Page MenuHomePhabricator

rollback action does not match abusefilter
Closed, DuplicatePublic

Description

Abusefilter does not match on rollback action only on edit, move, upload, ....
On dewiki many people (all Editors) can use rollback so that's a big problem.


Version: unspecified
Severity: major

Details

Reference
bz22713

Event Timeline

bzimport raised the priority of this task from to Lowest.Nov 21 2014, 10:58 PM
bzimport added a project: AbuseFilter.
bzimport set Reference to bz22713.

matthew.britton wrote:

I would consider that a feature, not a bug.

Rollback can only restore previous versions of pages, and said previous versions can only exist if they got past the abuse filter in the first place.

Sure, if you really wanted to, you can search through recent changes and find a page where rollback will reintroduce vandalism, and then do the rollback. But if you have members of one of your privileged groups deliberately doing that, you have worse problems than the abuse filter.

It is better to allow rollback to fix vandalism and trust your privileged group to use the tool correctly, than to end up having vandalism that *cannot* be removed because the previous version of the page trips some obscure filter. Because then you have a situation where you have to get an administrator in to fix the problem (I have no doubt you've given your administrators a free pass on all filters, just like en.wikipedia has).

Why do you even want to create filters that apply to your privileged group (did I mention this is a *privileged group* here?) If you don't trust them to edit properly without your supervision, why are they in the group in the first place? Enough of the 'us vs. them' mentality.

It's not Vandalisms, but Edit-War. And i made very good experience by limiting the edit count on a single page per user on dewiki. I set a limit of three edits per day for each user and article and that reduces die edit-war and in most cases solves the problem by time.
Some full locks in the past weren't not successful because the party of the blocked version did not participate at the discussion.

I tried the editlimit at some pages which had sysop protected since years and after a week the edit war stopped in about 80% of the pages. Now there aren't many sysop protected pages anymore.

But with rollback you can bypass the edit limit.
I think you know that it is not always easy to revoke editor right for people committing since years and the editlimit filter is more effective than i ever expected in my dreams before.

matthew.britton wrote:

(In reply to comment #2)

It's not Vandalisms, but Edit-War.

Again, this is a *privileged group* you've given rollback to. If they're abusing their position, the solution is not restrictive abuse filters, it's to remove them from that position of trust.

And i made very good experience by limiting
the edit count on a single page per user on dewiki. I set a limit of three
edits per day for each user and article and that reduces die edit-war and in
most cases solves the problem by time.

Ugh... you really did that? I'm glad German isn't my native language. Yes, I'm sure it has drastically reduced the edit warring problem. It's also lost you a lot of useful contributions.

Some full locks in the past weren't not successful because the party of the
blocked version did not participate at the discussion.

Then you unprotect the page, and block the edit warring user if they continue. I don't see why the abuse filter needs to get involved.

I tried the editlimit at some pages which had sysop protected since years and
after a week the edit war stopped in about 80% of the pages. Now there aren't
many sysop protected pages anymore.

Sounds to me like you'd been leaving those pages protected far longer than necessary. Full protection because of an edit war should never last "years". I think you're making a false correlation here -- I assure you if you'd just unprotected those pages without putting any abuse filters in place at all, you would still see an 80% reduction in edit wars, because years later, most people have either resolved the issue through discussion, or more likely completely forgotten about and abandoned the argument.

And even if they haven't, you remove the protection anyway and block them if they resume. Otherwise, nobody ever gets to contribute. Full protection because of an edit war should never last more than a few weeks.

But with rollback you can bypass the edit limit.
I think you know that it is not always easy to revoke editor right for people
committing since years

Type their name into [[Special:Userrights]], click a few buttons. It sounds pretty easy to me. If they're edit warring, they have no excuse. If they're not only edit warring but intentionally bypassing your project's policy on edit warring after you've told them to stop, then frankly I think a block would be in order, let alone de-grouping.

and the editlimit filter is more effective than i ever
expected in my dreams before.

Only because you have a warped definition of 'effective'. You could full-protect every page on the wiki and be 100% 'effective' at preventing vandalism, spam, edit warring and everything else. Do you see the fallacy there? If your filters are anything like those on the English Wikipedia -- and everything you've said points to them being much, much worse -- you are losing thousands of useful contributions, both because your filters are preventing them, and because new users frustrated with your filters are never becoming engaged enough with the project to regularly contribute.

I swear the abuse filter is the worst thing to happen to this project in recent years.

happy.melon.wiki wrote:

(In reply to comment #3)

I swear the abuse filter is the worst thing to happen to this project in recent
years.

I'm not sure I'd go quite that far; but fixing this would certainly not be a step in the right direction. WONTFIX.

  • This bug has been marked as a duplicate of bug 23087 ***