![]() | The Evidence phase for this case is closed.
Any further edits made to this page may be reverted by an arbitrator or arbitration clerk without discussion. If you need to edit or modify this page, please go here and create an edit request. |
Case clerk: Amortias (Talk) Drafting arbitrator: Opabinia regalis (Talk)
Wikipedia Arbitration |
---|
![]() |
|
Track related changes |
Any editor may add evidence to this page, irrespective of whether they are involved in the dispute. You must submit evidence in your own section. Editors who change other users' evidence may be blocked without warning; if you have a concern with or objection to another user's evidence, contact the committee by e-mail or on the talk page. The standard limits for all evidence submissions are: 1000 words and 100 diffs for users who are parties to this case; or about 500 words and 50 diffs for other users. Detailed but succinct submissions are more useful to the committee. This page is not designed for the submission of general reflections on the arbitration process, Wikipedia in general, or other irrelevant and broad issues; and if you submit such content to this page, please expect it to be ignored. General discussion of the case may be opened on the talk page. You must focus on the issues that are important to the dispute and submit diffs which illustrate the nature of the dispute or will be useful to the committee in its deliberations.
You must use the prescribed format in your evidence. Evidence should include a link to the actual page diff in question, or to a short page section; links to the page itself are inadequate. Never link to a page history, an editor's contributions, or a log for all actions of an editor (as those change over time), although a link to a log for a specific article or a specific block log is acceptable. Please make sure any page section links are permanent, and read the simple diff and link guide if you are not sure how to create a page diff.
The Arbitration Committee expects you to make rebuttals of other evidence submissions in your own section, and for such rebuttals to explain how or why the evidence in question is incorrect; do not engage in tit-for-tat on this page. Arbitrators may analyze evidence and other assertions at /Workshop, which is open for comment by parties, Arbitrators, and others. After arriving at proposed principles, findings of fact, or remedies, Arbitrators vote at /Proposed decision. Only Arbitrators (and Clerks, when clarification on votes is needed) may edit the proposed decision page.
The section on cosmetic changes was added to Wikipedia:Bot policy by these May 6, 2009 edits, following this four-day village pump (technical) discussion. The initial version did not include the most applicable clause, and simply advised that "Scripts that apply cosmetic changes... should be used with caution." The applicable clause, "Cosmetic changes should only be applied when there is a substantial change to make at the same time", was added by this July 11, 2010 edit with the rationale "per long-standing practice
", so it appears there was no discussion before this policy insertion. This May 2, 2011 edit asked "why is the most important sentence last?
", and moved that sentence up to become the lead sentence of the section. On May 13, 2011, the "main clause" was separated from the pywiki specific examples, and for the first time AWB general fixes was listed as the only "main clause" example. The shortcut WP:COSMETICBOT wasn't created until May 19, 2011 (diff), but there are now over 150 links to WP:COSMETICBOT. Among this "what links here" list, three editors stand out for their multiple links: Bgwhite, Rich Farmbrough and Magioladitis, though they are not the only editors found there. Wikipedia:Policies and guidelines § Life cycle states that "Many of the most well-established policies and guidelines have developed from principles which have been accepted as fundamental since Wikipedia's inception. Others developed as solutions to common problems and disruptive editing. Policy and guideline pages are seldom established without precedent, and always require strong community support.
" And, most importantly, "Proposals for new guidelines and policies require discussion and a high level of consensus from the entire community for promotion to guideline or policy.
" A four-day village pump discussion doesn't cut it. Nor does an edit asserting "per long-standing practice". To my knowledge, there has not been a 30-day RfC about this; this hasn't been advertised at Wikipedia:Centralized discussion. Before subjecting these three editors to more grief over this, there should be a proper policy proposal to determine whether there is broad community support for WP:COSMETICBOT, or this is just something important to a vocal minority. Thankyou. – wbm1058 (talk) 02:36, 4 January 2017 (UTC)
The request to make hiding work correctly finished at #47 in the 2016 Community Wishlist survey. Unfortunately this flew under my radar, but with 265 items up for vote, the full survey was TL;DR for me. Retroactively, I would support this in a heartbeat. However, as things stand now, this did not come close to breaking into the top 10 on the list – required to deem it a community priority. Thus, by extension, WP:COSMETICBOT is not a priority for the entire community. It seems it may only be a priority for the subset of the community that uses watchlists to patrol for vandalism. "Watching" article does´t work – this issue was brought up on the village pump in regard to AnomieBOT edits. I noted that AnomieBOT should do a better job of detection of vandalism of maintenance tags. Arguably, setting the date parameter on maintenance templates is a "cosmetic" edit – this is certainly an edit that not only "hid", but "endorsed" vandalism. Anecdotally I believe I revert this bot's edits more often than I revert Magioladitis' bot's edits. – wbm1058 (talk) 19:02, 6 January 2017 (UTC) Noting that the number of community wishlist requests more than doubled from 107 in 2015 to 265 in 2016, it would be nice if community priorities changed so that the community was committed to address 20, rather than 10 items – better yet, increased five times to address 50 items, but this hasn't happened.
There is no mention of the concept of "cosmetic" edits on the Help:Minor edit information page. No edit is so trivial that a human editor must wait until they have a more substantial contribution to make before they may make their "cosmetic" change. Thus, it is impossible for a human to violate the WP:COSMETICBOT rule, as the rule does not apply to edits by humans. The only way a human editor can evade COSMETICBOT when their bot is blocked, is by commanding another bot they control to make the edits. wbm1058 (talk) 22:38, 6 January 2017 (UTC)
Per Wikipedia:AutoWikiBrowser/CheckPage, there are only 74 approved editors that can run AWB as bots. These editors have a Bots tab in their user interfaces. Yobot has a Bots tab. Magioladitis does not. So, at most, Magioladitis can only be a "WP:MEATBOT", but the policy on bot-like editing only advises to pay attention to the edits, and don't sacrifice quality. This section of the policy says that "disruptive editing must stop", implying that poor-quality edits are disruptive, but it does not indicate that cosmetic changes are disruptive. This section doesn't discuss "cosmetic changes". Merely editing quickly, particularly for a short time, is not by itself disruptive. Indeed, the "cosmetic changes" clause does not indicate a rationale for why they should be applied only when there is a substantive change to make at the same time, nor does it say that cosmetic changes are disruptive.
Not to be confused with Help:Minor edit. |
Avoid making extremely minor edits such as adding or removing a single space.– Original rule first added by AWB author Bluemoose on 8 January 2006 (diff)
Avoid making extremely minor edits such as adding or removing a single space or replacing an underscore in a template call with a space.– clarified rules of use, 26 January 2006 (diff)
Avoid making extremely minor edits such as adding or removing some white space or moving a stub tag.– 12 April 2006 (diff)
Avoid making extremely minor edits such as only adding or removing some white space or moving a stub tag. The Spaces section in the Wikipedia:Manual of Style Headings guideline contains details on different acceptable spacing options in and around section headings.– 13 April 2006 (diff)
Avoid making insignificant minor edits such as only adding or removing some white space, moving a stub tag or something equally trivial.– 26 June 2006 (diff)
Avoid making insignificant minor edits such as only adding or removing some white space, moving a stub tag, converting HTML to Unicode, removing underscores from links (unless they are bad links), or something equally trivial.– 23 July 2006 (diff)
Avoid making insignificant minor edits such as only adding or removing some white space, moving a stub tag, converting some HTML to Unicode, removing underscores from links (unless they are bad links), or something equally trivial. This is because it wastes resources and clogs up watch lists.– 10 August 2006 (diff)
Avoid making insignificant or inconsequential edits such as only adding or removing some white space, moving a stub tag, converting some HTML to Unicode, removing underscores from links (unless they are bad links), or something equally trivial. This is because it wastes resources and clogs up watch lists.– 9 July 2007 (diff)
Avoid making insignificant or inconsequential edits such as only adding or removing some white space, moving a stub tag, converting some HTML to Unicode, removing underscores from links (unless they are bad links), bypassing a redirect, or something equally trivial. This is because it wastes resources and clogs up watch lists.– 22 September 2010 (diff)
Avoid making insignificant or inconsequential edits such as only adding or removing some white space, moving a stub tag, converting some HTML to Unicode, removing underscores from links (unless they are bad links), bypassing a redirect, or something equally trivial. This is because it wastes resources and clogs up watch lists. Insignificant edits include, but are not limited to, edits which solely introduce changes which have no noticeable effect on the rendered page. If in doubt, or if other editors object to edits on the basis of this rule, seek consensus at an appropriate venue before making substantial numbers of edits.– 23 February 2011 (diff)
Avoid making insignificant or inconsequential edits such as only adding or removing some white space, moving a stub tag, converting some HTML to Unicode, removing underscores from links (unless they are bad links), bypassing a redirect, or something equally trivial. This is because it wastes resources and clogs up watch lists. With some exceptions (such as changes to the emitted metadata or categorization of the page), an edit that has no noticeable effect on the rendered page is generally considered an insignificant edit. If in doubt, or if other editors object to edits on the basis of this rule, seek consensus at an appropriate venue before making further edits.– 23 February 2011 (diff)
Avoid making insignificant or inconsequential edits such as only adding or removing some white space, moving a stub tag, converting some HTML to Unicode, removing underscores from piped links, bypassing a redirect, or something equally trivial. This is because it wastes resources and clogs up watch lists. With some exceptions (such as changes to the emitted metadata or categorization of the page), an edit that has no noticeable effect on the rendered page is generally considered an insignificant edit. If in doubt, or if other editors object to edits on the basis of this rule, seek consensus at an appropriate venue before making further edits.– 1 March 2011 (diff)
There have been six new version releases in the past year. See Wikipedia:AutoWikiBrowser/History for the version history. Randomly click on a few of the Phabricator links (the T + 6-digit numbers). Observe how often Magioladitis' name comes up as the task creator. I came away with the impression that he is the head of the quality control department for this tool. Imagine how product quality might suffer if Magioladitis wasn't submitting these. wbm1058 (talk) 05:45, 9 January 2017 (UTC)
Humanbot was the inspiration for AWB, per Bluemoose's 29 November 2005 messsage on Wikipedia:Bot requests. The Wikipedia:AutoWikiBrowser page was created on Dec. 1, 2005, a week later the first version (0.2) was up, and barely a month later – on January 1, 2006 people were already complaining about its edits on the village pump. Oh, the horror! This new tool had efficiently removed all the links to dates in Israel, including [[1967]], [[1517]], [[1881]], [[1917]], [[1920]]... see WP:YEARLINK. wbm1058 (talk) 06:06, 9 January 2017 (UTC)
RMCD bot searches talk pages which have open ((Requested move))s for templates following the naming convention ((WikiProject
name of WikiProject. There's an open move request on Talk:Stewart Island, and because the bot found the template ((WikiProject Ecoregions))
near the top of that page, it placed a notice of the requested move on Wikipedia talk:WikiProject Ecoregions. However, if instead ((WP Ecoregions))
were used, RMCD bot wouldn't recognize that as following the naming convention, so wouldn't recognize it was a WikiProject template, and thus the notice would not have been posted. Observe function templateprocess
near the top of User:Merge bot/proposedmergers.php. It's a special conversion table for the 16 aliases of ((Merge)), 14 aliases of ((Merge from)) and 17 aliases of ((Merge to)), among others. What happens when an editor decides they don't like one of the existing 17 aliases and adds an eighteenth? My bot doesn't recognize the template, and fails to work correctly. Coding up tables like this is tedious. How many immediately recognize what ((R ud)) does, without looking it up? Would you be annoyed that you needed to make extra effort to look it up? Unbypassed template redirects are not without costs. We have a trade-off here. Which is more important to the Wikipedia project? Avoiding template redirect issues such as these, or avoiding alleged watchlist issues? I haven't seen these watchlist concerns clearly demonstrated, documented and explained. Do such edits still "waste resources" today? Some resources are a lot cheaper today than they were ten years ago.
This bypass is necessary to remove that page from User:AnomieBOT/Nobots Hall of Shame § Pages with broken exclusion templates (diff). wbm1058 (talk) 00:31, 10 January 2017 (UTC)
WikiProject Check Wikipedia was initiated on 6 September 2008 by Stefan Kühn, an administrator and former bureaucrat at the German Wikipedia. |
WikiProject Check Wikipedia (Checkwiki) helps clean up syntax and other errors in the source code of Wikipedia. Yobot and BG19bot are the two primary bots that fix the most comprehensive set of errors; see User:Magioladitis/AWB and CHECKWIKI. Shutting down Yobot likely has a significant negative impact on Checkwiki. – wbm1058 (talk) 21:11, 10 January 2017 (UTC)
AWB will run with "Skip if no changes", "Skip if only whitespace", "Skip if only casing is changed", "Skip if only minor genfixes" etc. activated.
In the continuing saga of BG19bot becoming Yobot's little brother, the last of Yobot's tasks that BG19bot will help with is fixing WP:CHECKWIKI errors."
This is great because it will split the task into 3 bots (Yobot, BGbot19 and MenoBot). This means faster response, less work for me and Bgwhite and more people reporting bugs on AWB. -- Magioladitis (talk) 09:43, 8 June 2013 (UTC)
Regarding #Magioladitis ran an unapproved adminbot to delete more than one thousand pages, he used Twinkle to delete redirects from a talk page of dab page to a non-dab page. There is nothing wrong with this; though my bot's task #3 replaces the redirects with ((WikiProject Disambiguation)) templates rather than deleting the pages. It's easy to find the pages that meet this criteria; they populate Category:Unsynchronized disambiguation talk pages (though Magio must have used a different method to find them back in September 2014 as I just created that category last May. My bot has done this over 9,000 times without any complaints. @Ramaksoud2000: if you remove that spurious section from your list of charges, that will help you get within your evidence limits. Then I'll remove this section, to try to stay within mine. wbm1058 (talk) 00:45, 11 January 2017 (UTC)
So Twinkle's D-batch module is a bot with administrative rights, eh? Perhaps, "broadly construed". You should get right on the job of tracking down all admins who used that tool without filing a bot request for approval first. wbm1058 (talk) 17:06, 11 January 2017 (UTC)
Magioladitis is not currently subject to any Editing restrictions. Past restriction:
I don't know whether there are any more. Confess that though I've been an admin for over a year now, this is an area of administration that I'm not that familiar with. wbm1058 (talk) 04:30, 11 January 2017 (UTC)
Pursuant to the bot policy there is an active discussion open to review the prior bot task approvals of Yobot. That discussion may be impacted by results of this case. — xaosflux Talk 05:03, 4 January 2017 (UTC)
Since the request for this case was initiated, Magioladitis was asked to stand for reconfirmation in the Bot Approvals Group. The reconfirmation was unsuccessful. — xaosflux Talk 05:03, 4 January 2017 (UTC)
The seven unblocks of YoBot by Magioladitis over a period of seven years were substantially in line with normal practice. If a bot has a problem with a specific task, there is generally a way to stop the task. There are various levels of granularity, and of authorisation, needed to invoke this, depending on the platform and configuration.
Notably AWB bots can be stopped by anyone who can edit their talk page: in the case of Yobot anyone even an IP can stop the bot. This will stop all tasks. Therefore there is prima facie no reason for anyone to block the bot, unless either:
Given that a bot is blocked for malfunctioning, rather than malfeasance, it has been normal for blocking admins to say or imply that the bot may be resumed once the issue is resolved. In some cases this may be a very trivial fix, changing a regular expression, or including a new template in a list, and may take minutes or even seconds. To impose some kind of re-authorisation would be unnecessarily onerous on the bot-meister.
There has been some debate over bot-owners unblocking their bots. Notably a finding of fact was made against me for doing this: this was, however, vacated. The general sense was that unblocking for technical reasons (i.e. issue fixed) was not a problem.
Blocking policy says:
Automated or semi-automated bots may occasionally not operate as intended for a variety of reasons. Bots ... may be blocked until the issue is resolved.
It is clear that in all cases Magioladitis unblocked because the issue was resolved.
It is axiomatic that changes that only effect the wikicode, should, in general, not be made in bulk, unless a substantive change is being made at the same time.
The reasons for this are somewhat fuzzy, and not rooted in solid evidence, but nonetheless are generally supported, in particular by the bot community and most very active Wikipedians, including myself and (I believe) Magioladitis. Substantially the over-arching reason is that these changes can be made at the same time as other changes, and that there is sufficient substantive work that the "cosmetic" changes will be done "eventually". (My only concern is that some of these background tasks would benefit from a permissio to complete the long tail, at say 5% of the original volume.)
It is in the nature of complex changes, using AWB, which rely on using regular expressions to parse that there will be edits which fail to achieve their goal. In theory improvements in the system make it easier now than it was five years ago to ensure that they either do so, or are skipped (quite probably Magioladitis was responsible for some of these improvements, as he has contributed a significant amount to the development of AWB). Nonetheless when we have source material which is created by human hands, and has such complexities of structure that even the MediaWiki software sometimes fails to parse it correctly, we may expect that exceptions will occur. Moreover we should not be surprised if a regression also occurs form time to time. This is a known issue in software engineering.
Given that, the majority of issues where someone has complained about a "cosmetic" edit should be considered an acceptable form of error, if they are not high volume. Certainly they are to be avoided, and indeed Magioladitis has spent considerable efforts in doing so. He has even passed some tasks to others, and has certainly asked me to take over several over the years (I have, regretfully, declined).
Magioladitis is quite self effacing about his bot work, but he has contributed enormously over the years. Moreover he is happy to discuss issues on-wiki or in person, and has reached out collegially to instigator of this case, both discussing the definition of cosmetic changes, and inviting him to face-to-face discussions in Canada later this year.
To even chastise such a hard-working, productive and co-operative Wikipedian over a relative handful of edits that are "not sufficient of an improvement" would be counter-productive.
Current word length: 1320 (limit: 1000); diff count: 28. Evidence is too long: please reduce your submission so it fits within limits.
Due to the length of time this case spans, there are truly too many pieces of evidence for me to list because of the evidence limits, and my own free time available to do this. If there is a specific topic that anyone wants to see more evidence for, just let me know.
I also have not included all 24 blocks, and the reasons behind each, because that would be too many.
This section shows that contrary to certain assertions made by Magioladitis and others (including some questions raised by some arbitrators), Magioladitis is fully aware of which kinds of edits violate WP:COSMETICBOT. This section consists of discussion with Magioladitis. For more, see Wikipedia:Arbitration/Requests/Case/Magioladitis#Prior_dispute_resolution.
Assortment of evidence chosen for no particular reason, because all the complaints are too numerous to list. Some automated cosmetic editing was from main account, in violating Wikipedia:Bot policy requiring all bot edits to be performed with a separate account, and with approval, in addition to violating WP:COSMETICBOT. See the block log of his main account.
Yobot block log shows scale of issue as well.
See this December 15 2016 diff for a list of 18 template redirects he orphaned at a rate of about an edit every 2 seconds from his account with unapproved bot runs. See this contribs link which shows a run for one of the templates, ending with him updating that page.
He improperly deleted most of these templates, including Template:Bigfinishbox, Template:GalvanicCells, Template:PeakOil, Template:Gmina Czermin, Subcarpathian Voivodeship, and more. See deletion log.
See the deletion log for this time period. About 400 county stub redirects were deleted in September with the log summary of "Housekeeping / Consensus to delete redirects to county stubs". "Housekeeping" was also the reason for deleting the templates mentioned in the above section. For his consensus claim, I could only find this, where he asked User:Od Mishehu why there was a page of stub redirects. That user said that there was consensus to remove stub redirects over time naturally with AWB, as a fix applied with other fixes. See also this recent section on that user's talk page where the user explicitly says that using AWB just to remove the redirects is not allowed. Of course, Magioladitis could not resist an opportunity to rid Wikipedia of template redirects, so he used AWB from his main account to run another unapproved bot to orphan all these redirects and then delete them. See this contribs link.
Response to User:Od Mishehu claiming it is acceptable because other administrators have done it: Putting aside the issue that there is no policy that authorizes it, the difference is that other administrators do not run unapproved bots to orphan the redirects then delete them on a massive scale. One deletion here and there is no cause for concern, even if done out of process. These mass automated improper actions are. As linked above, you yourself inform Magioladitis that it is improper.
As part of his ongoing crusade against template redirects, he has deleted many other template redirects improperly, too many to count. I spot-checked and did not find TfDs for any deletions where he didn't specifically mention TfD. Likely orphaned in the same manner as above: unapproved bot runs from the main account.
On further review, in response to Magioladitis' comment, these appear to be mainly subpage deletions, although not all of them. However, upon further examination, I discovered the following:
See this deletion log. At a rate of more than one deletion per second, Magioladitis ran an unapproved adminbot to delete more than one thousand "redirects from a talk page of dab page to a non-dab page".
In response to some concerns, I'd like to clarify that these were performed with the Twinkle massdelete tool. However, my understanding is that Twinkle needs a list of pages to delete, like there would be at a large TfD. The pages deleted were obviously generated by a bot that scanned and found the pages meeting the specific criteria. What is the difference if a bot generated a list of pages, and he used Twinkle or a different program to delete them all at once? 00:52, 11 January 2017 (UTC)
Also, wbm1058, you cannot state that it is okay because you have a bot task that was approved recently that does something similar. It isn't a free-for-all. There is good reason why people can't just run unapproved bots from their main account, much less WP:ADMINBOTS. 00:52, 11 January 2017 (UTC)
See the block log
See Wikipedia:Administrators'_noticeboard/Archive279#Magioladitis. This was a very long AN discussion about Magioladitis' quick violation of his temporary unblock conditions prohibiting semi-automated editing. He was re-blocked, but then unblocked solely to participate in the AN discussion. There was wide agreement of a serious issue, but the discussion fizzled out, and Magioladitis restarted the mass automated cosmetic edits from his account and Yobot.
In response to User:Stevietheman's assertion that there may not have been block evasion if AWB was not operated in a bot-like manner: Magioladitis operated AWB in a bot-like manner after Yobot's block. See contribs where he performs an edit every 2 seconds, when the generally allowed max rate for bots is one edit every 10 seconds. I picked this edit randomly out of those contribs, and all it did was bypass template redirects, the reason that Yobot was blocked. Can't be more clear-cut block evasion than this. He was blocked for this block evasion then both him and Yobot were unblocked. Soon after the unblocks, Yobot was reblocked for continuing to make cosmetic edits. After this second block of Yobot last month, he started running the bot program on his account again. See this contribs link where you see him include the standard Yobot edit sumamry with the bot task number, then remove the task number after he notices it to conceal his block evasion.
See WP:BOTAPPROVAL: "All bots that make any logged actions (such as editing pages, uploading files or creating accounts) must be approved for each of these tasks before they may operate." WP:BOTACC requires a separate account. 02:17, 8 January 2017 (UTC)
In response to User:DeltaQuad: As demonstrated above, Magioladitis consciously runs AWB bots that have the sole purpose of bypassing specific template redirects. This is not a "bug". In addition, as demonstrated above, Magiolaidits has been referencing "bugs" as the reason for cosmetic bot edits from both his account and Yobot since 2010. No other bot operators, including those using AWB, seem to have these bugs. Frankly, 7 years and 24 blocks is quite enough time and incentive to either resolve the bugs, or stop using buggy programs. 02:17, 8 January 2017 (UTC)
Check the logs given by Ramaksoud2000 carefully. The 2014 deletions were talk pages or things like Template:UruguayProject/doc. The deletions were based on User:Scott/Notes/WikiProject template redirects. See for example Template:AirlineProject by other admin.
Od Mishehu posted a large lists of stub template redirects in WP:AWB/TR to be orphaned as a low priority task 2 years ago, I asked the rationale behind it and they told there is consensus to do it. In September 2016 we started deleting them from the list we had. I used Twinkle to finish the job 5 days later. Then I asked Od Mishehu to clean the AWB list and we both removed the unused templates.
See evidence presented by CBM.
Some of the complains I get conflict each other: Do this, don't do this, do this but only if you do something else, even if you have approval for this you can't do it (as sole task), etc., etc. I think the ArbCom should take under consideration the following problems:
In conclusion there are two options: The community will have to decide or to allow certain "cosmetic changes" or define some changes as "not cosmetic".
For example FILEMOVE policy requires all instances of the filename to change. A bot under COSMETICBOT won't be able to do that.
Another example: Persondata that does not affect the visual outcome was added by bot and later removed by bot. Persondata had ~2 million transclusions.
One of the biggest examples is the removal of Persondata from almost 2 million pages. The same holds for bot renaming files, etc.
Note here that BG19bot and MenoBot have been approved to perform the same st of edit with Yobot.
For example not all reflists redirects are listed in CHECKWIKI, WPCleaner and AWB core lists. This causes the problem of false positive detections (CHECKWIKI, WPC) of duplicated reflist addition (AWB). This could be hard-coded but the number of pages using redirects of relfist is extremely low. I periodically replaces these redirects to help both detection and avoid bot errors.
In contrary to the claim that Yobot hides vandalism, Yobot tries visit each page only once while other bots visit pages multiple times: [32]. Each time a main error is not fixed WPCleaner finds it and me or someone fixes it. This means the page is checked for vandalism too.
I keep asking people to take over tasks since 2010. This would save me a lot of time. In some cases this would also help in further automatisation
I keep an incomplete list of AWB bug fixes related to CHECKWIKI here: Wikipedia_talk:WikiProject_Check_Wikipedia/Archive_8#AWB_fixes.2Fdetects_more_of_some_errors
The only problem today is the difficult to handle CHECKWIKI part. There were never complains about the section renaming tasks. One more evidence that shows that the problems could be solved by a simple stop of the bot and a talk page message instead of a block.
Someone complained to BasilicoFresco. The edit was actually fixed something. I was the one to reply to the message. Same for BG19bot.
Note: I have been using AWB vigorously for the past 2 1/2 years, so much of the following comes from my experience with it. |
AutoWikiBrowser (AWB) is the software used behind Yobot. AWB has current and former limitations that may have some bearing on this case (Note that this list is not to cast aspersions on anyone involved with the AWB project but merely to lay out relevant facts):
1. Q: So is the assertion that AWB itself is causing the errors to show vs. any negligence by the bot operator?
This is a bit of a sticky wicket to answer. The existence of the cosmetic-related bugs in terms of running a bot (Yobot) do make it within reason to not blame the operator entirely as they have no opportunity to review the saves as they occur. However, it does seem clear the bot operator (Magioladitis) knew of these limitations in that he appears to have accepted an "error rate" of a substantial amount of saves being wholly non-substantive in nature. He is actively participating in running a process he surely knows has this "error rate". As for running Yobot's tasks in his regular user account (Magioladitis), given evidence is shown he was running it bot-like even though in semiautomated mode, saving everything without appropriate review, with repetitive wholly non-substantive saves, I would say he's entirely "on the hook" for knowingly violating AWB Rules of Use, and that's on top of MEATBOT/COSMETICBOT.
Now, here's more of an opinion -- this "negligence" should be weighed against the so-called damage of cosmetic edits, or whether COSMETICBOT was properly created policy, both of which I find are somewhat disputed, and ultimately stemming from a decade-old watchlist bug.
I would guess from a lot of reading about this matter that Magioladitis was ignoring rules for the cause of improving articles. Outside of the wholly non-substantive edits, I don't think there is disagreement that his edits were indeed improving them. And he has improved a huge number of them. There is something to be said about whether the work one is doing is accepted by the community or not, and how that extends to how one responds to complaints, but if the complaints ultimately become hollow (like after that old watchlist bug is fixed), I wonder about the long-term effect of stopping Yobot's work, and the tremendous dearth of article improvements to come. We can complain about Magioladitis' seeming lack of response to complaints, but we should consider pros/cons, looking at his vast positive contributions, incredible amounts of development work, and continually working to address problems in Yobot's tasks (even if not reducing the aforementioned "error rate" to zero, which, by the way, can't be done without a completely new approach per AWB's limitations as described in my earlier testimony). I need to stop before I veer off into a proposed final decision.
2. Q: How close is the gap between the bug fix and software release to the public?
First of all, as a former professional software developer, significant gaps between bug fix and new version availability are normal and tend to be dependent on the severity of the bug involved. With AWB, new version availability after fixes of severe bugs (maybe 'severe' depends much on one's POV) has seemed to be variable. Sometimes, the new versions have come quickly, then others, there seems to be a weeks-long wait. You may want to get AWB's development team to answer this more accurately. Given Yobot's problems related to cosmetic edits, I would hope related bug fixes are considered as severe in that context, and would see reasonably quick availability, even as interim/"nightly" releases.
3. Q: Is there a beta branch that would serve Magioladitis better in these fixes?
Not that I know of. I would suggest interim/"nightly" releases like in #2.
4. Q: Is it practical to stop Yobot until the changes are deployed and updated with the software?
Yes, with qualifications. Yobot does mostly valuable cleanups, but it can wait the hopefully short period until AWB's latest cosmetic-related bug with regards to template redirect bypasses has its fix available in a new release. If other AWB updates are expected, I don't believe that has gelled at this point, so I can't really speak to that. But the rub is that given AWB's continuing limitations, we are going to continue to see an "error rate" with respect to wholly non-substantive saves. So, if Yobot is to be unblocked and run again, the wiki community will, in my humble opinion, need to at least understand and perhaps accept via consensus that there will be this error rate (RfC?), At the same time, the bot operator should be required to "skip only cosmetic changes" on all runs, and "skip only whitespace" on runs unrelated to fixing whitespace. I'm veering off into a proposed final decision again, but I'm just trying to imagine what's in the checklist, so to speak.
4a. Q: Are these bugs being fixed in a timely manner at all?
Per my testimony on AWB's limitations, I reported two cosmetic-related bugs that seem to have a bearing on the case in 2014, with a fix for one coming in 2015 and the other one a few days ago. The one just fixed was clearly triggering repeated strife per other presented evidence. Like I already suggested, with respect to Yobot's troubles, these bugs should have been considered to be severe, and fixed quickly. However, I don't think we should go outside of the scope of Yobot in terms of inquiring about AWB's fix/release approach. This case really isn't about AWB as a whole -- just its aspects pertaining to problematic uses leading to this case.
See this bug reported May 4, 2007. Isn't it ultimately the reason we're all here? If it wasn't for (usually) necessary bot edits "hiding" vandal edits because of this nearly-decade-old bug, would nearly anyone have cared at all about Yobot's edits? Yobot is known to be a highly prolific bot. See also wbm1058's findings. Connect the dots. Anyone please push to have this bug fixed.
Re: wbm1058's contention that fixing this bug is not a community priority, tying at #45 in the 2016 Community Wishlist Survey puts it in the top fifth of proposals, and besides, bugs can be worked on by any volunteer developer at any time, and this bug is currently assigned a 'High' priority. The survey was about special attention from the WMF in fostering the development of proposals that reached the top 10 (and some proposals that help "smaller groups"). Even if you take COSMETICBOT out of the matter, bots with this watchlist bug not fixed still "cover up" previous edits in watchlists when "hide bots" is checked. So, whoever this "covering up" upsets might be especially upset if a bot edit looks useless to them.
I have reviewed WP:BOTPOL, and I am unable to pinpoint how in the case where a bot task has been blocked, the regular user who runs that bot cannot run that same task in AWB's semiautomated/regular mode if they are not running it "like a bot", in other words, going slowly and checking all the suggested changes before saving. I am contesting the concept of "bot block evasion" in this sense as something necessarily rooted in policy. The only way policy seems to come into play is when this regular user is running through a list of articles quickly and with no or inadequate review before saving; in other words, running like a bot on their regular account (therefore, within reason, earning an accusation of evading a blocking action). Whether this aspect of policy was violated depends on reviewing evidence presented by others.
I bring this up not just for this case but to hopefully avoid an overly broad sanction that may affect other bot operators down the road. If a bot is blocked, the operator in their regular user account should be able to run the same AWB task in a semiautomated mode, given they are not acting like a bot and adhering to AWB Rules of Use (esp. #4).
For User:Yobot, clearly WP:COSMETICBOT applies. For both User:Yobot and User:Magioladitis, AWB Rule of Use #4, which begins "Do not make insignificant or inconsequential edits", applies. For User:Magioladitis if running AWB in a bot-like manner (referenced in previous section), there is WP:MEATBOT, which, within reason*, triggers COSMETICBOT, both on top of AWB Rule of Use #4.
* If a regular non-bot account's actions are bot-like, they are behaving as a bot, and therefore bot-related policy applies. If evidence is presented that consistently shows such an account acting as a bot, then any potential sanctions applicable to a standard bot are applicable to them. For this case, that's on top of any sanctions related to use of AWB. Not having a Bots panel in AWB doesn't mean an AWB user cannot or doesn't exhibit bot-like behavior using the tool in its semiautomated mode, as quick repetitive saves without any or proper reviews is what MEATBOT effectively addresses. MEATBOT and COSMETICBOT are part of the same WP:BOTPOL, therefore it is sensible they are used together to form conclusions about behavior.
I would also like to register an objection to the wordplay used in other evidence that suggests that using AWB in regular mode (not bot mode) constitutes "meatbot mode". This is a misreading of WP:MEATBOT and I hope ArbCom will understand that MEATBOT is about a particular behavior/approach while running AWB, and doesn't apply to its use in regular mode in general.
The deletion of unused stub templates is apparently considered acceptable, as is seen by the fact that it has previously been done by other admins, such as Fastily and Wizardman; and I have been doing it myself as a result of seein this being done. Magioladitis didn't do it until after I said it should probably be done. It should also be noted that, in the same message, I told him not to intentionally orphan any of them.
@Izno:I have little doubt that if ArbCom were to forbid Magioladitis from using AWB, he would obey that restriction; if he were to violate it, a temporary block would be palced on him. Even if you think he should be disallowed to use AWB (I don't think so), this should be plennty to prevent him from doing it; only if he unblocks himself would desysoping be ppropriate there. עוד מישהו Od Mishehu 14:07, 16 January 2017 (UTC)
I'm a WP:Template editor and WP:Gnome without a direct connection to Magioladitis, but at least indirectly affected by matters raised in this case. This is more "expert witness testimony" than "evidence". —SMcC |
I have to concur with wbm1058's detailed and well-researched analysis. I would remind that WP:POLICY and WP:NOT#BUREAUCRACY are clear that our policies exist to serve us, not vice versa, and are intended to codify actual best practices not try to force the community to change, nor to favor one camp within it.
An "obvious" consequence of WP:EDITING policy that is not stated frequently enough: No edit that is important to make is "trivial", "cosmetic", or otherwise dismissible as not permitted to be made. An edit may be important or necessary or needed (take your pick) for a technical reason, to comply with a policy or guideline, to fix a definite error (whether it be one of fact, of grammar, etc.), or simply because it's clearly an improvement to the encyclopedia. This WP:Common sense cannot be WP:WIKILAWYERed and WP:GAMEd away by saying "well, we can still prevent an editor from using a tool to make those edits", since if that would make the task so manually onerous that it would not get done, it's obviously subject to WP:IAR. (Guideline matters with multiple acceptable approaches generally shouldn't be performed en masse, but this is a broad principle not a bot or AWB one in particular; WP:MOS's lead section, for example, covers this as a general editing matter.)
Magiolatitis drew a distinction between "make changes that do not affect what the reader sees" and "the definition of 'cosmetic'". This distinction is vital; whether an edit is something infrastructural and technical, or pertains to directly visible changes in the reader's rendered output, has no bearing on whether the edit was necessary/important/an improvement. Undefined pejorative labels like "cosmetic" and "trivial" mean nothing concrete, and it's grossly inappropriate for "enforcement" action to be taken with regard to such "it means whatever I want it to mean" twaddle.
See table below distinguishing "minor", "trivial", "cosmetic", "not reader-visible", "important", etc., and providing at least 7 rationales for why COSMETICBOT has serious policy problems. The gist as it pertains to Magioladitis: COSMETICBOT is intolerably counterproductive, and does not actually represent community practice or common sense. Magioladitis's edits are often being mischaracterized as "cosmetic" or "trivial" simply because some aren't reader-visible, without regard to their purpose; this is fallacious. There is a strong implication being made that Magiolatidis's edits in question (reader-visible or not) were not important, but this has not been demonstrated at all.
Terminological analysis table
|
---|
There's a lot of terminological confusion going on here. Let's clear that up right now:
Hopefully this will help address these matters more clearly and reasonably. |
See also Giraffedata using AWB to perform comprised of → comprises corrections. Many editors who apparently don't own dictionaries have tried to stop him, at ANI and otherwise, and failed in their attempts to paint this grammar cleanup as "trivial"/"cosmetic" (much less controversial/wrong) and thus impermissible. I think this closely parallels much of what some people are trying to pillory Magioladitis for. Wanna-be rules that have been inserted without sufficient consensus for no reason other than to hinder gnomes from bringing content into compliance with guidelines, HTML specs, accessibility standards, reliably sourced English grammar, and other "rules" or common sense, do not magically trump WP:EDITING policy, which represents top-level, site-wide consensus since Wikipedia's earliest days. — SMcCandlish ☺ ☏ ¢ ≽ʌⱷ҅ᴥⱷʌ≼ 12:31, 11 January 2017 (UTC)
Many if not most attempts to prevent bots and other tools, or even humans without them, from performing "trivial", "cosmetic", or reader-invisible edits seem to be to be primarily of one sort (when they are not connected to the "bot-hiding-vandalism" bug discussed in detail above). To wit: attempts to evade the application of guidelines and policies (especially WP:MOS but also WP:AT and the naming conventions, WP:CITE, WP:SAL, and others) to particular articles or topics that an editor or faction feels excessively territorial or possessive about. Most of the rest are simply "wiki class warfare" of large-scale content editors against small-edit gnomes doing maintenance. This is yet another variation of WP:ARBINFOBOX and the other cases where ArbCom has repeatedly had to tell wikiprojects and other clusters of editors that all constructive editorial input and choices of volunteer focus are valued here, that WP:CONLEVEL policy really does apply to your wikiproject too, and that you don't get to make up your own "local consensus" pseudo-rule against site-wide consensus and try to force other editors to abide by it at "your" topic or page.
When editors who perform tedious and often thankless "polishing" work are hounded by others crying "disruptive editing!" simply because the latter have not internalized WP:MERCILESS and WP:OWN, the latter are actually the disruptive editors and are clearly gaming the system for control purposes. This kind of wikipoliticking needs to be sharply and firmly curtailed again, but in more general terms, so the typical "that RfARB was about infoboxes and classical music, and that RfC was about capitalization of species names, so they don't apply to [my pet topic]" rationalization is shut down. Making it clearer that it's a general principle would prevent an amazing amount of future drama.
— SMcCandlish ☺ ☏ ¢ ≽ʌⱷ҅ᴥⱷʌ≼ 12:39, 11 January 2017 (UTC)
Magioladitis deleted multiple template redirects after orphaning them without discussion. This is abuse of the administrator tools, as it circumvents WP:RFD. Other editors have argued that deleting unused template redirects is uncontroversial. That's highly questionable, but even if that were uncontroversial, gaming the system by orphaning the redirects yourself is clearly unacceptable. See, for example, this and this.
AWB Rules of Use #4, discussed in other sections, forbids cosmetic-only editing. Magioladitis has engaged in large-scale tasks using AWB semi-automatically from his main account to replace template redirects with the template name, which is clearly cosmetic-only. For instance, see [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54]. This would normally be grounds for removing AWB access, but that's bundled with the administrative tools. Note that these edits are all from a one minute time period, and considering other edits with non-cosmetic components, there were 28 total AWB edits in that minute. This also violated WP:BOTASSIST, which requires bot approval for semi-automated editing at extremely high rates. For context, even approved bots are usually required to operate at rates of around 6 edits/minute per WP:BOTREQUIRE. These edits are not abnormal from the Magioladitis account.
Until recently, Magioladitis was a BAG member of 2.5 years, meaning that he is expected to be intimately familiar with the entire bot policy. As such, claims of ignorance of the bot policy are dubious.
Additionally, Magioladitis is listed as the lead AWB developer at WP:AWB. He's claimed that many of the errors were the result of AWB bugs, which he is not responsible for as a bot operator, but AWB is his project. If he had ever made it a priority to follow the community's wishes, it was well within his power to fix the AWB bugs causing errors before restarting his bot.
This supplements the explanation of basic issues in other sections of evidence.
Magioladitis has publicly stated this month that he, as a regular practice, "run[s] Yobot in huge untested links". See Special:Diff/757725689. The practice of making changes to Yobot and then running it without any testing falls well below the standards expected of a bot operator. Changes to a bot are allowable without additional approval if the main aspects of the task are not altered, but operators are expected to rigorously test and debug their bots before making large volumes of edits without oversight. It is negligent to run Yobot on large new lists of pages without any testing or oversight.
Magioladitis often asks editors to report errors on his talk page, not the bot's talk page, as posting to the bot's talk page stops an AWB bot from running. See Special:Diff/756871132. This highlights the underlying practice of leaving the bot running while cosmetic-only errors persist. Buggy bots should not run until fixed, but the operator doesn't appear to see it that way.
Special:PermaLink/757674203#Clarification on Yobot task shows that Yobot has greatly expanded its Task 16 since initial approval, including incorporating entirely novel fixes, without returning to the community for additional approval. The original approval enumerated and tested 38 specific fixes. Per WP:BOTACC, bot operators must seek additional approval when performing new tasks unconnected to the ones they were previously approved for. The additional CHECKWIKI tasks without approval are effectively an unauthorized bot.
Magioladitis has claimed that AWB is the cause for many of these errors, but other AWB bots don't have the same error rate that Yobot does. My own AWB bot, BU RoBOT, has certainly had some errors before, but never beyond the trial phase to this degree. Other AWB bot operators have also commented, including botops for CheckWiki (See Bgwhite's section in the collapsed uninvolved section), and confirmed that they don't have problems with cosmetic-only edits. This is a Yobot-specific problem.
Based on the many discussions linked in the original case, it's clear that the community isn't confused here. At most, Magioladitis is confused, but I struggle to even characterize things that way. Please take a (long) moment to read through Wikipedia:Administrators'_noticeboard/Archive279#Magioladitis in its entirety. In that discussion, there was a comprehensive community discussion about COSMETICBOT and how it applies to Yobot and Magioladitis. In that discussion, Magioladitis engaged with community members, asking questions and responding to concerns directly. These are the same concerns we have here today that Magioladitis is claiming he's never seen before. That's more WP:IDONTHEARTHAT than genuine confusion over the policy. Even if the bot policy was as unclear as Magioladitis claims, such discussions make obvious that community consensus is against such edits. An administrator and bot operator is expected to be able to understand consensus, not refuse to hear repeated community demands.
Magioladitis' response to OR in their preliminary statement makes clear that he intends to resume edits that the community regards as cosmetic-only as soon as this case is completed.
before using the last evidence template, please make a copy for the next person
I earlier mentioned that Magioladitis unblocking Yobot seems to contravene WP:INVOLVED.[55] But I also feel it contravenes WP:BLOCK. Which says Unblocking will almost never be acceptable: ... To unblock one's own account (unless an administrator blocked themselves).
Now I would argue that a bot account is "one's own account", or at least an account allowable under WP:VALIDALT. After all, the bot owner knows the password of the account, and is directly responsible for the edits it makes.
I would very much like ArbCom to rule over whether a bot owner who is also an admin, should allowed to be able to unlock their bot's account (unless it was the owner who blocked in the first place); I feel it contravenes the two policies mentioned. --Jules (Mrjulesd) 18:39, 13 January 2017 (UTC)
Wikipedia:AutoWikiBrowser/CheckPage describes how one may come to use AWB on English Wikipedia. Notably (at the time of this case):
There are two consequences:
Among the users who have interacted with Magioladitis in regard to AWB have been several administrators. They include (non-exhaustively) the following:
It is entirely possible that if Magioladitis were not an administrator, then one of these users would have been able to remove his AWB permissions by removing his name, or that of one of his accounts, from the CheckPage.
Yobot is listed as a name on the CheckPage.[60]
While in a dispute which hinges about interpretation of guideline or policy, best practice is to a) interpret the policy or guideline under the most restrictive interpretation to editing while b) seeking clarification via community input (as e.g. with an RFC).
Magioladitis (and others involved with Magioladitis, for or against) is a long-standing user who should follow best practice. He, among others above, seems to dispute either the validity or certain interpretations of WP:COSMETICBOT. An arbitration case is not the place for disputing such validity or interpretations because it is a) too late to edit under the most restrictive interpretation and b) not a seeking for clarification via community input.
As the comments show, Magioladitis has often claimed the problem is fixed, although the previous section shows it was not fixed.
Regarding the pattern of errors re-occuring: [61] [62] [63] . The latter ANI thread has many accounts of attempts to communicate: [64]. In the links above, Magioladitis did not argue that the cosmetic edits rules are vague; his comments over the years reflected an understanding of their generally understood meaning.
It is impossible to tell, in many cases, what Yobot was trying to achieve, apart from general allusions to CHECKWIKI, e.g. these 500 edits [65]. Compare WP:BOTREQUIRE.
In Wikipedia:Bots/Requests_for_approval/Yobot_16, which was for the CHECKWIKI edits, the issue of insignificant changes was raised, and Magioladitis wrote ""Skip if only minor genfixes" and "skip if only whitespace changed" will be activated. I don't think there will be any insignificant changes. We can make some test edits to see and we discuss it again. -- Magioladitis (talk) 14:47, 20 July 2010 (UTC)"
In [73] see ""Restrictions on specific tasks" and "Assisted editing guidelines".
In [74] see "Rules of use", especially "Don't do anything controversial with it." and "Avoid making insignificant or inconsequential edits".
"Accounts performing automated tasks without prior approval may be summarily blocked by any administrator." WP:BOTAPPROVAL since at least 2013 [75].
Disclosure: I'm biased as I consider Magioladitis my closest friend on Wikipedia.
First off, there are some things that are wrong some evidence sections.
Additionally, Magioladitis is listed as the lead AWB developer at WP:AWB. He's claimed that many of the errors were the result of AWB bugs, which he is not responsible for as a bot operator, but AWB is his project. This is incorrect. The developers listed at WP:AWB's infobox are in alphabetical order. The lead developer is rjwilmsi . This is further evidenced by the Changelog. Magioladitis does know some areas of the code, but not even close to 50% of it.
This also violated WP:BOTASSIST, which requires bot approval for semi-automated editing at extremely high rates.WP:BOTASSIST doesn't say required.
Other AWB bots do not have these issuesand gave his own bot as an example. This is misleading. Most AWB bot operators, including BU Rob13, don't have "general fixes" turned on. General fixes are causing cosmetic edits such as whitespace only edits and template renaming. Every CheckWiki error that can be fixed by an AWB bot is contained in general fixes. So, Magioladitis and I have general fixes on. I do have the same issues, but not at the rate Magioladitis does. I too have been threatened with blocks.
Magioladitis ran an unapproved adminbot to delete more than one thousand pagesMagioladitis was deleting these via his account and not a bot. WP:ADMINBOT states,
Administrators are allowed to run semi-automated tools (assisted use of administrative tools) on their own accounts but will be held responsible if those tools go awry.
Good luck on this one. Some here and on Yobot's talk page insist a cosmetic edit is one that is not rendered on the page as seen by the reader. At User talk:Yobot#Simplify wikilinks, SpinningSpark, who blocked Yobot, says, Bots are allowed to make cosmetic edits, but not cosmetic-only edits unless they have explicit permission (which is never granted).
As I mentioned in my preliminary statement, I run a bot task that only removed blank lines between list items. This is a purely cosmetic edit per SpinningSpark, BU Rob13 and others definition of a cosmetic edit. It does nothing to the rendered page. It does make a difference to those with screen readers. My bot also fixes DEFAULTSORT, which also doesn't change the rendered page. Bots have and/or are currently removing PERSONDATA, removing duplicate template parameters, removing spaces and underscores in maintenance templates, remove deprecated parameters from infoboxes and changing http to https.
A listing of fixes in AWB that were CheckWiki related can be found at Wikipedia talk:WikiProject Check Wikipedia/Archive 8#AWB fixes/detects more of some errors. I'm not sure how comprehensive the list is.
Magioladitis handles CheckWiki/AWB issues. If a new CheckWiki error is added or one is modified, Magioladitis does the phab ticket requests and testing.
Magioladitis and I do most of the CheckWiki fixes. At 0z, CheckWiki is run on the majority of articles edited that day. Around ~6z, I start fixing errors. I usually fix via bot and manually the articles listed in the "high" or "middle" priority section. Magioladitis usually does the "low" priority and starts editing from 1-4 hours after me. From today's errors, my bot would run on ~210 articles. Yobot would have run on ~1300 articles. That is a huge discrepancy. The bot will not fix all the problems, so those will have to be fixed manually. It takes me an average of four hours to manually fix the new daily errors.
Notice the time discrepancy. There can be hours between when an error was detected to when we visit the article. This is the source of cosmetic edits with CheckWiki work. The error may have been fixed by the time the bot checks the article. The time discrepancy only gets worse when we aren't doing our normal routine. For example, I usually don't do the Sunday 0z run until 24 hours later.
Yobot's December 13, 2016 block
Magioladitis was running Yobot based on a list I gave him. The list dealt with a modification of CheckWiki error #104. This discussion tells the whys and chronology. Dexbot also ran on the same list and got into trouble. I agree with BU Rob13 in blocking Yobot. However, the discussion on the talk page turned into a rehash of Yobot's sins, thus Yobot was kept blocked even though the issue was solved.
Yobot's December 27, 2016 block
BU Rob13 and Ramaksoud2000 gave 7 examples of cosmetic edits and BU Rob13 then blocked Yobot. BU Rob13 did not look into why Yobot was there or the history log. Those 7 edits had CheckWiki errors that an AWB bot can't fix. They were fixed before BU Rob13 complained. By BU Rob13's same logic, I should also be blocked. My bot won't fix everything. What isn't done by my bot, I then fix manually. Bgwhite (talk) 21:53, 17 January 2017 (UTC)
According to data reported by Rich in /Workshop, around half of all Wikipedia mainspace and talk edits since 2009 have been made with AWB. I'm sure this tells us something, though I'm not sure what. (Update: this is unclear, per further discussion there. But 110 million edits is a lot.)
This isn't the most damning diff in the world but I remember being annoyed by it, and it's the edit that brought me to the case, so I'm writing it up:
Yobot changes ((cite)) templates to ((citation)) in 38 places in the article if I counted correctly. Note that ((citation)) is a redirect target from ((cite)), so they both do the same thing. I see the edit as not just wasteful of human editing time (I had to sit there reviewing all 38 changes) but also as wrong on editorial (WP:CITEVAR) grounds. I'm not crazy about either template but I used ((cite)) since that's what was already in the article when I got to it.
I brought the diff up on Yobot's talk page; see the resulting discussion here.
Magioladitis replies that the change enables some general citation fixes for bots (says "bits" but I think bots was meant) including GoingBatty. The bots should just accept both variants instead of having other bots mess up Wikipedia for bot convenience (bots are supposed to work for humans, not the other way around).
Bgwhite says "AWB will rename redirects to their proper names. See WP:AWB/TR for more info." I looked at WP:AWB/TR at the time, and it didn't say anything about changing ((citation)) to ((cite)). This "fix" should be disabled if it's still there.
I wrote but decided not to post the following (off-wiki editing note of Tue Feb 10 12:10:28 PST 2015):
I decided reverting would be pointy so I kept using ((cite)) in new citations, leaving Yobot's ((citation))s in place for the old ones. But my general annoyance was memorable enough that I'm posting this as another example of robots (Yobot in this instance) interfering with human enjoyment of editing, plus being a bad specific edit imho. 50.0.136.56 (talk) 09:46, 18 January 2017 (UTC)
Note: I had no idea at the time of the above, that Yobot had been in controversy. I knew it was a very active bot but that was my first (and I think still only) direct encounter with it. 50.0.136.56 (talk) 09:51, 18 January 2017 (UTC)