Wikipedia talk:Bot Approvals Group

From Wikipedia, the free encyclopedia
  (Redirected from Wikipedia talk:BAG)
Jump to: navigation, search

Requests for BAG membership[edit]

Requests to join the Bot Approvals Group are currently made here, although other methods have been proposed. Users wishing to join BAG, or to nominate another user to become a member, should start a new nomination page via the form below (replacing "UserName" with the nominee's) and transclude the discussion in a section below. Please note that notification to WP:AN, WP:VPM, WT:BOT, and WP:BON is required. After a suitable length of time (usually one week unless the nomination has not received a reasonable level of support), the discussion will be closed by a bureaucrat.


Other discussion[edit]

Bot unblocking policy[edit]

Duplicate: Discussion already happening at Wikipedia_talk:Bot_policy#Changes_to_.27dealing_with_issues.27_section --slakrtalk / 07:09, 10 February 2017 (UTC)

I initially made this suggestions at a ARB case, but was advised to bring it here. Basically: If a non-critical bot is blocked for technical malfunction, the bot operator must request its unblock here, and the BAG must have consensus that the issue has been fixed. A critical bot would be defined as one that impacts the English Wikipedia's running as a whole, examples would be Cluebot NG, but not any of the other Cluebots. Thoughts? Iazyges Consermonor Opus meum 16:02, 7 February 2017 (UTC)

Some questions/issues: What is a (non-)critical bot, who decides this (bots are almost all external, so none are truly critical to Wikipedia's running; Wikipedia ran fine before Cluebot NG, if less efficiently)? Where is "here" for unblock requests (this being BAG page), would it be WP:BON or some new page? What constitutes BAG consensus (since approvals need only 1 BAG member), does BAG need to sign-off or actively verify it? How does this affect existing WP:BOTISSUE practice? Who unblocks the bot -- a closing admin, BAG member, whoever? What is "technical malfunction" and who decides this -- ranging from an obvious bug in code to simply a task not done to human standards? If it's really just a technical issue (i.e. bug), why do we need consensus -- surely bugs happen all the time? What if it's a one-time issue due to its nature? How would BAG determine the bot is fixed if it's blocked -- do we unblock for a trail, do we ask for code? How often do you expect these requests, as there are probably at least several bot issues daily (BAG is swamped at WP:BRFA, not sure how long these will take)? I think this idea needs a lot of details ironed out before bringing it to a proposal. Personally, I'm not sure BAG has the manpower to do this effectively (depending on how many blocks need this unblocking). —  HELLKNOWZ  ▎TALK 17:40, 7 February 2017 (UTC)
I don't understand why the unblock procedure should be different for bots. Neither I understand the definition of non-critical. I would like also to underline the fact that many bots allow themselves to be stopped without a block. -- Magioladitis (talk) 17:51, 7 February 2017 (UTC)
I don't agree with this proposal as-is. WP:BON would be the place to discuss if needed - however I think that should only be if there is disagreement with the blocking admin. In any case, I think the blocking admin should be able to unblock if they are satisfied. — xaosflux Talk 17:55, 7 February 2017 (UTC)
@Iazyges: That's never been the policy, and personally I believe this is overly bureaucratic. Unblocks for technical malfunctions are routine, and having BAG review every instance would add a huge burden to its mandate. That's why we have BRFA that evaluates among other things not only the technical merits of the bot, but also the compliance of bot operators with WP:BOTCOMM and other aspects of the policies so that once bots are unleashed, they are trusted to fix technical issues as they arise and maintain the bot's code to be technically sound, and in line with changes in consensus.
However, you can take a look at this discussion, which in part aims to clarify guidance on how do deal with malfunctioning bots, and when malfunctioning bots may be unblocked. Headbomb {talk / contribs / physics / books} 19:21, 7 February 2017 (UTC)
I don't know why someone told you to come here when WT:BOTPOL would have been a better place. Especially since there's already discussion on this topic started at WT:BOTPOL#Changes to 'dealing with issues' section. Anomie 00:47, 8 February 2017 (UTC)
The policy to discuss with the blocking admin or post an unblock request works fine as-is. No need to reinvent the wheel here. ~ Rob13Talk 03:12, 9 February 2017 (UTC)

Ramp up and BAG Trial & Closure Guide[edit]

Two things.

The first is I took a look at Wikipedia:Bots/Requests for approval/JJMC89 bot 7 and I think xaosflux (talk · contribs) did something quite interesting with the closure, as approving a bot with a ramp up deployment. I think this is something BAG ought to do more often, especially in the case of bots that will make several edits, in potential tricky areas (technically tricky, or in areas where historically there's been issues with similar bots).

The second is I think we should probably write a BAG Guide detailing best practices and guidelines for bot trials and the closing of BRFA. I'll take a stab at a draft over the weekend if no one else does it, but I encourage other to give their ideas in the meantime (or even start the guide).

Headbomb {talk / contribs / physics / books} 11:55, 10 February 2017 (UTC)

I am very much a favor of ramp up schedules, and have used them on runs with very large initial edit expectations - some problems only present over larger scales and getting complaints from other editors is a final safety valve. — xaosflux Talk 12:16, 10 February 2017 (UTC)
Personally, I would have done the ramp-up as part of the trial (requesting flag, if needed) and approve it after its done. I don't think I like trial-like ramp-ups after the bot is approved (since BRFA page is the logical place to discuss any issues, but approval closes it). But then again I am not really against them, if they make sense and the time period is lengthy. —  HELLKNOWZ  ▎TALK 13:52, 10 February 2017 (UTC)
I agree with Hellknowz on this one, if the point is to get feedback then it should be part of the trial rather than a post-trial thing. Anomie 14:31, 10 February 2017 (UTC)
Agree, on the largest ones I've done (say an job for 100,000 pages+) I usually give them a short demonstration trial (25-50 edits), then extended trials (100-1000 edits depending on the type of job and if there have been any issues) - extended trials are repeated if errors came up that needed to be retested. I only occasionally use the ramp up go-live - I've never had an operator complain about. — xaosflux Talk 18:47, 10 February 2017 (UTC)
What we really need is a safety valve of some sort which provides oversight past an approval. Sometimes, complaints won't be registered until the bot edits are in operation, even with large trials. This could be done a few ways:
  1. Allow BAG members to unilaterally rescind their own approvals or require alterations of the task/additional trials/additional consensus-seeking in the event of a complaint filed at WP:BON.
  2. Allow any uninvolved administrator to require a bot operator to cease operating a task until a BAG member reviews a particular complaint or issue if, in the opinion of the admin, that complaint or issue was not brought up and addressed during the BRFA.
  3. Require bot operators to cease operating a task until a BAG member reviews a particular complaint or issue if any three extendedconfirmed editors complain about a given task.
I'm sure there's more options as well, but basically, our current system of no oversight after approval isn't working. We need something more. ~ Rob13Talk 19:08, 10 February 2017 (UTC)
Option #1 isn't too bad. Option #2 would probably lead to lots of arguments about whether a particular admin is actually uninvolved. #3 doesn't seem that good to me either, it seems to be going beyond what extendedconfirmed is supposed to be. Anomie 21:34, 11 February 2017 (UTC)
@Anomie: I do agree. I was just trying to weed out sockpuppetry or totally uninformed users as a concern when throwing out ideas. I really don't care how we get this done as long as we're working toward some accountability for all bot operators (including me!) after a BRFA is approved. I think we can all agree that there have been some cases (and not just the most recent high-profile one) where editors disagreed with a bot's current function and had no recourse. Hopefully, serious issues can be side-stepped entirely if we can come up with some method for a BAG member or other party to come in, hit the pause button, and force some discussion. That would be better than the current system, which requires a full discussion rescinding an approval before we really have any way to get a botop "to the table" to discuss their bot's behavior. ~ Rob13Talk 21:42, 11 February 2017 (UTC)
#1 sounds good, though so far BOTPOL has omitted any BAG authority to do anything about approved bots unless the issue is already brought back to re-approval. #2 is same as #1 in that BOTPOL doesn't give BAG this authority, though having an admin request this would be an extension to #1. It sounds a little bureaucratic, but I guess it's a logical step. #3, however, does sound bureaucratic. I think (from reply above) new editors and socks and such can just be listened to or ignored by common sense. I think that consensus is key rather than any specific editor counts or their "ranks". If there is an obvious issue, then there is an obvious issue, regardless how many editors notice it. Anyone could just ask BAG to review something, given #1 is ever a thing. —  HELLKNOWZ  ▎TALK 22:07, 11 February 2017 (UTC)

Focusing the discussion back on the topic of the guide, I've started Wikipedia:Bot Approvals Group/Guide, please comment on the talk page. Headbomb {talk / contribs / physics / books} 02:30, 12 February 2017 (UTC)

Trials and subtasks[edit]

Looking at past BRFAs, I realized we often approve bots (e.g. Wikipedia:Bots/Requests_for_approval/MenoBot_4) with multiple subtasks for "50 edits", without really making sure that each subtask has properly been tested. I've added this to the WP:BAGG guide. Feel free to tweak the wording on it. Headbomb {talk / contribs / physics / books} 20:45, 27 February 2017 (UTC)

I'm going to ping all active members so everyone is on board.

Headbomb {talk / contribs / physics / books} 20:47, 27 February 2017 (UTC)

Hehe, I literally just replied to similar concern in a BRFA. Personally, I treat each subtask in a BRFA as a separate task and make sure that a trial has examples of all or as many as possible, or at least sufficient details are specified. After all, you could file 20 BRFAs or just 1 with the same scope. So this addition is pretty much the practice, or should be. I would go as far as to explicitly add to WP:BOTREQUIRE that each subtask has the conditions mentioned and that WP:BOTAPPROVAL can request each one to be tested, regardless of overall scope. —  HELLKNOWZ  ▎TALK 21:10, 27 February 2017 (UTC)

I'd support making it part of WP:BOTREQUIRE personally. We can wait on upcoming WP:BOTPOL RFC to make the change. I mean I doubt it's contentious, but it would be a change in policy. Headbomb {talk / contribs / physics / books} 21:26, 27 February 2017 (UTC)

In general, sure. But I recently looked back at AnomieBOT 54, where one of the subtasks wasn't tested during the trial because it would have required a four-month wait to test the yearly "add a new archive link to WP:BRFAAA" bit. Anomie 01:28, 28 February 2017 (UTC)
Pretty sure BAG wouldn't have killed one of its own for doing BAG-related business on a low-traffic BAG-related page. Common sense still applies :p. If policy is updated, I'm sure we'll have some adequate wiggle room in whatever wording we settle on. The concept of testing archival however is valid. If it's critical/core to the task, and the potential for disruption is high, then a sandbox page with mock entry can be set up, or a spoofed entry made to trigger archiving conditions. Headbomb {talk / contribs / physics / books} 01:54, 28 February 2017 (UTC)

Concerns about syntax fixing bots[edit]

Collapse off-topic discussion. Take it elsewhere, please. Anomie 01:33, 28 February 2017 (UTC)

Exactly. As I thought. Now there is going o be witch hunt against more bot. Not only Yobot. Let's see where this ends. -- Magioladitis (talk) 20:50, 27 February 2017 (UTC)

Headbomb Wikipedia:Bots/Requests for approval/BG19bot 7 take this too. -- Magioladitis (talk) 20:52, 27 February 2017 (UTC)

Headbomb This is because you use the false assertion that some tasks "should be done in addition to others". Where is this written anyway? -- Magioladitis (talk) 20:55, 27 February 2017 (UTC)

I'm not quite sure what you're referring to here with "false assertion that some tasks "should be done in addition to others"". If you mean WP:COSMETICBOT, then yes I stand by the idea of "cosmetic changes should be applied only when there is a substantive change to make at the same time." Even in the draft for a revision User:Anomie/Sandbox2 that remains a core idea. But the above isn't about cosmetic fixes, it's to fix a flaw in our reviewing process. In some cases, it might require re-doing some BRFAs, but in most cases the bots have run for long enough that they effectively had trial by fire, and bot ops have adjusted to the feedback accordingly. Headbomb {talk / contribs / physics / books} 21:13, 27 February 2017 (UTC)

Headbomb So you are in favour of approving AWB bots with general fixes activated. That's good. I support this. -- Magioladitis (talk) 21:14, 27 February 2017 (UTC)

The only thing left is to define "cosmetic changes" with a wiki definition. -- Magioladitis (talk) 21:16, 27 February 2017 (UTC)

Yes, but I'm also not giving a blank cheque for unleashing cosmetic genfixes or trivial checkwiki fixes. I have AWB genfixes enabled for my own User:CitationCleanerBot, but that never got me in trouble because the logic is such that cosmetic genfixes will rarely be done on their own even if they are miscategorized as "major" genfixes. Other bots may have a higher likelihood of doing such edits, and might not get approved because of that. But this is a discussion that should be held in some other place than WT:BAG, so if you want to discuss that, I suggest you take it to WT:AWB, WT:CHECKWIKI, WP:VPT, or possible WP:BOTN. Headbomb {talk / contribs / physics / books} 21:23, 27 February 2017 (UTC)

Headbomb What looks trivial to you it took more than 6 years to establish tools to fix it. It helped improving AWB/WPCleaner logic, etc. -- Magioladitis (talk) 21:31, 27 February 2017 (UTC)

Again, stop putting words in my mouth. It's the third time I have to ask you to stop doing that in the last hour. As for how long "it" took to develop, whatever that may be, that is completely inconsequential on whether or not "it", whatever it is, should be done. Headbomb {talk / contribs / physics / books} 21:35, 27 February 2017 (UTC)