Last modified: 2011-03-13 18:04:32 UTC
Currently, a system of wild guessing is used to determine is someone is a 'bot. Edits performed using the python suite of code should be flagged differently, especially if performed from an account that does not have the bot flag set. Obviously, there are certain times where this is considered acceptable (i.e. less than 50 edits.) But identifying this class of edits separately would be helpful.
Keep in mind that this is totally unreliable; a malicious bot could pretend to be human, or pretend to be a bot, as it wills.
(In reply to comment #1) > Keep in mind that this is totally unreliable; a malicious bot could pretend > to be human, or pretend to be a bot, as it wills. Exactly. But with no way of discerning which is which, no judgement (with respect to a wiki's local policies) can be made. All kinds of spoofing exists - bogus or unrelated edit summaries, "minor" etc. This one simply will indicate that someone seems to be using the pywikipedia.py code. Some people will want to pay attention only to bot edits (sometimes) while other people will want to exclude all of them all of the time. Right now, only the server "knows" which is which.
Marked enhancement. What you're suggesting is somewhat equivalent to HTTP user agents.
(In reply to comment #0) > Currently, a system of wild guessing is used to determine is someone is a 'bot. No, it's not - it's quite simple. A local bureaucrat assigns a user "bot" rights, which suppress that user's edits from showing up in the default recent changes view; it's an additional filter. This is typically done because someone would like to perform a large number of edits, semi or fully automated, without flooding changes pages, but it doesn't have to be. Detecting apparent bots is dangerous, because if one manages to spoof as such, and we grant it bot rights, then potentially unpleasant editorial changes will go unnoticed by the vast majority of users.
Rob, that is the *opposite* problem. What you say is true on Wikipedia, perhaps. But on Wiktionary, it would be really helpful to see which users are actually using automated tools to create lots of entries quickly (not marked with the bot flag) so that they can be slowed down or coerced into compliance with existing formats, before they go too far. That's why my original request wasn't to filter in any way; merely to _flag_ such edits in Special:RecentChanges. It is quite the opposite response to suggest that someone pretending their bot might be given the bot flag. The natural result that I see, is that they'd be given a short term block, while at the same time a lot more people would suddenly pay attention to those edits.
The only thing I can think is to show in a special way in recent changes such eidts (for example in a way similar to the red exclamation mark used for the markaspatrolleddiff mecchanism. But the main problem is how to know that an edit is made by a bot? One way can be using the HTTP user agent string (see #3), but this way you will able to catch the edit made by compliant bot that agree to use a specific string in the HTTP user agent identification. One other way would be to trig a mechanism when too many edits are made on a limited number of time by one user or one IP (this would not have to be done in live)
