Wikipedia:Bots/Requests for approval/Femto Bot 4


 * The following discussion is an archived debate. Please do not modify it. To request review of this BRFA, please start a new section at WT:BRFA. The result of the discussion was Symbol keep vote.svg Approved.

Femto Bot 4
Operator: Rich Farmbrough

Automatic or Manually assisted: Automatic

Programming language(s): Perl/AWB

Source code available: AWB Perl no.

Function overview: Manage my BRFAs

Links to relevant discussions (where appropriate): N/A

Edit period(s): Continous

Estimated number of pages affected: 1-3 per day

Exclusion compliant (Y/N): N, but Y on baggers talk pages.

Already has a bot flag (Y/N): Y

Function details:
 * Generate BRFA from spec and post
 * Update bot pages and my pages with appropriate status changes
 * Tag stale BRFAs
 * Ping BAG for very stale BRFAs
 * Generate code for bot
 * Run trial once authorised
 * Post results
 * Switch on task once authorised

Discussion
As Femto Bot doesn't have a bot flag yet I will be using it manually to test the code. [Update it now has a flag, and is effectively permitted to edit my userspace anyway 2010-10-08.]

Rich Farmbrough 01:49, 6 October 2010 (UTC)


 * Where do BRFA specs come from? Do you write them and tell the bot to post them? How are stale BRFAs tagged? I don't think pinging BAG members should be done automatically. What does "generate code" for bot mean? How is the trial run — is your bot automatically told to run a trial once a trial is given? I don't think such automation should ever be done for trial runs of a bot. What are the posted results — is this a bot generated report or just the contribution list? — HELL KNOWZ  ▎TALK 11:47, 8 October 2010 (UTC)
 * Yes I write them at the moment, the bot will post them automatically.
 * Stale BRFA's will be tagged with the bag assistance needed template.
 * BAGBot was I think supposed to ping BAG members.
 * Code generation will of course be done by AI. (and is there for information as much as anything - since it happens off wiki.)
 * The trial is usually "50 edits" or similar. If the run is automatic (which, in fairness, will only be possible some of the time) there is no reason that the trial run can't be "5 edits", which can be reviewed almost immediately followed by 10 more, or 50 or whatever the BAGGER thinks appropriate.
 * Posted results will be the contribs list - that's the current plan. It might be possible to generate a little more information - edit conflicts, time taken etc, but generally I would think this is not stuff BAG is interested in.
 * Rich Farmbrough, 14:01, 8 October 2010 (UTC).


 * I suppose automatic trials are useful for the reviewing BAGger, as long as everything is fine. But if anything goes wrong, they would have to revert the changes themselves or wait for you. That's my concern. But given you are a long-standing bot programmer, I hope this shouldn't be an issue. "Code generation will of course be done by AI.". Do you mean the code will be posted for review? Because it sounds like there is going to be an AI writing the code. :) —

HELL KNOWZ ▎TALK 16:05, 8 October 2010 (UTC)
 * Well, the AI may take a little work yet... But your other point is valid: however it is just as valid for "manual" trials.  I can do a trial and it not get reviewed for a couple of days.  And that's why I said they can always say, "hmm 5 edit trial please." and either "That's borked, 5 rollbacks" or "Looks good, give me 10 more". It's also true to say that, for example AWB edits can't be put on this basis just yet. Rich Farmbrough, 23:20, 8 October 2010 (UTC).


 * "I can do a trial and it not get reviewed for a couple of days." — but you would have to promptly revert any errors after the trial run, as you would be present. This is left to BAGger if you are not available and the trial was automated. Also, you didn't mention BAGger being able to ask for reverts as well, so that should balance it out. Regarding AI, I am still unsure if you are serious, but It'd be nice to see the first AI to make programs on demand. :) — HELL KNOWZ  ▎TALK 00:50, 9 October 2010 (UTC)
 * Well, could build in reverting (like revert them all - bot spelled X with a Y). Maybe this is an area we can feel our way, if baggers are uncomfortable they can ask for 1 edit, 1 edit, 2 edits... And it's also true that reviewers pick up errors that the botmeisters don't - that after all is one purpose of the review. As to the AI, yes it's tongue in cheek, but I certainly have written programs to write programs to write programs. Rich Farmbrough, 04:24, 9 October 2010 (UTC).

Isn't this edit a little in advance of getting trial approval for this? VernoWhitney (talk) 22:49, 8 October 2010 (UTC)
 * Just a little but I wanted to see what colour the pie was. Rich Farmbrough, 23:14, 8 October 2010 (UTC).

Rich Farmbrough, 23:24, 10 October 2010 (UTC).


 *  MBisanz  talk 23:53, 14 October 2010 (UTC)

The point of a trial is that the operator reviews every single edit as it is made; there's no point to an "automatic" trial. Moreover, RF has a poor record of cleaning up mistakes when his tasks go wrong. I have had to revert innumerable broken edits by SmackBot. So I can't see how automated trials are going to improve things. &mdash; Carl (CBM · talk) 11:48, 18 October 2010 (UTC)
 * Some errors here - Kingpin13 (talk) 12:01, 21 October 2010 (UTC)
 * Also error here, should have been placed in the Requests to add a task to an already-approved bot section rather than Current requests for approval - Kingpin13 (talk) 10:28, 22 October 2010 (UTC)
 * Thanks, one was known and not yet implemented, the other was implemented but untrialed code. All being well both should work now. Rich Farmbrough, 13:49, 22 October 2010 (UTC).


 * And space suppression too. Rich Farmbrough, 14:22, 22 October 2010 (UTC).


 * Please note the trial is over and the bot is still making (erroneous) edits for this task - please shut that portion off. – xeno talk 15:16, 2 November 2010 (UTC)
 * My fault for giving it a bad BRFA name. Rich Farmbrough, 11:00, 5 November 2010 (UTC).


 * Am I missing something or is this bot still making edits for this expired trial? VernoWhitney (talk) 13:54, 10 November 2010 (UTC)
 * Yes but they are manually supervised. Rich Farmbrough, 15:55, 12 November 2010 (UTC).

Rich Farmbrough, 21:44, 21 November 2010 (UTC).

Second trial
It's easy to appreciate why this BRFA hasn't been touched for three weeks, but since all the problems were with the code (and therefore fixable), - Jarry1250 [Who? Discuss.] 11:51, 12 December 2010 (UTC)
 * D Was this trial done? Mr.Z-man 04:13, 2 January 2011 (UTC)
 * It's running now, it has submitted BRFA SmackBot 42 43, only yesterday. Rich Farmbrough, 10:05, 2 January 2011 (UTC).


 * Looking at it's past edits and discussion:


 * 1) Generate BRFA from spec and post – O.K.
 * 2) Update bot pages and my pages with appropriate status changes – O.K.
 * 3) Tag stale BRFAs – what does "stale" mean? Is it a time period? The bot can't detect things like wider discussion requests or some related discussion taking place elsewhere, etc.
 * 4) * Means no templated status, last edit is by me, and more than 24 hours ago. Rich Farmbrough, 10:42, 3 January 2011 (UTC).


 * 1) * 24 hours is hardly "stale" in current activity. Some BRFAs live on with weeks of no replies. The description page itself says "If you feel that your request is being overlooked (no BAG attention for ~1 week) you can add BAG assistance needed to the page." This number was probably based on experience rather than consensus, but is still more realistic. Are you O.K. with this being a week? — HELL KNOWZ  ▎TALK 11:04, 3 January 2011 (UTC)
 * 2) Ping BAG for very stale BRFAs – which BAG members? All? Those who commented?
 * 3) * 72 hours after bag assistance is requested with no BAG response it will ping one "active" BAG member, wait for 24 hours then ping another, after that it will move to "inactive" members at one per 12 hours. Rich Farmbrough, 10:42, 3 January 2011 (UTC).


 * 1) * Can it first ping those that have already participated in the discussion, i.e. before the BAN template. Also, don't ping inactive members, they don't participate for their own reasons and you can't tell who may get agitated by a random ping to a BRFA they have never seen before. — HELL KNOWZ  ▎TALK 11:04, 3 January 2011 (UTC)
 * 2) **Yes that makes sense. I don't realistically expect it to get to non-active members, but some of them have been classified non-active by me. If a ping makes them "agitated"... then well, they are less than inactive - they have effectively left, and should be removed from the roster completely - or at least classified as "on leave" or something.  Rich Farmbrough, 19:49, 3 January 2011 (UTC).


 * 1) ***I thought you meant inactive at WP:BAG list? Inactive in BAG does not mean inactive. Well, anyway, as long as you don't get complaints and WT:BAG/BRFA doesn't, I suppose it is O.K. — HELL KNOWZ  ▎TALK 20:00, 3 January 2011 (UTC)
 * 2) Generate code for bot – This is really your side of things and you may choose to generate the code as you wish, but this isn't something a blanket approval can be given for.
 * 3) * Yes this is only for completeness, and out-with BAG's purview. Rich Farmbrough, 10:42, 3 January 2011 (UTC).


 * 1) * You are welcome to mention this, but completeness of your feature documentation is not really the same as a list of actual WP-related tasks. — HELL KNOWZ  ▎TALK 11:04, 3 January 2011 (UTC)
 * 2) Post results – (I assume of the trails) O.K.
 * 3) Run trial once authorised – as below
 * 4) Switch on task once authorised – I really prefer you activate the tasks yourself, especially those that edit fast or between other tasks. That is the point of BRFA after all. I can see how it can be easier for BAG member who already know and are aware of your automated system though. I prefer that some BAG members post how they feel about this. — HELL KNOWZ  ▎TALK 10:37, 2 January 2011 (UTC)
 * The last two points are again here for completeness, it may very much depend on the task, those where there is a simple matter of grabbing N pages and applying a fix are clearly more amenable to controlled trials (human or bot initiated) than those that require an error condition that is normally absent to occur. Rich Farmbrough, 10:42, 3 January 2011 (UTC).


 * Last two points really need more BAG input. — HELL KNOWZ  ▎TALK 11:04, 3 January 2011 (UTC)
 * More BAG input: Yes, I agree that the tasks need to be started manually by Rich himself. - Kingpin13 (talk) 02:14, 9 May 2011 (UTC)

BAG assistance needed Rich Farmbrough 23:19, 4 January 2011 (UTC)


 * That was annoying. Is there a good reason this request states that the bot will not respect bot exclusion? Anomie⚔ 23:45, 21 February 2011 (UTC)
 * Yes there is, it was a bot limitation when I posted the BRFA, and not relevant to most of its work. However I will modify the code to check on BAG members talk pages. Rich Farmbrough, 23:54, 21 February 2011 (UTC).


 * On the other hand if you wished to recuse yourself, as Xeno has you could simply have told me. Rich Farmbrough, 23:57, 21 February 2011 (UTC).


 * Any updates?  MBisanz  talk 10:13, 21 April 2011 (UTC)
 * Yes I ran the BAG notifying part, the result is Anomie's response above, and elsewhere they recuse themselves form my BRFAs.  Rich Farmbrough, 18:50, 1 May 2011 (UTC).


 * Incidentally the Bag assistance needed template above is still active from 4 January. This type of delay is the reason that I wanted to ping BAG. If however BAG members are unsympathetic to automated pings, and BRFAs are really going to take maybe a year to get through, there's really little point bothering. I noticce however that one of Anomies BRFAs took a few hours or days, ushered on a spurious concept of urgency. Rich Farmbrough, 22:58, 1 May 2011 (UTC).

Alright, I'm late to the party it seems. The end goal of the bot seems like something most BAG members would support, and would mostly concern BAG members and bot operators. I understand that if the notices annoy them, people can opt out of them. However, I'm unclear about what's the "notifying" logic, at least in terms of what exactly is considered an "inactive" BRFA, and who gets noticed. Could you clarify these two aspects? Headbomb {talk / contribs / physics / books} 05:57, 9 May 2011 (UTC)
 * The trial seems to have been done with 24 hours in mind, while a week is more sensible (at least if we're following BAG-related templates). So the bot would probably have to stay quiet for one week without BAG / Bot op activity (whichever applied).
 * As far as BAG-related notices go, they should be first be given to BAG members which posted in the BRFA. Then failing a response (say in the next 24 hour period), a notice to another BAG member, preferably drawn at random from the active BAG member list (repeat ad-nauseum until you run out of BAG members or that you got a response).
 * If a week is the minimum delay, then so be it.
 * Yes, I haven't implemented the "related bagger" functionality but I can do that. Rich Farmbrough, 08:27, 10 May 2011 (UTC).

Alright, then. Let's have one-week thing for now. If people feel that this is too slow/fast, just get a straw poll at WP:BAG or something. Headbomb {talk / contribs / physics / books} 09:33, 10 May 2011 (UTC)


 * The above discussion is preserved as an archive of the debate. Please do not modify it. To request review of this BRFA, please start a new section at WT:BRFA.