Wikipedia:Bots/Noticeboard

Bots noticeboard

Here we coordinate and discuss Wikipedia issues related to bots and other programs interacting with the MediaWiki software. Bot operators are the main users of this noticeboard, but even if you are not one, your comments will be welcome. Just make sure you are aware about our bot policy and know where to post your issue. Do not post here if you came to discuss non-urgent bot issues, bugs and suggestions for improvement. Do that at the bot operator's talk page discuss urgent/major bot issues. Do that according to instructions at WP:BOTISSUE discuss general questions about the MediaWiki software and syntax. We have the village pump's technical section for that request approval for your new bot. Here is where you should do it request new functionality for bots. Share your ideas at the dedicated page

sboverride userright

How would I get User:GreenC bot the new sboverride userright? c.f. T36928 recently closed resolved. -- GreenC 20:12, 29 August 2022 (UTC)

Oh, that's a neat user right. User:AAlertBot could use it since it occasionally encounters urls users used that it cannot report and has to trim the report. I spent way too long fixing it when I first encountered this because I assumed bots would surely be exempt from this. I doubt there's any process yet for granting the right though. —  HELLKNOWZ ∣ TALK 21:06, 29 August 2022 (UTC)

You'd probably need to lobby for the permission to be added to an existing user group such as "bot", or for the creation of a new user group such as "sboverride". Any preference? –Novem Linguae (talk) 21:18, 29 August 2022 (UTC)

Probably should just get added to bot user group. Izno (talk) 21:27, 29 August 2022 (UTC)

The bot user group is "trusted" enough to have sboverride added to it imho — this is proposed at T313107 — TheresNoTime (talk • she/her) 21:28, 29 August 2022 (UTC)

I guess the question is if there is a case when it would be desirable to block a bot's edit due to the blacklist? —  HELLKNOWZ ∣ TALK 21:30, 29 August 2022 (UTC)

IIRC (and I may not), if I try to fix a typo in a section that contains a blacklisted URL, I can't save the edit, even if I am not editing near the URL. If that workflow still exists, it is frustrating. If bots can add blacklisted URLs but regular editors are then unable to edit the sections that contain those URLs, that would be undesirable IMO. If I am misdescribing or misremembering the workflow, or if I am misunderstanding this conversation, let me know in a nice way and I will strike this comment. – Jonesey95 (talk) 21:43, 29 August 2022 (UTC)

[...] the link filtering is based on what links existed before the edit vs. what links exist after (exist meaning interpreted as an external link by the software). Do you have any evidence that an edit that did not try to add a link was prevent by this extension? See the code - this part makes it so that if the page already existed, the links that are checked are only those that were added in the current end. --DannyS712 (talk) 00:18, 1 August 2022 (UTC)
— m:Requests for comment/Allow sysops to override the spam blacklist

So, it looks like you don't recall correctly. * Pppery * _{it has begun...} 21:52, 29 August 2022 (UTC)

Thanks for that; that's twice today that you have set me straight. Maybe I'm thinking of trying to revert vandalism, section blanking, or other undesirable edits and being stopped because I would be restoring a blacklisted link. I can't think of a situation where a bot would put a human editor in such a situation, so we're probably OK. – Jonesey95 (talk) 23:02, 29 August 2022 (UTC)

I guess the question is if there is a case when it would be desirable to block a bot's edit due to the blacklist? AnomieBOT's rescuing of orphaned references. It would probably be better if the bot didn't reinsert blacklisted links, but continue to complain on its talk page for humans to do a proper removal. Anomie⚔ 01:27, 30 August 2022 (UTC)

You could probably theoretical construct such a bot. But in general, I think whatever bots are doing, if it's an approved task, overrides those concerns.

I wouldn't let an AWB user overide the blacklist, but an AWB bot should be able to plow through. IMO. Headbomb {t · c · p · b} 15:19, 11 September 2022 (UTC)

Why don't we just create a dedicated sboverride group instead of speculating about whether there is some bot that might be harmed by having the right? * Pppery * _{it has begun...} 15:24, 11 September 2022 (UTC)

I could theoretically create a bot that has already existed for 14 years? Anomie⚔ 17:19, 11 September 2022 (UTC)

Bots, AWB, and 'crats

The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

---

There is a discussion at Wikipedia talk:Requests for permissions § AWB and bot access about the appropriate venue for requesting AWB access for bots. Your input is requested. Primefac (talk) 16:28, 8 January 2023 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

AnomieBOT Disruptive bot edits and dismissive operator

AnomieBOT (t · th · c · del · cross-wiki · SUL · edit counter · pages created (xtools · sigma) · non-automated edits · BLP edits · undos · manual reverts · rollbacks · logs ^{(blocks · rights · moves)} · rfar · spi · cci) (assign permissions)^{(acc · ap · ev · fm · mms · npr · pm · pc · rb · te)}

User:AnomieBOT has a pattern of "resucing" references by finding a similarly named reference anchor in the history of an article, or in some related article, and injecting it as a replacement for an undefined reference. Sometimes, this works fine. Other times, it intorudces a referenes that's completely irrelevant or inappropriate for the context. Of course, any reader, it just looks like a reference.

As far as I can tell, edits made by this bot aren't reviewed by any human. The bot churns out these edits mechanically, and they're unlikely to be caught unless someone happens along to check the "resuceing" edits the bot has made.

I've cataloged errors made by this bot and others, but received no response from pings in those notes. I have also directly reported these problems to the operator and received a very curt and dismissive reply, inviting me to -- myself! -- review the edits the bot was making.

For sure, the bot is trying to do something very aggressive because it's hard to verify its correctness. But because it is operating unsupervised and trusted to be helpful, it should be held to a high standard. If it can't guarantee its correct behavior, it shouldn't take action at all.

If the bot is making bad edits and the author isn't interested in addressing the issue or even directly monitoring the bot's actions, what can be done? -- Mikeblas (talk) 19:26, 15 January 2023 (UTC)

@Mikeblas the page you linked to shows rare, rather stale, edits. Is this bot currently making bad edits (i.e. an edit that would be reverted immediately if it were made by a non-bot editor)? Please provide a few very recent diffs. — xaosflux ^Talk 19:30, 15 January 2023 (UTC)

Operator notified of discussion. — xaosflux ^Talk 19:33, 15 January 2023 (UTC)

I've linked several pages. You must mean the catalog? "Directly reported" references an edit I fixed this morning, along with another that I brought to the operator's attention last month. This issue is ongoing and has been happening for at least a couple of years, which is part of the point. -- Mikeblas (talk) 19:35, 15 January 2023 (UTC)

@Mikeblas please provide a few recent diffs showing "disruptive" edits here so we are all on the same page. Don't need a big explanation right now, just a few diffs from this year will be fine. — xaosflux ^Talk 21:51, 15 January 2023 (UTC)

Here are the ones I linked above, explicitly copied here

• Here's one from Harry Styles: irrelevant reference injected
• From Anuradha Koirala: user erroneously edited reference name; AnomieBOT claims a fix, but just removed a space from the bogus text and copied bogus text to subquent invocation of that ref
• Again from Anuradha Koirala: disrupts reference name again ; AnomieBOT fixes by removing some spaces but error remains; user correctly removes bogus reference def
• From Iranian Revolution: user copies text with references (but not reference definitions) from another article; AnomieBOT arbitrarily copies an irrelevant reference previously deleted from this article; Proper fix gets reference from source of original copy

Here are some new ones:

• From Lewisville, Texas: user removes reference for 2019 stats, adds 2021 stats without references; AnomieBOT replaces the 2019 reference, which doesn't support the 2021 numbers
• At Lviv Oblast: user partially updates population and reference-generating template causing undefined ref error; AnomieBOT replaces missing reference with dead link to oudated stats from another article that doesn't support the content
• At List of best-selling albums of the 21st century in the United Kingdom: user mistakenly (?) removes several entries, references; AnomieBOT replaces only the reference to help the mistake (?) blend in
• At Steamboat Willie: user removes IMDB as a non-RS, but incompletely; AnomieBOT replaces it
• At Snoqualmie Valley School District: user updates stats with undefined source; AnomieBOT places reference for stats -- from a completely different school district
• At 2023 Nigerian Senate election: user tries to update, provides results without defining reference; AnomieBOT places reference from a different Nigerian state, no mention of this state and no support for this state's election results
• at 2023 24 Hours of Daytona: User tries to add driver and team, gives bad reference; AnomieBOT chooses a reference from another article that mentions neither that driver, nor this race. The user probably meant "TurnerBMWLineups".
• At Crystal Mountain, Michigan: user slams an update by changing stats, then renaming the reference anchor name; rescue just slams another use of the old name to the new name, so the new stats are falsely referenced and the old stats have an anachronistic anchor name
• At Indian Armed Forces: user forces an update with new stats but no references, just renames existing reference anchor; AnomieBOT finds another use of the old reference name and updates it to match the new one, even though it still refers to the old stat
• At Holmdel Township, New Jersey: User tries to update population and density, makes a mistake or two; AnomieBOT replaces partially deleted references with old links that don't support the 2019 population they claim to reference, probably should've deleted the invocation of the reference and the outdated values instead

-- Mikeblas (talk) 04:08, 16 January 2023 (UTC)

Also, thanks for notifying the operator. The header here says by following the steps outlined in WP:BOTISSUE, but I didn't see any specific steps (certainly not any kind of numbered list), and wasn't sure if notification was required or how it was required to be done. -- Mikeblas (talk) 19:37, 15 January 2023 (UTC)

I note Mikeblas has been fairly tendentious on this topic; for some time he was in the habit of edit-summary pinging me every time he reverted an edit by the bot that he did not like, even simple reversions of the bot's edit concurrent with reversion of IP vandalism. This is also the first I've heard of this "User:Mikeblas/Robots Behaving Badly" page, nor do I see any evidence that I was ever pinged on that page. Overall the tenor of complaints I've seen from this user suggests to me that they expect human-level AI from the bot, which is IMO an unreasonable request. The request here and at the bot's talk page to "stop the bot" or convert it to a semi-automated task similarly seems far beyond what is reasonable considering how often other editors have specifically thanked me or the bot for the service it provides.

On the other hand, I have been considering looking into the error rate related specifically to orphaned references rescued under Wikipedia:Bots/Requests for approval/AnomieBOT 6, to decide whether that specific portion of the task should be stopped or modified to also check for the source article containing text similar to adjacent text in the article being processed. I haven't yet found time to actually conduct this analysis or look into the feasibility of such modification. If anyone unbiased (e.g. not Mikeblas) feels like conducting that analysis, I'd appreciate the data. Anomie⚔ 20:01, 15 January 2023 (UTC)

I usually mention whomever's work I'm editing in edit summaries. This gives them the opportunity to review my revisions of their work and re-check my changes. I also think that Wikipedia will automatically notify a user when their contributions are reverted or undone.

My expectation for the bot (and any other, as well as any other user) is that they'll do what's possible to avoid making regressive or disruptive edits. Mistakes happen when people make edits, and we forgive those and try to repair them; but we do rely on policies to decide what to do when that behaviour isn't corrected or appears malicious. Bots are held to a higher standard because we expect them to do work for us, and correct them promptly if they're doing something wrong.

Aside from the inappropriate reference insertion I raise here, I often discover the bot is cleaning errors in vandalism edits rather than cleaning the vandalism itself. And so it only follows that the bot's edits are reverted along with the vandalism.

This bot is somehow (I don't know how) tasked with "rescuing" referencing errors. This task itself requires far more than simple algorithmic replacement or reaction, and instead demands that the bot know what a reference means in the context of the article, gauge its propriety, understand the previous editor's intention about editing or removal, choose an appropriate replacement (or other action), and evaluate: does it make the situation any better?, and so on. Nothing like fixing a date format or deleting hyphens, or even rebalancing parenthesis. The "rescuing" task is an ambitious order, but certainly is not something that I established or expected. OTOH, since the robot is trying to do it, anyone should expect it to do it correctly and reliably.

Fact is this bot often violates that standard and I've provided feedback to that effect. Perhaps worse, it's making edits that have subjective results (or, at least, unclearly defined outcomes) and its efficacy and safety aren't being monitored or checked. In what percentage of its edits is it being productive or disruptive? We don't objectively know. Maybe I'm wrong in that, but I don't think it's a tendentious position.

When pointing out these errors, I've met with excuses (just responding to "garbage input") and dismissal as in the current examples. That the bot's owner actively explains they're not listening to me (biased? because I've reported and documented problems?) and themselves won't monitor the bot at all unless someone else volunteers to do it indicates they're not a responsible bot operator.

Letting the bot make automated edits, dismissing and ignoring feedback, and "considering looking into" problems until some volunteer offers to own the task on their behalf doesn't seem like a responsible way to run a bot. Instead, the position should be to not release wide-spread active behaviour until that behaviour can be verified reasonably correct. Or, to set aside their own bias and face the realization (and feedback) that their goal is too ambitious to be completed reliably without human intervention or any monitoring of its outcome and instead abandon it. -- Mikeblas (talk) 20:57, 15 January 2023 (UTC)

You have just posted another 513 words of generic complaint without the diffs of recent problems that you were asked for above. Johnuniq (talk) 01:05, 16 January 2023 (UTC)

After reviewing the so called 'bad' edits, I can't say I see any problem with them. AnomieBot is functionally exactly as intended. E.g. someone removes the 'declared' instance of a reference that's reused multiple times in an article, causing issues like [1] without removal all references to it. AnomieBot then rescues the reference, solving the issue. This is neither disruptive no regressive. Headbomb {t · c · p · b} 05:45, 16 January 2023 (UTC)

That specific edit is a valid complaint, and is why I'm considering disabling or changing the AnomieBOT 6 task. The problem is that the generic reference name "Canada" was used for an unrelated reference in just one linked article; no idea where the IP got the reference name from in Special:Diff/1124983363. One of the other linked complaints is the same. OTOH, other of the complaints are less valid. For example Special:Diff/1102679930 did fix an error, just not in the "right" way; in Special:Diff/1133658542 Mikeblas would rather have had the orphaned copy removed instead of rescued; and in Special:Diff/1133439631 he'd rather the IP edit had been reverted (I guess). Anomie⚔ 13:22, 16 January 2023 (UTC)

I'll once again dubtifuly AGF in face of the sneering response and spell out the details:

• Special:Diff/1102679930 made an error message go away, but didn't fix or objectively improve anything. It's pretty obvious that the original root cause was the addition of article prose to the name of a reference anchor. Instead of making the prose visible, the bot codified the error by reusing the article prose in other invocations of the anchor: <ref name="Anuradha Koirala named CNN Hero of the Year 2010 In 1993, Maiti Nepal started with two rooms to protect women from abuse and trafficking. After establishing Maiti Nepal, Ms. Koirala plunged into the service of humanity. Her first work was setting up a home so that women and girls who have nowhere else to turn could find themselves a place to call theirs. After almost three decades today, Maiti Nepal has one Prevention Home, sixteen Transit Homes, two Women Rehabilitation Homes, one Child Protection Home, two Hospice Centers, one Information and Surveillance Center at Ministry of Foreign Affairs (MoFA) and a Formal School (Teresa Academy). More than 1000 children receive direct services from Maiti Nepal every day. All of these were possible because of Ms. Koirala’s firm determination and unprecedented leadership." />

Ignoring the more obvious cause, WP:ILCLUTTER explains that long anchor names aren't appropriate. Why would the bot enforce them? Now, a human must come along and clean up both the robot's meddling and the original error.

• In Special:Diff/1133439631, the anonymous edit was indeed unreferenced -- the new 2020 stats use a reference from 2010 that doesn't support their verifiability. AnomieBOT abetted the addition of unreferenced material by correcting the newly mismatched anchor name. Why is it acceptable to update objective statistics but not update their sources?
• In Special:Diff/1133658542 a user was trying to remove a reference to IMDB, which is a listed WP:USERGENERATED source. But AnomieBOT insisted on keeping the reference present and subverted them.

In every one of these cases, AnomieBOT demonstrates that it does not (because it largely can not) consider user intent or surrounding context when making edits. Instead of realistically acknowledging that it doesn't have a high probability of making a constructive edit, it instead brashly assumes it knows best and takes action without regard to the outcome.

I strenuously object to this pattern. (And, sorry for the awkward anthropomorphic wording. I just can't figure any better way to write it just now.) -- Mikeblas (talk) 16:37, 16 January 2023 (UTC)

In the Harry Styles example, the article claimed the Harry's House album was certified platinum by Music Canada, but used an undefined reference anchor name to do so. AnomieBOT inserted a new reference to an article about Amazon using Alexa brining its Prime Music service to Canada -- with no mention of Harry's House or its certifications, or even Music Canada itself.

Can you help me understand why you think mechanically adding completely irrelevant references is acceptable? -- Mikeblas (talk) 16:05, 16 January 2023 (UTC)

I have provided the requested examples and answered other questions, but haven't heard anything back. What are the next steps? -- Mikeblas (talk) 18:50, 21 January 2023 (UTC)

Walk away and find an issue that can actually be solved? No bot is perfect, and expecting perfection is unreasonable. Your primary concerns have been responded to, and now you are just nitpicking. Primefac (talk) 19:25, 21 January 2023 (UTC)

It's hard for me to see either the injection of irrelevant references or the automated undoing of user-intended beneficial edits as "nitpicking". 130-word reference anchor names aren't any kind of "fix".

The solution is easy: if a bot can't make edits reliably, then it shouldn't attempt them ... particularly when no human is actively monitoring the edits to try and improve the behaviour of the bot.

Let's try it another why: why is your position is that the status quo can't be improved upon at all, and asking for improvement is to be met with outright dismissal? Why isn't it important to conisder that bots could be made better -- both more accurate and less invasive? -- Mikeblas (talk) 23:31, 21 January 2023 (UTC)

Bots can be made better - to a point. Errors, especially when dealing with this particular subject, are inevitable. Most of the examples you give in the "Here are some new ones" list are GIGO situations; predicting, let alone fixing, these types of errors is unreasonable. Additionally, the point of this bot is not to evaluate orphaned references, it is to fix them, so the argument that "restoring non-RS isn't acceptable" is also unreasonable (specifically w.r.t. the IMDb ref you mention). The "horribly long named reference" issue also falls under that umbrella: it's not the bot's job to evaluate the names of refs.

Out of curiosity I looked at the bot's contributions for the period you pulled your examples from (13-15 January). Out of 172 orphaned references that were fixed, you listed 10, which is a 5.8% failure rate, which to me is perfectly acceptable for a task such as this.

So yes, the status quo can be improved, to a point. Asking for improvement is acceptable, to a point. Telling a bot operator to turn off the robot until its correctness can be verified (source) when it is already operating at a 94.2% accuracy rating is not acceptable. To paraphrase/build on what Anomie said on the bot's talk page - if there is a specific or often-repeated issue that needs to be looked at, that's one thing, but "this bot isn't perfect so shut it down" is not. Primefac (talk) 13:19, 22 January 2023 (UTC)

It's hard to overcome this defeatist attitude -- that the problem can't be fixed or the errors are inevitable or that these issues are "just GIGO". They aren't unfixable, obviously: turning off the bot stops the errors completely. Of course, the bot does do desirable changes and does them reliably. My suggestion is to turn off the unpredictable parts until they can be made acceptably reliable.

GIGO is a cop-out. Properly-programmed computers evaluate their input. In the face of invalid input, good software raises an error and does not take further action. If we truly think these are GIGO cases, then we're describing a bot that either doesn't care to validate its input and acts regardless. Or are we describing a bot that does evaluates its input, finds it invalid and still acts anyway? A great improvement would be to do away with the unconditional action and instead either evaluate further or quit trying to do the impossible when it's so likely to be a debatable change.

Here, I was asked to provide "some a few recent diffs", and I did. I did not set out to comprehensively evaluate all the edits made in a particular period. Had I tried to, I'd undoubtedly find more problems than I reported and would've driven your computed error rate higher. Note that nobody -- and most irresponsibly, the bot operator -- hasn't evaluated the accuracy rate either and that's another thing that I suggest be done.

But why would you deliberately misapply my list in order to compute statistics in this way?

I also wonder: why is it your position that any bot should be free to run before its correctness can be verified? That seems contrary to anything we know about automation in IT. Yet from the same talk page, the bot's owner {{diff2||1125613668|1125568193|dismissively invited me to monitor the bot for problems}} rather than themselves consider any solution or monitor their bot's edit themselves. -- Mikeblas (talk) 04:32, 23 January 2023 (UTC)

Do you even know how BRFA or bots operate? Every bot task gets evaluated before it is approved, and you can't just "turn off the unpredictable parts" of a bot's code. Code can be updated if something specific is found to be a problem (see for example my Task 17 and its code updates), but not everything can be planned for or prevented. As I believe multiple BAG members have said in this discussion - if you find a repeated, often-problematic bug, it can likely be fixed, but 100% is simply not feasible, and 90-95% is a pretty decent target to shoot for. Additionally, Anomie has said they will look into some of the coding in the Task 6 expansion. Stop expecting bot operators to bend over backwards to fix these edge cases and you might actually get somewhere; flies, honey, etc. Primefac (talk) 09:37, 23 January 2023 (UTC)

These aren't edge cases, and I've provided plenty of examples. Is the code for AnomieBOT publicly available?

Meanwhile, you've decided to ignore the inconvenient questions I've asked you. -- Mikeblas (talk) 14:31, 23 January 2023 (UTC)

Feel free to ask it again if I've missed a question that still needs a reply, instead of playing coy. Primefac (talk) 14:33, 23 January 2023 (UTC)

Nothing coy here: the questions are above and you know which you didn't answer. To me, the most intersting ones are about your misinterpretation of my list of issues as an exhaustive evaluation; and about your expectation that bots can run before they're demonstrated correct -- and further that raising concerns and evidence about bots making errors "is unacceptable" -- are outstanding.

You also didn't answer my question about the availability of source code. But I found it anyway!

After reviewing the code for just a few minutes: Looks like OrphanReferenceFixer is the "task" that implements some of the dubious edits. I'm not much of a perl guy, but it seems like the bot starts in bot-instance.pl by reading all the tasks in the supplied tasks directory.

Seems like we could, then, remove this task just by removing its package code from the tasks directory. Or, by adding code in the enumeration of that directory that finds that file and skips creating the task object for it.

Looks like each task has an "approved" method that returns -- well, maybe there's some way to modify "approved" for the OrphanReferenceFixer task to have it not run without actually removing the code. But I couldn't find code that uses the "approved" value in a task. (Maybe that's because I'm surfing a bunch of web pages instead of browsing a directory full of files, so searching isn't so facile.) Probably could also disable a single task by returning failure from its init() method.

Maybe we want to keep the ORF task but modify its behaviour more granularly. Within the task, it's pretty easy to see that more direct replacements are done with a series of regex replacements in process_page. After the regexes are applied, I don't see any checks that test the results of the change. That is, anything that trips up the regex is going to produce undesirable output, and the code won't notice. This is how we end up with reference anchor names that are 600-something characters long.

This code could check for new errors after its changes are applied and abort its intended change if new errors appear. Such a check would help make the code resilient against input that it wasn't prepared to work. That is, it's at this point in the code where the "it's just GIGO" argument falls over: the code itself doesn't test the validity of its input, or of its own output. GIGO is avoidable.

These regex substitutions could be simple, but since they're implemented with regexes they're quite involved.[2] But more complicated is the reference copying code -- which we find is sometimes injecting irrelevant references and unconditionally claiming victory. This behaviour could be selectively disabled by not executing the code starting at the loop commented # Any orphaned refs?.

And so I think your claim that we can't just "turn off the unpredictable parts" of a bot's code is entirely false. -- Mikeblas (talk) 15:26, 23 January 2023 (UTC)

If I had known which question I hadn't answered, I would have answered it, and you still haven't told me so I am done with this farcical conversation. Ask for a question and get eight paragraphs of response... jeez. Primefac (talk) 19:39, 23 January 2023 (UTC)

I directly answered: To me, the most intersting ones are about your misinterpretation of my list of issues as an exhaustive evaluation; and about your expectation that bots can run before they're demonstrated correct -- and further that raising concerns and evidence about bots making errors "is unacceptable" -- are outstanding. -- Mikeblas (talk) 23:52, 24 January 2023 (UTC)

What is the error rate, how bad is it? No one expects 100% perfection due to the great benefit of the bot, but at the same time we might expect at least 95% accuracy or something and be quite concerned if it was 80% because the 80/20 Rule suggests the first 80% are easy and will be error free, the remaining 20% are very hard - does it at least get beyond the 80/20 Rule? IMO any bot that is 80% accurate should not be running. -- GreenC 14:38, 22 January 2023 (UTC)

That's part of the problem -- we don't know, since nobody is monitoring the bot for correctness. This bot is very active, so manually tallying its actions would be quite a chore. The "some new ones" list I present above spans about 25 hours, but I certainly did not exhaustively examine every edit the bot made in that time span. -- Mikeblas (talk) 04:04, 23 January 2023 (UTC)

I don't have the time to read all the above, but I skimmed it and have one driveby comment: It seems to me that an article which this bot task has to edit already contains an error, the orphaned reference, or some very weird wikitext which confuses the bot. Without the bot's edits, far more articles would contain errors. Some of the nitpicks, like internal reference names containing vandalism obscured by bot edit, are far less problematic than that. It's a task that's been running for over a decade with success, I'm not convinced a strong case has been made here for the approval to be revoked or modified. ProcrastinatingReader (talk) 15:51, 23 January 2023 (UTC)

I think people would like to believe that, but there's no collected evidence to support it. That is, I don't think there's a strong case to comprehensively claim these tasks are running "with success". Meanwhile, not thoroughly reading the evidence can't mean you're upholding your duties as a BAG member in addressing concerns raised about a bot and its operator. Doesn't your voicing a decision about the same mean that you've ignored the presented facts and are acting on bias instead of reason? -- Mikeblas (talk) 23:51, 24 January 2023 (UTC)

Doesn't your voicing a decision about the same mean that you've ignored the presented facts and are acting on bias instead of reason? Or discarding them because they don't think they're sufficiently valid to be of concern. Please try to avoid presenting false dichotomies.

At this point, with multiple BAG members, including the bot author, responding in the negative to changing how the bot functions in the way you would prefer, you should consider an WP:RFC. Otherwise, you should drop the stick. I have some doubt an RFC would resolve in the way you would prefer also, but it does remain an option available, and who knows, maybe people would indeed prefer that the ref-saving task be turned off. Izno (talk) 00:25, 25 January 2023 (UTC)

Inactive bots

Moved from Wikipedia:Bureaucrats' noticeboard

(copied from moved-to header)The bot policy asks for this to be handled, with some notifications, over at WP:BOTN, so moved there. Once done if a crat doesn't handle it there, the removal request can be placed here. — xaosflux ^Talk 10:38, 20 January 2023 (UTC)

Hello. Per Wikipedia:Bot policy#Activity requirements, we're supposed to be de-flagging bots that haven't been active in two years. xaosflux has this helpful query showing about 150 bots that meet the criteria currently. I dug up a query after noticing that LaraBot still has a bot flag even though that bot hasn't edited since 2014. Can someone please do the necessary notifications and de-flaggings for these inactive bots?

For LaraBot and BernsteinBot, you can just go ahead and remove the bot flags immediately. --MZMcBride (talk) 07:09, 20 January 2023 (UTC)

The criteria also needs that the listed operator has also had no logged actions or edits for two years. The query is based only on bot activity, not operator, so most of the 150 in the result do not meet the criteria for removal. I checked about 20 and all of them had active operators. ಮಲ್ನಾಡಾಚ್ ಕೊಂಕ್ಣೊ (talk) 08:02, 20 January 2023 (UTC)

Oh, that's stupid. Why would operator activity make a difference? I think account activity requirements are a bit daft in general, but keeping bot flags on inactive bots because their operators are still around is particularly puzzling. We should change the policy. --MZMcBride (talk) 09:09, 20 January 2023 (UTC)

I keep a bot account that I use for one-offs and it's simply easier to flip the bit once and keep it flipped than request it possibly multiple times for arbitrary runs. Anyway, I'm starting a policy discussion shortly on the topic of inactivity that targets what I think are low-hanging fruit in the report. Izno (talk) 09:19, 20 January 2023 (UTC)

Nobody is suggesting de-flagging your bot. There's a clear and substantial difference between a bot such as IznoBot that last edited in September 2021 and a bot such as Stwalkerbot that last edited in June 2010. We both know this. :-) --MZMcBride (talk) 09:29, 20 January 2023 (UTC)

Just a note for the record that Izno's discussion is at Wikipedia talk:Bot policy § Bot and operator inactivity - blocks. Primefac (talk) 13:07, 20 January 2023 (UTC)

@MZMcBride: Why would operator activity make a difference? If the operator is around, and there's an issue with the bot account, an active bot operator can take action.

This doesn't mean every other bot should keep its flag, it's just that by policy, automatic flag removal is unwarranted. Stwalkerbot is a bot that's both inactive and by the admission of its operator, unlikely to resume. That makes it a good candidate for a flag removal IMO. Headbomb {t · c · p · b} 21:01, 20 January 2023 (UTC)

Wikipedia:Bots/Noticeboard/Archive 16#Inactive bots - February 2022 was the last go around. The report of interest is User:MajavahBot/Bot status report. Izno (talk) 08:24, 20 January 2023 (UTC)

Below are the inactive ones by the definition (TAP bot is a couple days from now). Izno (talk) 08:38, 20 January 2023 (UTC)

Bot account	Operator(s)	Total edits	Last activity (UTC)	Last edit (UTC)	Last logged action (UTC)	Last operator activity (UTC)	Extra details
PDFbot	Dispenser	7943	12 Feb 2012	12 Feb 2012	04 Feb 2007	15 Mar 2020
BG19bot	Bgwhite	1005055	09 Feb 2017	09 Feb 2017	08 Feb 2017	15 Mar 2020
Makecat-bot	Makecat	103877	05 Apr 2013	05 Apr 2013	10 Feb 2013	28 Mar 2020
Lonjers french region rename bot	Lonjers	11910	15 Mar 2016	15 Mar 2016	—	19 May 2020
TAP Bot	Thine Antique Pen	1920	14 Sep 2015	14 Sep 2015	13 Dec 2014	24 Jan 2021

• Moved from WP:BN. Will send the operator notices. — xaosflux ^Talk 10:38, 20 January 2023 (UTC)

Operator notices sent to the 5 operators listed above. — xaosflux ^Talk 10:41, 20 January 2023 (UTC)

---

• As there has been no response form the operators, these bots have been deauthorized and deflagged. Should the operator want to use them again, they will need to request it at WP:BRFA. — xaosflux ^Talk 00:46, 1 February 2023 (UTC)

  I noticed that BG19bot and Makecat-bot are now showing up on Wikipedia:List of Wikipedians by number of edits/1–1000 - see this edit. Is there a way to resolve this? Thanks! GoingBatty (talk) 05:46, 1 February 2023 (UTC)

  @GoingBatty: I have added them to Wikipedia:List of Wikipedians by number of edits/Unflagged bots, which will remove them from the list in the next edit. ಮಲ್ನಾಡಾಚ್ ಕೊಂಕ್ಣೊ (talk) 12:03, 1 February 2023 (UTC)

Remove bot flag from two bots

LaraBot (t · th · c · del · cross-wiki · SUL · edit counter · pages created (xtools · sigma) · non-automated edits · BLP edits · undos · manual reverts · rollbacks · logs ^{(blocks · rights · moves)} · rfar · spi · cci) (assign permissions)^{(acc · ap · ev · fm · mms · npr · pm · pc · rb · te)}

BernsteinBot (t · th · c · del · cross-wiki · SUL · edit counter · pages created (xtools · sigma) · non-automated edits · BLP edits · undos · manual reverts · rollbacks · logs ^{(blocks · rights · moves)} · rfar · spi · cci) (assign permissions)^{(acc · ap · ev · fm · mms · npr · pm · pc · rb · te)}

Please remove the bot flags from LaraBot and BernsteinBot. I don't know what the passwords for these accounts are off-hand and I have no intention of using them again. We should independently fix the bot policy, but my request got a bit buried by me noticing that the bot policy is goofy and not in alignment with actual practice. --MZMcBride (talk) 16:01, 20 January 2023 (UTC)

 Done — xaosflux ^Talk 16:49, 20 January 2023 (UTC)

Another bot flag removal request

WelcomerBot (t · th · c · del · cross-wiki · SUL · edit counter · pages created (xtools · sigma) · non-automated edits · BLP edits · undos · manual reverts · rollbacks · logs ^{(blocks · rights · moves)} · rfar · spi · cci) (assign permissions)^{(acc · ap · ev · fm · mms · npr · pm · pc · rb · te)}

As prompted by Izno, please can you remove the bot flag from User:WelcomerBot? The functionality of the bot has been completely replaced within the ACC tool and edits are now done via OAuth instead stwalkerster (talk) 23:31, 20 January 2023 (UTC)

 Done thanks for the note @Stwalkerster:. — xaosflux ^Talk 23:58, 20 January 2023 (UTC)

User:MalnadachBot is running amok

The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

---

The bot User:MalnadachBot is currently going crazy, making edits to thousands (?) of talk pages with completely useless edits to decades-old user signatures. The signatures are completely fine and do not need to be changed. The "font" tags they use work in every browser ever and will continue to work in every browser forever into the future. If font tags are causing someone's linter to complain, then either (a) the linter should not be run on talk pages, or (b) that should be considered a bug in the linter, and fixed.

This bot should be stopped to stop it from spamming everyone's watchlists, and should not be allowed to run again until it can be shown it won't make so many worthless no-op spam edits.

After hitting the E-stop on this one, one thing the bot author may want to do is configure the bot to only examine main namespace. –jacobolus (t) 00:37, 1 February 2023 (UTC)

@Jacobolus that bot appears to properly asserting the 'bot' flag on edits, you can avoid seeing bots on your watchlist by checking hide next to "bots" on the filter. — xaosflux ^Talk 00:42, 1 February 2023 (UTC)

There are plenty of good reasons to keep an eye on bot edits. Telling people they should stop paying attention to meaningful edits if they don’t want to sift through the massive piles of no-op spam being created by a different bot is not an adequate response. The edits being done by User:MalnadachBot are literally 100% worthless. This is not some kind of important maintenance task. It’s pure distraction and make-work. We are talking people's signatures on talk page comments from 15 years ago. There is no reason whatsoever to care whether these pass someone's made up linter rules. –jacobolus (t) 00:45, 1 February 2023 (UTC)

@Jacobolus have you discussed your concerns with the bot operator already and are at an impasse? — xaosflux ^Talk 00:52, 1 February 2023 (UTC)

The only response is "this is an approved bot" and "you can hide [the edits from your watchlist]". Which is not the response I am looking for of: "I will shut the bot down and stop spamming everyone".

I’m bringing it here for hopefully more eyes, including maybe from someone with some better perspective or authority. Whatever process was involved in "approving" this might also be broken. I don't really know too much about Wikipedia bot policies. And I also would really rather not wade into a big fight about it more generally.

I just want the bot to stop. The current behavior is insane. –jacobolus (t) 00:55, 1 February 2023 (UTC)

Thanks for reporting this. I have over 16,000 items in my watchlist, and this was annoying me too. I have checked the no-bots box, but I'd rather not have to do it for too long. So if this is all about something optional that doesn't need fixing in the first place, then I too would like to see that stopped ASAP. BilCat (talk) 04:13, 1 February 2023 (UTC)

• Operator notified. This discussion appears to primarily be about Task 12. — xaosflux ^Talk 01:24, 1 February 2023 (UTC)
• Unlike what the heading suggests, this section isn't based on any malfunction or policy issues, you are demanding that we stop working on replacing erroneous and obsolete markup because you don't like it. These are safe edits designed to replace all font tags in one go, the bot has replaced over half a million font tags in 4 days. There are multiple options to hide them in your watchlist, yet if you don't want to use them you will have to put up with it. The only thing that matters in regard to how mediawiki sofware works and its future plans is what the developers who maintain the software say. And they have clearly done so by marking them for replacement, random opinions not based on policy has no bearing on how mediawiki works. Mine is hardly the only bot working on Lint errors. ಮಲ್ನಾಡಾಚ್ ಕೊಂಕ್ಣೊ (talk) 05:22, 1 February 2023 (UTC)

  replacing erroneous and obsolete markup There is literally no benefit whatsoever in doing this. These html tags have worked for the past 20+ years and are going to continue to work forever. No browser is ever going to drop support. If you linter is giving you a hard time about "deprecated" tags on historical talk pages, that's a linter bug that should be fixed (or better still, you should stop trying to run a linter on 15-year-old discussions).

  If you want to go tell current users to stop using the font tag in their current signature, fine. If you want to fix linter errors in main namespace, it's probably unnecessary busywork for you, but fine.

  But trying to force every bit of historical talk page markup to match new made-up standards based on some kind of pointless crusade with no technical merit is a huge waste of everyone's time and attention, and is grossly disrespectful of everyone else on the wiki.

  how mediawiki sofware works this has nothing at all to do with “how mediawiki software works”. –jacobolus (t) 05:51, 1 February 2023 (UTC)

  pointless crusade with no technical merit describes your comment perfectly. My bot (as well as others' Lint fixing bots) has sound technical merit. If you want to propose changes to how mediawiki software should work, do so at Phabricator instead of railing against users working on known issues. ಮಲ್ನಾಡಾಚ್ ಕೊಂಕ್ಣೊ (talk) 06:01, 1 February 2023 (UTC)

  My complaint is a social one, not a technical one. You are spamming everyone for literally no benefit. I have the same complaint when a telemarketer calls me in the middle of dinner or when my email inbox fills up with viagra ads and Nigerian prince scammer solicitations. –jacobolus (t) 06:09, 1 February 2023 (UTC)

  Please tone down the rhetoric, it's entirely unnecessary. Legoktm (talk) 06:31, 1 February 2023 (UTC)

  There has been no reason articulated for why these edits should happen at all, besides "it's a linter error". Why should anyone care if there's a linter error on old talk pages? Who is "linting" those pages and why? What problem does it cause if there are linter errors there? No attempt has been made to answer any of those questions.

  There was as far as I can tell no community discussion or consensus that millions of old talk page comments needed to be modified. It was just unilaterally decided by a tiny number of people who went ahead and started making millions of no-op edits, polluting everyone's watchlist.

  When asked to explain, they don't make any attempt at an answer, or even indicate that they understand the complaint, but instead hide behind bureucratic "this was already approved". Which again, does not answer any of the relevant questions. –jacobolus (t) 06:48, 1 February 2023 (UTC)

  Font tags are deprecated according to MDN docs. 0xDeadbeef→∞ (talk to me) 06:05, 1 February 2023 (UTC)

  Yes, but "deprecated" here is a term of art that in the browser/html world means "has worked in every browser for decades (cf. browser compatibility table) and will continue to be supported until the end of time, but one spec editor one time wanted to discourage new uses". This is not something that anyone needs to bother "fixing" on ancient talk page discussion signatures. Imagine if even in the worst case, some new VR-based browser in 2050 drops support for the deprecated "font" tag. Suddenly archived Wikipedia talk page discussions from 2007 will have their signatures revert to default colors of Wikipedia-VR-skin-2045 instead of the originally-intended-in-2007 rainbow effect. Is this really a pressing problem anyone should care about? –jacobolus (t) 06:06, 1 February 2023 (UTC)

  (As an aside, "deprecated" in browser specs also occasionally means "was a tricky new experimental feature that was never widely supported across browsers and has already been replaced by something better, so is likely to break in the future"; but that's an entirely different situation vs. tags like center, font, or tt, which are used on billions of web pages and can never be removed from implementations.) –jacobolus (t) 06:15, 1 February 2023 (UTC)

  For what it's worth, this was previously discussed this last June. Nothing has really changed since then, aside from the main complaint back then of making multiple edits per page being fixed AIUI (thank you). Legoktm (talk) 06:24, 1 February 2023 (UTC)

  It looks like the main complaint back then was the annoyance of sifting through spam from useless no-op edits. Which is exactly the same as the complaint today. –jacobolus (t) 06:26, 1 February 2023 (UTC)

  The main complaint was repeated visits, with each visit fixing just one or a couple of individual signatures. This complaint has been addressed by the bot operator. The bot now skips editing a page entirely if it is unable to fix all of the obsolete font tags in a single edit. – Jonesey95 (talk) 06:34, 1 February 2023 (UTC)

  This is a mischaracterization:

  • “I'd like to ask that the MalnadachBot be halted or its impacts on watchlists be removed... We don't allow cosmetic edits for a reason.”
  • “Are we certain the errors need to be fixed?”
  • “I'm really unclear on why fixing these lint errors in old discussions is worthwhile. Is there a pointer to a discussion on this?”
  • “Do we really need to fix Lint errors on pages from 10+ years ago that nobody reads?”
  • “If either browsers or MediaWiki dropped support, then we could go about updating the old HTML, right? There's no reason to think support will be dropped anytime soon?”

  (none of these questions were ever addressed) –jacobolus (t) 07:01, 1 February 2023 (UTC)

  @Jacobolus: See also WP:HIDEBOT to hide a specific bot from your watchlist. 0xDeadbeef→∞ (talk to me) 06:35, 1 February 2023 (UTC)

  Thanks for that. I've asked about that before (can't remember who or where) within the last couple of years, and was told it wasn't possible (at least then). I'll definitely try it out on several annoying note (and users too!) BilCat (talk) 06:52, 1 February 2023 (UTC)

  has replaced over half a million font tags in 4 days I am curious whether people would complain less if it ran slower, as there really is no urgency here. Legoktm (talk) 06:36, 1 February 2023 (UTC)

  Last 4 days I was running it on pages with most errors. Previously when I ran slower, people complained that it has been going on for a long time, someone or the other will be unhappy regardless of pace. ಮಲ್ನಾಡಾಚ್ ಕೊಂಕ್ಣೊ (talk) 07:20, 1 February 2023 (UTC)

  Legoktm asks a reasonable question. The bot currently edits roughly 10,000 pages per day, which I think is a typical bot pace (about 10 edits per minute). There are currently 8 million Linter errors, reduced from 21 million thanks primarily to this bot. If we estimate that something like 6 million of those errors are fixable by bot, and each edit fixes one error (unlikely, but the worst-case scenario), we can have 10,000 edits per day for 60 days, or 1,000 edits per day for 600 days, or some other trade-off. For most bots, we put up with a fast pace because they are only doing a few thousand edits, or at most a few days of edits, at that pace. The scale of this cleanup is different, and it may be worth a community conversation. – Jonesey95 (talk) 07:33, 1 February 2023 (UTC)

Style guidance for bot User pages?

As an unrelated aside, is there any kind of CSS style guidance for the about pages of active bots? I find this one has an almost entirely illegible (to me) page. It looks like this in my browser:

–jacobolus (t) 07:29, 1 February 2023 (UTC)

As I have said in my talkpage, Jacobolus was involved in a recent dispute with me, casting aspersions and editing disruptively following which they were briefly blocked, as seen here and here. They are bringing this up hoping something will stick, after reading the page and going through the links that has all the explanations they are demanding here. Everybody is allowed freedom with styling their userspace. However if any uninvolved editor has a problem with it, I will replace it with another font. ಮಲ್ನಾಡಾಚ್ ಕೊಂಕ್ಣೊ (talk) 08:01, 1 February 2023 (UTC)

I was annoyed at you in October for pointlessly deleting a few templates that I had put a lot of effort into a decade ago without offering any reasonable justification for it, hiding behind bureaucratic legalese when called out, and refusing to explain yourself or engage in discussion. I had forgotten about that whole episode by now, because I decided it wasn't worth the considerable hassle and community organizing effort it would take to try to fix the currently broken and abusive "templates for deletion" process. But that is all entirely off topic here.

I am annoyed at your bot now in February because it is filling up my watchlist (and everyone else's) with huge numbers of no-op "fix linter error" edits to 15-year-old talk page discussions which do not need to be made.

I really couldn't care less what font you use on your user page. But if your bot is going to put a link to an explanatory page in every one of its millions of edit summaries, you might as well try to make the page legible to people who click through, no? (While you are at it you might expand the explanation on the page to answer common questions such as "why does this need to be done at all?" and "where is the community consensus supporting this change?") –jacobolus (t) 08:11, 1 February 2023 (UTC)

@ಮಲ್ನಾಡಾಚ್ ಕೊಂಕ್ಣೊ, this is indeed hard to read. — Qwerfjkltalk 08:22, 1 February 2023 (UTC)

@Qwerfjkl: Thanks, I have removed Brush Script MT font. ಮಲ್ನಾಡಾಚ್ ಕೊಂಕ್ಣೊ (talk) 08:33, 1 February 2023 (UTC)

Well, it's pretty rare for people to come to this noticeboard to talk about a bot when they're happy about something it did, but I am pretty glad that all those goddamn lint errors in subpages of WP:SIGNPOST have been fixed -- manually fixing that would have been a gigantic pain. jp×g 11:17, 1 February 2023 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

Bots operated by Cyberpower678

Cyberpower678 might be on a little R&R break from Wikipedia, as they are not responding. Meanwhile, his bots seem to not be functioning as they should. I know the admin stats have not updated for a while. And apparently other tasks are stalled out. Please see User talk:cyberpower678 and its recent archives. Is there an alternate admin who can get things back up and running? If not, I guess we just wait it out. — Maile (talk) 01:16, 1 February 2023 (UTC)