This has been brought up several times, you may want to review [[Wikipedia:Requests for rollback privileges]] on en: for more information on the past findings.
Having read through that page I just have to say: Gosh, what an amazing amount of debate and polling, with most of the people in the discussion having zero capacity to fix it, and all leading to no real conclusion.
The worry of course is that it makes vandalism very, very easy. If the threshold were low then it would be open for abuse by vandals.
I agree, and that's why I'm thinking a high threshold.
User contributions follow a power law (example graph: http://wikiq.nber.org/figs/png/eswiki-20060620-hist_log_total-edits.png ). i.e. the total number of users who would have edits in the 1000's is a very small percentage of the total userbase.
Using the figures from http://en.wikipedia.org/wiki/Wikipedia:List_of_Wikipedians_by_number_of_edit... (with hasn't been updated since March, but it'll do for a quick calculation), and the figure of "English Wikipedia had about 1,064,259 registered users" on that same page:
To put this into numbers, looking at just edits (and disregarding account lifespan), the edit threshold, number of users who meet or exceed that threshold, and the percentage of the userbase that this represents are as follows: Edits Num users % of userbase 1500 2500 0.23% 2000 1923 0.18% 3000 1293 0.12% 4000 924 0.09% 5000 711 0.07%
In other words, only a tiny percentage of users would get this, and if a vandal is willing to make thousands of good edits just to have a window of opportunity of a few minutes in which to vandalise: a) It's rather unlikely. Let me put it this way: If someone wanted to damage the Wikipedia, to me this seems to be an enormously inefficient and ineffective and labour-intensive way to do it. b) It's probably a net win anyway. In other words, the vandals would have to do enough useful stuff to cross a threshold, more than the damage they would do in those few minutes. In other words, you harness the power of the vandals, by getting more out of the vandals than you put in, so you're still better off.
Thoughts: not to be a jerk, but this is more of a Wikien-l issue
You're not being a jerk, and I understand what you're saying.
However, there *is* a significant technical aspect to this, and ignoring that technical aspect is a good way of guaranteeing that it won't work. There are two sides to this: 1) Getting general agreement. 2) Implementing it.
Finding something that can be agreed upon does little good if there is no interest in implementing it, or if implementing it causes so much load, or difficulty, or takes so much effort that it's not worthwhile, or just simply can't be done. Therefore the technical aspects cannot be ignored, and need to be discussed too.
There's a user JavaScript tool that was developed some time ago which provides about the same one-touch rollback capability for any user
I know, but as you say it's not ideal, and just not as good as a server-side solution. Furthermore its very existence is evidence of a desire that the Wikipedia were better in this regard.
we have an extension suitable for this.
Yes, your GiveRollback extension ( http://svn.wikimedia.org/viewvc/mediawiki/trunk/extensions/GiveRollback/ ). I've read through the code, and (please correct me if I'm wrong) it seems to be a way that bureaucrats can add and remove rollback abilities for named auto-confirmed existing user accounts.
That's definitely an improvement, although I'm wondering if the addition and removal can't be largely automated to enormously reduce the load on bureaucrats (i.e. this would be a good UI for the exceptions, with the general case being handled automatically).
I think that the important aspect of manual reverting is the ability to enter summary of the edit, i.e. to explain why someone is reverting.
When I revert someone they are I think always: a) an anon b) have written something like: "YOUSE ARE ALL BOGANS WHO CAN KISS MY 733T FURRY BUTT!!!! WOOT!!! YOUSE HAVE BEEN HAX0R3D BY S0nIc WaV3!!!!" c) my edit description is usually "revert" or "rvv" or just "rv".
So: 1) Is a more detailed description really necessary? (e.g. "Revert: Buttocks need branding with hot poker iron" ?) 2) The default "Revert edits by [[User:x.y.z.w]] to last version by [[NiceGuy]]" that I think you get on rollback are nicer anyway than the descriptions I use.
Not only that, it can generally be rolled back by *anyone*, even an anonymous user.
Exactly. This is about making what people can already do easier for those who have stuck around long enough to know what's expected, and know outright vandalism when they see it.
At the moment, here's what I'm thinking:
General idea / spec:
Technical rules: * Upon exceeding 2000 edits and also having an account for more than 2 months and having never been blocked for any period of time, the user shall by an automated process be granted the ability to do limited rollbacks. * These limited rollbacks can only be used on anons (as a technical restriction). (For named accounts need to do it the old slow way). * You lose limited rollback ability immediately and automatically for violating 3RR (i.e. if you use the limited rollback on the same article within 24 hours, then the limited rollback permission is automatically removed from your account). [Or perhaps it should just refuse to let you do a limited rollback in this situation?] * Maximum of one limited rollback every 5 seconds (e.g. by server-side sleep() for a few seconds, or by saying "too many rollback requests in too short a time. Press refresh in a few seconds to try again"). * Bureaucrats can add or remove the limited rollback permission at their discretion.
Social rules: * You can only rollback for blatant vandalism or a very obvious mistake (i.e. an edit description should not be required because the reason for the revert should be obvious. If an edit description would be required, then revert manually).
I would appreciate comments on whether: a) it seems technically feasible and/or sensible b) if I can deal with the political slog to get agreement, if there is anyone out there that might potentially be interested in building it, or a bit of it? c) if you think it can be improved / simplified d) Are any of the technical rules not required, or are more required? (e.g. is 3RR restriction required? Is limited rollback throttling required? Is there a better way of doing it?)
Comments / thoughts?
All the best, Nick.