Page MenuHomePhabricator

Rollback to a flagged revision should flag the new revision automatically
Closed, DeclinedPublic

Description

Author: mike.lifeguard+bugs

Description:
When there are edits that get reverted with rollback, and the revision being reverted to is flagged to some level, the top revision should be automatically and silently flagged to the same level, as it is identical to the earlier, flagged, revision.


Version: unspecified
Severity: enhancement

Details

Reference
bz13751

Event Timeline

bzimport raised the priority of this task from to Lowest.Nov 21 2014, 10:07 PM
bzimport set Reference to bz13751.
bzimport added a subscriber: Unknown Object (MLST).

This is already done, unless the user doesn't have the right permissions. Otherwise, there are some exploits.

However, I'll probably make it at least basic reviewed.

Folloquing bug 13978 comment 30 I reopen this bug.

(In reply to comment #2)

Folloquing bug 13978 comment 30 I reopen this bug.

How is this different then that bug? It is a wontfix for the same reason since it's the same request (autoreview on rollback). Comment #1 from 2008 should be ignored.

If this is for an "autoreview on rollback" right to be available to user groups, then that would not be this bug.

(In reply to comment #3)

How is this different then that bug?

It applies to rollback only and not to reverts. RobLa said the summary was too wide.

It is a wontfix for the same reason since
it's the same request (autoreview on rollback). Comment #1 from 2008 should be
ignored.

If this is for an "autoreview on rollback" right to be available to user
groups, then that would not be this bug.

As I said in bug 13978 comment 22, I may agree, but I would prefer such right to be implicit in "rollback" right: otherwise, it should be added to all sysop and rollbacker groups and to global rollbacker, so I don't see the point of creating a new right.

What RobLa, Chad, and I discussed was always "autoreview on rollback" (not just any revert) and that is still a wontfix.

A new right, that when granted, lets someone do that is possible though.

Can you explain why you want to create a new right?

(In reply to comment #6)

Can you explain why you want to create a new right?

I'm not saying *I* want to, I'm saying it is acceptable if the behavior is really desired.

We don't want the software hardcoded to autoreview edits just because someone has 'rollback'. 'rollback' and 'autoreview' should be kept separate rights. Having one right do part of what another right does like this is usually bad design and historically always backfires later on.

(In reply to comment #7)

We don't want the software hardcoded to autoreview edits just because someone
has 'rollback'. 'rollback' and 'autoreview' should be kept separate rights.

That's exactly what we're asking, because we don't want to add "autoreview" to all rollbackers.

Having one right do part of what another right does like this is usually bad
design and historically always backfires later on.

I really don't understand this. "Autoreview" is about the flag of a new revision the user creates when he edits the page, isn't it? And according to http://www.mediawiki.org/wiki/Extension:FlaggedRevs#User_rights those revisions (the page says "edits", I suppose it's a typo) are marked as sighted.
But when you use rollback you restore the most recent revision apart from the rollbacked one[s]; currently, the "new" revision, identical to the restored one, is not flagged, but even if you had "autoreview" or "autoreview on rollback" (as you propose) that revision would be only "sighted", wouldn't it?
And why should a "sighted" recent version suddenly become unflagged, or even a "good" recent revision suddenly become only "sighted"? What has changed after the rollback?
Why should e.g. rollbacked vandalism require additional verification of the retored revisions?

What if a wiki community doesn't want rollback to make the result page revision have the same flags as the stable one for people with no actual review rights? Maybe they don't trust such people to only revert unacceptable content.

I don't want to "hard code" this behavior because it's a huge pain to pull it out if people on a wiki don't want it (requiring going to a configuration based strategy and transitioning over the existing wikis). I'd rather keep rollback and review-related rights separate in the "hard code" of the software, but leave it open configuration as needed.

Also, one could always make a "restorereviewtags" permission that fully uses the old tags of the previous stable version (thus a rollback to a "quality" revision itself becomes "quality").

(In reply to comment #9)

What if a wiki community doesn't want rollback to make the result page revision
have the same flags as the stable one for people with no actual review rights?
Maybe they don't trust such people to only revert unacceptable content.

I still don't understand: even if you rollback an acceptable edit, why should the previous revision change flag?

(In reply to comment #11)

(In reply to comment #9)

What if a wiki community doesn't want rollback to make the result page revision
have the same flags as the stable one for people with no actual review rights?
Maybe they don't trust such people to only revert unacceptable content.

I still don't understand: even if you rollback an acceptable edit, why should
the previous revision change flag?

If someone rolls back good edits to a "sighted" version, and this new version (with the same text as the last "sighted" version) is also marked "sighted" automatically, then stable version is considered "up to date" and does not show up as having anything to review (e.g. Special:PendingChanges). People may or *may not* want this behavior for users just because they have rollback.

There are probably additional reasons people might not want rollback and and reviewing rights to be mixed in a hard coded way; such things have a nasty habit of revealing themselves after code is already in use for a while.

(In reply to comment #12)

If someone rolls back good edits to a "sighted" version, and this new version
(with the same text as the last "sighted" version) is also marked "sighted"
automatically, then stable version is considered "up to date" and does not show
up as having anything to review (e.g. Special:PendingChanges). People may or
*may not* want this behavior for users just because they have rollback.

Well, if I understand correctly your point, this may be a problem only if the restored revision has some problems which the rollbacked good edit was trying to address, because in that case something has changed (the evaluation of the restored revision may change because we know something new about it); but such a use of rollback is considered an abuse everywhere, as far as I know: rollback should basically be used only for vandalism.

There are probably additional reasons people might not want rollback and and
reviewing rights to be mixed in a hard coded way; such things have a nasty
habit of revealing themselves after code is already in use for a while.

As I said above, unless some wiki uses rollback in a different way than the most common and correct one (which would be very strange, but I understand that we can't assume it won't happen) there shouldn't be any problem.
If you think that the technical tool of rollback is likely to be used with different meanings, then I think that you could create that "rollback-restorereviewtags" right, or even an additional "rollback-autoreview" right, and add it (or them) to the default rights of sysops and rollbackers, so that wikis which want to use rollback in a different way can do so, but most wikis are not annoyed by unexpected and inconsistent behaviours; this would close this bug.

(In reply to comment #13)

Well, if I understand correctly your point, this may be a problem only if the
restored revision has some problems which the rollbacked good edit was trying
to address, because in that case something has changed (the evaluation of the
restored revision may change because we know something new about it); but such
a use of rollback is considered an abuse everywhere, as far as I know: rollback
should basically be used only for vandalism.

Well, that too. But I was thinking of the fact that a good edit was reverted and "lost" unless someone decides to look through the page history, notices it, and restores it.

Closing this since 'rollback' won't imply 'autoreview on rollback' per above.

New bugs can be filed for things like:
(a) Site config requests to give a user group (like Rollbackers) 'autoreview'
(b) Requests for a new right that makes rollbacks autoreview the page (if rolled back to the stable version)

(In reply to comment #15)

(b) Requests for a new right that makes rollbacks autoreview the page (if
rolled back to the stable version)

Bug 27582.