you will only be using hard redirects for some special kinds of categories? I thought you wanted to use it for all category redirects?
 * R'n'B: Extreme delays in category updates is not an urban legend. I see it every now and then, in all kinds of situations. Are you guys never around when Wikipedia runs on a reduced number of servers? I mentioned the delays since the text on the front of this page claims bot update delays as a reason why hard redirects would be better, which simply isn't true. Both hard and soft redirects have delays of several kinds: Time until the bot does the change, time until a human fixes the cases the bot can't handle, and category update delays and so on. For both soft and hard redirects the delays are annoying, but as long as the bot is working then in both cases the problem is mostly just temporary. But it seems to me the delays and visibility issues are worse for the hard redirects, thus I think that soft redirects are better. And if the bot goes off-line for some weeks then hard redirects cause much more problems.
 * --David Göthberg (talk) 18:57, 17 November 2009 (UTC)


 * What I meant about the "kind of category we have in mind here" was that we wouldn't (or shouldn't) be using templates to generate the sort of categories that get redirected (categories intended for readers). Project maintenance categories (the kind that templates ought to generate) are presumably not left permanently redirected (or if they are, then the templates are updated to use the correct target name), so the question of what kind of redirect we use for them doesn't really arise - or if it does, then it doesn't matter much what answer is chosen.--Kotniski (talk) 07:11, 18 November 2009 (UTC)


 * Kotniski: Take a look at the templates listed under Template messages. There's literally hundreds of templates used in article space that auto-categorize. And they are supposed to do that. And causing problems with category redirects in project space would also be a bad thing.
 * Just waving the hand and saying "templates will be updated when the category is redirected" doesn't solve it. Since if we use hard redirects we won't see that the template uses the old category name. Since if we click on the category name on the page we get redirected and don't notice that the category name on the page is wrong. The user who made the template might not be around anymore, or as usually is the case, has no idea that the category has been renamed. And other users might not be aware of or understand the inner workings of the template.
 * Sure, as long as the bot is running, the bot could list pages that need manual checking, so we can find which templates need updating. But that only works if the bot is running, and if it dumps such a list, and if people are checking that list. That's a lot of if's...
 * (R'n'B: I suppose RussBot dumps such a list somewhere, right? But I looked around, I can't find if and where RussBot dumps lists of pages that need manual handling.)
 * Sort of; User:RussBot/category redirect log doesn't list the pages specifically, but it lists each redirect category that was not empty, and the number of pages found and moved. If the number moved is less than the number found, whatever is left in that category requires manual attention. --R'n'B (call me Russ) 22:10, 18 November 2009 (UTC)
 * But even if all the if's above are fulfilled, that is; even if the templates are updated, there still are several problems:
 * After a redirect has been created there will be a delay until the template is fixed, and then a delay until the system moves most (but not all, see below) of the pages to the new category. Editors that go to the old maintenance category will be hard-redirected to the new category, and will then see an empty category. Thus they will think there is no work to be done. With soft redirects, if they go to the old category they will see the pages, and they will see the category redirect template so they know they should also look in the new category.
 * But note, not all pages will change their categorization just because the template has been updated. Pages also have to be re-rendered, that is visited, for their categorization to be updated. (And sometimes even purged or even null edited.) So if you for instance put a stub template on an article, and then later change the category that template points to, and that stub page is not visited in a long time (or worse: never visited again), then that stub page will continue to only be visible in the old category. So we (the editors) need to be able to see the content in the old category, not just be hard-redirected to the new category.
 * R'n'B: Does your bot have any handling of this problem? And even if it does, all the if's above need to be fulfilled. And it shows this needs a pretty complex bot.
 * Again, sort of. If the bot finds a page in a redirected category but can't find a category link in the wikitext, it will save it anyway. If the template has already been fixed, this null-edit will update all the category links; but if the template is fixed afterwards, it won't do anything until the next daily run. --R'n'B (call me Russ) 22:10, 18 November 2009 (UTC)
 * Again: There are two things that have to be fixed in the MediaWiki software before we can trust hard redirects:
 * 1: Pages in the redirected category should also be visible in the target category. (And it would be nice if the page names would be marked in some way so we can see that they came through a redirect.)
 * 2: We need a "Special:" page that lists redirected categories that still have pages in them, and it needs to be sorted with the category with the most pages in at the top. Thus allowing human editors to find the pages and templates that need to be fixed.
 * If those two things are fixed then hard redirects would be a good thing. And then hard redirects could be used even without a bot, thus could be used on all Wikimedia projects. Although the bot would still save us a lot of manual labour.
 * --David Göthberg (talk) 21:10, 18 November 2009 (UTC)


 * R'n'B: Okay, good that your bot does null-edits to the pages that need it.
 * But I see that only one talk page (+this talk page, +5 archive pages), link to your User:RussBot/category redirect log, and there is no explanation on the log page what the different kinds of items in the log means. So I take it the only person watching that log is you? Which would mean that when you are on vacation, even if your bot is running, no one is taking care of the cases which the bot can't handle. Right?
 * --David Göthberg (talk) 23:39, 18 November 2009 (UTC)
 * All these concerns about template cases (and about what the software ought to do) are doubtless genuine, but I don't see that they have much to do with hard v. soft redirects. If someone views a category that has another category redirecting to it, they won't know what's in that other category regardless of what kind of redirect it is. And if a WP maintainer (i.e. someone who knows a bit about these things) clicks on a category link on a page and finds an empty category, they're going to wonder what's happened and hopefully notice the redirect link. Frankly I think the problems you're predicting are likely to be very rare, and anyway will do no more harm than cause posible slight confusion to WP editors, so should not be used as an excuse for holding back an improvement which will give all readers a better experience in a common everyday situation.--Kotniski (talk) 07:23, 19 November 2009 (UTC)


 * Kotniski: You are constantly oversimplifying things. If the target category isn't empty, but has other pages, then even experienced users won't look in the upper left corner to see if they were hard-redirected.
 * And you seem to totally disregard that if you use hard redirects, then the entire system stands and falls with that R'n'B is watching his bot and the logs it produces.
 * And readers only rarely see the soft-redirected categories, since after all most of the time we do have a bot that moves the pages. And when the users see them they too (just like us editors) have the possibility to see all the pages, both the pages which are still in the old category, and those that are in the new category. So readers that land in the old category actually get better information than readers that followed an updated link or a hard-redirect directly to the target category.
 * So I am not holding back an improvement, it's you who want to deploy a broken system. I'd rather have a slightly rough looking system that works, than a shiny system that doesn't work.
 * --David Göthberg (talk) 13:17, 19 November 2009 (UTC)
 * Doesn't work is rather an exaggeration, but I see we're not going to agree on the relative frequency or seriousness of the two scenarios where this makes a difference. I would rather get ordinary readers straight away to the page where almost certainly almost all of the pages they're looking for are situated, rather than stick in an extra confusing technical page as an obstacle which they might fail to negotiate correctly. We already have to tell people that the category lists might not be up to date, for other reasons.--Kotniski (talk) 13:54, 19 November 2009 (UTC)

I am closing this discussion as no consensus, the RFC period having expired. Kotniski, thank you for your efforts on this issue, but I am unwilling to make any major changes in the bot's operation without a more general agreement among those users who are active in this area of the encyclopedia. --R'n'B (call me Russ) 22:26, 14 December 2009 (UTC)