One bot per task?
Let's talk about this. The "odd policy" may be a relic; it may not reflect the consensus of the currently-active editors and administrators.
To my mind, there are two main reasons why bots need to be approved:
- To ensure that the bot software is properly written; to give interested bystanders an opportunity to review the code if necessary.
- To ensure that the bot owner is someone who can be expected to run the bot responsibly.
There are several facets of point 2: use the bot carefully so that it doesn't run amok, clean up after the bot if it does run amok or if you have it make some edits which the community decides in retrospect were unwarranted, and most importantly, only use the bot to perform edits in the first place which you're sure there is or would be good consensus for.
That last point, I think, if properly applied would override the need to preapprove tasks for task-specific bots.
Like all the wikis, we encourage editors to be bold and just go ahead and make the changes they want to make. We don't insist that editors ask permission before making every little change (in fact it's not unheard of for a too-cautious editor to be chided for asking permission before making some change). If an individual editor makes an ill-advised edit, it can always be reverted.
The problem with bots, of course, is that an ill-advised class of edits performed on hundreds or thousands of pages at once can't reasonably be manually reverted. How do we prevent such widespread but unwanted bot-performed edits? One way is to require preapproval of each task, but another is to: trust the bot owner. Now, of course, trust can be a difficult-to-achieve commodity, and we don't trust everyone, but then again, building trust is what the approival process (any approval process) is all about.
I believe that an aspiring bot owner is making several promises:
- I won't use the bot to perform any edits that I wouldn't, in good conscience, perform manually myself, except that
- I will not "be bold" with the bot, and specifically, before giving the bot a task to do,
- I will ask around for consensus, perhaps at the Beer Parlour, or
- I will make sure that the task is so innocuous that no one could possibly object. Finally, if I should screw up,
- I will stop the bot if objections are raised to a task it's performing, and not restart it until there's positive consensus, and
- I will un-do (perhaps with the same or another bot) any and all damage inadvertently caused by my bot, and any edits which consensus decides were unwanted.
I believe that, with good understanding and appreciation for these points all around, by aspiring bot owners and by concerned bystanders, we could drop the "one task per bot account" aspect of the bot policy, and instead approve bot accounts as a combination of a trustworthy bot owner, a well-written bot program, and a set of promises to use the bot responsibly.
What do other people think?
—scs 19:24, 31 July 2006 (UTC)
- Generally agree. Large parts if not all of this should become official policy, as long as we keep the focus on the fact that even though someone is trusted to run a bot, that doesn't endorse them running all kinds of tasks never discussed anywhere. Any operation that might have the least controversial nature should not be done without discussion. — Vildricianus 20:19, 8 August 2006 (UTC)
- This is a healthy and sensible direction for English Wiktionary. Bot owners here seem to have a good understanding of the types of edits are obviously controversial, the types that have wholesale consensus, and the types that may require further deliberation. They also seem responsible enough to clean up after themselves if they guess wrong. So, I support formalizing the "Bot Owner Vows" above (which until now have been unwritten rules) and dropping the "one task per bot account" aspect of the bot policy. Rod (A. Smith) 04:27, 9 August 2006 (UTC)
- Support. Although, I think the reasons not mentioned above are still salient. That is:
- We are not Wikipedia, and 'bot geared toward Wikipedia activities are rarely appropriate.
- Likewise, capable 'bot operators on Wikipedia are highly likely to fall into common pitfalls here.
- Bot operators such as myself are apt to consider a bot run on sketchy premises, or not fully discussed premises (as I nearly did with the whole TranslationBot thing, out of ignorance.)
- 'Bots can run with the "minor edit" flag set, as a regular user. Most who are savvy enough to use Special:Recentchanges are also savvy enough to click "Hide minor edits" during one-time runs.
- "Most" runs/tasks are of the one-time-only variety.
- "Most" tasks are more lexically complicated than they appear, especially to one of a programming mind-set. (e.g. To a programmer, a spelling correction 'bot would seem at first blush to be almost reasonable, when in fact, that would be one of the single most vandalistic things possible.)
- With those (and several others, that Ec could enumerate better,) I think there is still some value to the current policy/practice (that I personally detest.)
- The argument that I think wasn't presented above, for revising the policy is this: If you find someone trustworthy, then trust them.
- --Connel MacKenzie 07:21, 9 August 2006 (UTC)
- Gee, Connel, if you detest the current policy so much, why are you arguing so persuasively in support of it? :-)
- Couple points:
- 4. "Bots can run with the 'minor edit' flag set, as a regular user." But this isn't a very good argument, is it? If everyone who wants to (responsibly or irresponsibly) run a bot, but who can't get the bot bit set due to a too-stringent bot policy, decides to do the expedient thing and run it as a regular user with the 'minor edit' bit set instead, then the bot policy isn't just too stringent, it's useless, and we've abandoned any attempt to guide bot usage in any way at all.
- 5. I'm not so sure about a preponderance of tasks being one-time-only. Most of our bots are in support of some formatting convention or another, and all of those conventions tend to need ongoing maintenance, because human editors won't always honor them.
- Also, if you have any specific examples of how bots or their operators are so different here than on Wikipedia, please mention them. Our bot policy ought to (I think) be an extension of Wikipedia's, so those differences would be worth explicitly discussing.
- —scs 13:43, 9 August 2006 (UTC)
- The major difference (that has lots of sub-differences) is that each Wikipedia article has a format depending on the topic. Each Wiktionary entry, OTOH, should conform to WT:ELE. So the Wikipedia 'bots in general, function on only a single topic subset of entries at a time. On Wiktionary, 'bots affect (often) all words in a particular language; a much, much larger subset. (You can draw comparisons, but you were asking me to identifiy fundamental differences, so...)
- 'Bots on Wikipedia have to therefore cater only to a small subset of the Wikipedia editors. 'Bots on Wiktionary need to satisfy all Wiktionary contributors, as the formatting discussions affect everyone here. A good example would be the NanshuBot snafu. By thinking he could address only zh language concerns, we've been innundated with a mess that conforms to no standard at all.
- Interwiki links are pretty much the opposite of what Wikipedia does.
- You mean, the way WP's point to other-wiki pages on the same topic (typically with a translated name), while ours point to other pages on the same spelling (typically meaning something completely different)? Yeah, I was just noticing that. (Thanks for your other points.) —scs 12:21, 13 August 2006 (UTC)
- So, a 'bot operators from Wikipedia might incorrectly assume that they can do discrete, small 'bot tasks for Wiktionary. Nothing is further from the truth; most non-one-time-only 'bot functions (such as Morobashi/Morohashi) actually do affect a much wider group than the 'bot operator is likely to realize.
- All that said, my vote firmly remains at Support. I think the guidelines will work themselves out, as the reform is seen. --Connel MacKenzie 21:08, 12 August 2006 (UTC)
- Support Kipmaster 15:26, 14 August 2006 (UTC)
- All this looks like a huge step in the right direction. I would add (if I didn't miss it in there already) a specific statement that the bot operator should thoroughly test the bot on a small number of entries (perhaps 10-100) that can be manually undone if necessary before letting the thing go on longer runs unattended. —Dvortygirl 23:04, 9 September 2006 (UTC)
- One more comment. I think the stuff that should be clearly stated in the bot request (your user name, what exactly the bot does, etc.) should also be clearly stated on the user page for the bot. —Dvortygirl 18:21, 10 September 2006 (UTC)
I object to the number of hoops that are required of a bot which are not required of a user, I object to the standard that bots are held to which users are not, and I object to how difficult it is to make any headway as it pertains to policy and change on en.wiktionary. It seems VERY silly to me that if I am to automate a task that I was doing anyway, with no objection from anyone, I then have to host a forum on the matter in the beer parlor, reach a concensus (5-20 days) wait a few more days for good measure, start the bot, and then stop it any time someone objects to what I am doing until another discussion has taken place and consensus has been met again. If we required this of every editor before they undertook some task nothing would get done (which is mostly the case with bot work, it seems). I propose a paring down of the rules to something more simple:
- In order to get a bot flag one must:
- be a trusted member of the community
- ask a 'crat nicely.
- In order to run a bot task one must:
- only do things which one knows are beneficial to wiktionary.
And that is it, let people who we trust to edit wiktionary anyway make their jobs easier, why put all kinds of legalism and red tape in the way of adding to and cleaning up wiktionary, which is why we are all here in the first place. - TheDaveRoss 06:46, 18 September 2006 (UTC)
- Um, were you given permission to speak rationally and logically here? No? Objection dismissed out of hand, then! --Connel MacKenzie 07:13, 18 September 2006 (UTC)
- Thanks, fixed. --14:09, 28 March 2007 (UTC)
I think that this page needs to be rewritten (and not just 'cos my bot is being voted on at the moment ;). I propose something along the lines of the following which seems more inline with current practise. Conrad.Irwin 01:03, 16 April 2008 (UTC)
- You are responsible for every edit your bot makes, as though you had made it yourself.
- If you haven't ensured that you have community consensus in addition to simply obeying all policies (esp. ELE) then don't make the edit with a bot.
- If the bot makes a mess you are responsible for cleaning it up and ensuring that it can't happen again.
- Respect the servers, you should obey any error messages you receive from them and not hammer them with too many requests.
- Don't flood RecentChanges, this means that for large edits and quick edits, you need to run a WT:VOTE to get a bot-flag. It also means that low-quantity, slow, automated editing is permissible through non-flagged accounts.
Bot to create enPR pronunciation guide from IPA?
Can anyone tell me if Wiktionary's enPR guide is created automatically from IPA, or are all those done by hand? I could use something like that at Wikipedia. - Dan w:User_talk:Dank55
- To my knowledge there is no automatic IPA → enPR conversion. --Bequw → ¢ • τ 18:44, 11 November 2009 (UTC)
- Such a thing isn't really possible, because a particular IPA vowel symbol may stand for more than one enPR vowel symbol. The choice depends on the particular dialect that the IPA transcription represnts. Also, some IPA symbols are not transcribed in enPR. --EncycloPetey 00:40, 12 November 2009 (UTC)
request for a bot status
Where could I request for a bot status in english wiktionary? --Aibot 20:01, 2 May 2008 (UTC)
- Well you could have on the beer parlour but it is unlikely to be accepted as we already have a much more efficient Interwiki bot - see User:Interwicket. Conrad.Irwin 19:07, 3 May 2008 (UTC)
...does everything except tell you how to make a bot. ---> Tooironic 07:42, 30 April 2010 (UTC)
- That's not its purpose. There is adequate coverage of that issue already at WP and Meta. --EncycloPetey 13:15, 30 April 2010 (UTC)
I would like to make a "bot-request", is that possible here like at the Wikipedia? Kreuzkümmel 17:45, 19 January 2011 (UTC)
- Do you mean you want a bot to do something for you? Ask in the Grease pit. Or do you mean you want to run a bot? Ask in the Beer parlour.—msh210℠ (talk) 17:47, 19 January 2011 (UTC)
I think it would be nice if we listed somewhere, maybe here, the wikitext accommodations that we make for bots so that they can parse the structure more easily.
- We don't put headers creating text (='s) in templates
- We don't put list creating text (#, *, #:, etc.) in templates (for the first line if template is multiline).
- We put a dividing line ---- between language entries
I don't want to start the discussion again, but having only one interwikibot is a simple point of failure. Could the botpolicy be changed that other interwiki bots then the Interwicket bot is allowed here. Interwicket bot is down for several months and is not very likely to comeback. If only one bot for one task is the policy, please remove interwiki botflag from Interwicket and give it to bot that is active and is willing to run on all the wiktionaries. 126.96.36.199 10:52, 22 April 2011 (UTC)