Wikipedia talk:Selective deletion

RevisionDelete

Oversighters now have the ability to delete log entries and page histories using RevisionDelete. And soon this ability will be extended to all admins.

I think it's very important that a discussion be held about when it is and is not appropriate to modify log entries or page histories. For example, I personally don't think hiding "simple" vandalism would ever be appropriate (for example, "Got HAGGER?").

I foresee conflict between admins unless there are at least vague requirements for when it is appropriate to hide things and when it is not. Thoughts? --MZMcBride (talk) 20:12, 1 February 2009 (UTC)[reply]

Off the top of my head, I can think of three instances where hiding revisions should be allowed. Libel, links to virus-ridden websites, and page move vandalism, either of the HAGGER variety or "Admin Z lives at 123 West Street" (which often enough is not oversighted). NuclearWarfare _(Talk) 21:05, 1 February 2009 (UTC)[reply]

For the time being I think that the best option is to treat Revision deletion as restricted at the same level oversight is. Prodego ^talk 21:45, 1 February 2009 (UTC)[reply]

I don't see any need to extend it to administrators. I assume that oversights will retain the "Apply these changes to administrators and lock this interface" option on the interface? --Deskana (talk) 00:01, 2 February 2009 (UTC)[reply]

I think that it's obvious that it should only be used in cases where you would otherwise oversight. I also think that it's obvious that admins should not have access to this tool. John Reaves 00:39, 2 February 2009 (UTC)[reply]

No, admins should have access to this tool, but not the whole thing. Selective deletion is stupid, and we don't want that to continue when we have a perfectly good way to do it. Prodego ^talk 01:44, 2 February 2009 (UTC)[reply]

Where was the discussion that resulted in consensus to extend this to all admins ? I'm not aware that the problem is large enough to need over a thousand admins being able to do this. I feel most strongly that this should be a very restricted function and can't see a reason to extend it to us admins. Oversight only looks like a good solution. - Peripitus (Talk) 02:04, 2 February 2009 (UTC)[reply]

I should clarify. It'd be great if admins could use this tool, but I don't think every admin is capable of using it without abusing it. John Reaves 02:26, 2 February 2009 (UTC)[reply]

(EC) It would be a useful feature for getting rid of redirect revisions when doing history merges, where Page A is created as a redirect to page B, then the contents of page B are cut and pasted into page A. For an example, see my logs at War in Afghanistan (2001–present). Deleting the three edits would have been much easier, both from my point of view and from the server's point of view. I may be a perfectionist, but I really don't see the point of having overlapping redirect revisions in an article's history. The feature could also be used for getting rid of copyvio revisions, where one editor has dumped a copyvio into an article, but in that case, the deletion summary should be clear about where the copyvio was from. Selective undeletion can be incredibly painful, both for the user and for the server, and its use should be minimised. I can think of absolutely no reason for admins to hide page logs, now that oversighters can do that. Graham87 02:36, 2 February 2009 (UTC)[reply]

Admins already are able to do this, just in a less transparent, less server-friendly, and less user-friendly way. Mr.Z-man 06:44, 2 February 2009 (UTC)[reply]

Indeed. I think selective deletion should simply follow the same policy than full deletion. If you wouldn't delete the whole page because of these revisions, you shouldn't use revdelete. With a small trick: hiding revisions can sometimes be very effective against btards (who tend to vandalize once, then simply use the permalink and save). -- lucasbfr ^talk 10:25, 2 February 2009 (UTC)[reply]

PS: Anything that can prevent this ugliness is good. -- lucasbfr ^talk 10:28, 2 February 2009 (UTC)[reply]

I think admin revision deletion should be used in the following situations:

If an admin sees anything which should be removed by oversight, they should (after alerting oversight) delete the revision to reduce the damage.
Personal attacks and harassment, under some circumstances.
Edits which would be deleted under CSD G7 or U1, including edits to pre-existing pages.
Vandalism or spam being linked to, usually by the vandal/spammer.
Copyright violations which were added to a pre-existing article.
Edits by banned users.

עוד מישהו Od Mishehu 11:12, 2 February 2009 (UTC)[reply]

I'm not sure about item #6, but the rest looks fine to me. -- lucasbfr ^talk 12:12, 2 February 2009 (UTC)[reply]

OK, I'm an admin and maybe this tool might be handy. On the other hand, there's legitimate concern expressed above that not all admins would use it correctly. Admins already have the power (through selective deletion) to stop any immediate damage. After that, only folks with admin privileges can see problematic versions until they're oversighted. I think oversight could probably wait a few hours or days at this point. And if we can't trust admins enough during that waiting period to not abuse their abilities to restore sensitive material, should we then be giving them (including me) oversight privileges?

So I have a question -- do we not have enough people already authorized to handle this sort of thing? If not, do we need 1000+ more (admins) or just another 5 or 10? If we don't need more, then I suggest not extending it to all admins. If we just need a few, then perhaps add it to bureaucrats' privileges or give the power to checkusers; both these groups are much smaller and have been subjected to additional vetting.

If this is extended to all admins, then MZMcBride is right -- we need very specific rules.

So that folks like me that may do it once every 6 months are on firm ground and don't use the privilege inappropriately through well-meaning inexperience
To ensure abusive admins (yes, we do have some, let's admit it) don't abuse it, leading to more wikidrama.

My bottom line: expand this privilege very conservatively, if at all.--A. B. ^{(talk • contribs)} 17:29, 2 February 2009 (UTC)[reply]

I don't get it, it's already there for a long time (just check the deletion log, you'll see that more than half the restores are in fact selective deletion). I don't link to anyone, check it yourself. The aim is to stop killing the servers and risking to screw things up each time we do it. -- lucasbfr ^talk 18:59, 2 February 2009 (UTC)[reply]

I agree that we need very, very specific rules about this if all admins will get this tool. Otherwise various admins will start deleting simple instances of vandalism, or revisions that might possibly be, in a way, a BLP violation, etc. --Conti|✉ 19:30, 2 February 2009 (UTC)[reply]

Wikipedia is going to end up looking like a government document, full of "[REDACTED]" and blocked out text if all admins have access to this. John Reaves 18:47, 2 February 2009 (UTC)[reply]

All admins can already do this (as Mr. Z-man said above), and the sky hasn't fallen yet. --Kbdank71 18:52, 2 February 2009 (UTC)[reply]

Yes, currently it just disappears from the history and can only be seen through Special:Undelete; I'm not even sure if its possible for non-admins to know how many revisions were deleted. This just adds transparency to the process. Mr.Z-man 19:06, 2 February 2009 (UTC)[reply]

Currently, it is a bit of a hassle to do it, so it's less prone to abuse. If we make it super simple, that's where the problems come in. John Reaves 19:37, 2 February 2009 (UTC)[reply]

I wonder how many people are going to use this to cover up their mistakes, things they said and then later regretted (e.g. being a dick in an edit summary). A line between what appropriate use and what is historical revisionism needs to be drawn. John Reaves 22:29, 2 February 2009 (UTC)[reply]

There are more than enough admins willing to call people on it if they see this being misused. --Random832 (contribs) 03:47, 3 February 2009 (UTC)[reply]

How much to hide

One of my concerns it that if a user makes an edit like this:

(cur) (prev) 18:37, 2 February 2009 VandalUsername (Talk | contribs) (9,553 bytes) (→RevisionDelete: http://inappropriate.link) (rollback | undo)

With RevisionDelete, you'll be able to hide the revision content, the username, or the edit summary. To me, if anything must be hidden, it should only be the inappropriate edit summary, nothing else.

However, my concern is that in an attempt to deny the existence of vandals, some admins will want to hide the username or the entire entry. Can we all agree that doing so would be a Bad Thing? If the username is inappropriate enough to warrant hiding, it should be renamed by a bureaucrat (which takes care of the issue completely). --MZMcBride (talk) 20:14, 2 February 2009 (UTC)[reply]

Revision content deletion is essentially the same as oversight is now. Edit-summary deletion is obviously useful, as you say. Changing the username leaves a record in the log.

We need a strict understanding that information is not hidden gratuitously and that the default is to keep the information visible, not to delete it, but certainly all three are necessary.

[[Sam Korn]] ^(smoddy) 20:21, 2 February 2009 (UTC)[reply]

I'd prefer if people stopped deleting/restoring articles altogether. But that'd mean having approximately the same amount of "privacy" than before. Personally I can live with the username remaining visible when admins use the extension (we usually catch problematic usernames before they even hit save), but is it useful? -- lucasbfr ^talk 20:40, 2 February 2009 (UTC)[reply]

There are occasions where a username that is in itself not problematic might be problematic in the context of a particular page. [[Sam Korn]] ^(smoddy) 21:17, 2 February 2009 (UTC)[reply]

Eh? I can't think of any such examples. I mean, obviously User:IHateJews shouldn't be editing the "Israel" article, but that username shouldn't be in other page histories either.... --MZMcBride (talk) 22:00, 2 February 2009 (UTC)[reply]

To give an entirely hypothetical and made-up example, a username that was the real name of a pseudonymous porn star, causing potential real-life damage to that person. Unlikely, but not impossible. I agree that its use should be rare, but not, I think, forbidden. It is a policy-level concern, not a technical one. [[Sam Korn]] ^(smoddy) 22:30, 2 February 2009 (UTC)[reply]

Quick question

I'm not sure if this is the right place to ask, but I don't know where else to ask, so here I go. Does this feature exist in addition to oversight, or does it replace oversight? And if it is the latter, is there still the option to completely erase a revision (not even leaving a "This revision has been oversighted" message behind)? --Conti|✉ 20:37, 2 February 2009 (UTC)[reply]

Both extensions are currently enabled for oversighters. If everyone used the new extension and oversight left a visible trace, that would greatly improve transparency and the level of trust users have in it though IMO. -- lucasbfr ^talk 20:42, 2 February 2009 (UTC)[reply]

Agreed. Where is the switch to turn off oversight, then? :) --Conti|✉ 20:47, 2 February 2009 (UTC)[reply]

I for one am still figuring out how this works -- I have no objection at all to Oversight's being turned off, but it might be sensible to let us have a few days to get used to it first with a safety net! [[Sam Korn]] ^(smoddy) 21:15, 2 February 2009 (UTC)[reply]

No, no, no. You all are horribly confused. :-) This new thing, "RevisionDelete" (which also hides log entries) is not an extension. That's the entire point. mw:Extension:Oversight is the extension; it is (was) a poorly-written temporary hack while this new system was being written. The Oversight extension can't be disabled until (or if) the old data is transferred to the new system. See also: http://lists.wikimedia.org/pipermail/wikitech-l/2009-January/041192.html and subsequent replies. --MZMcBride (talk) 21:58, 2 February 2009 (UTC)[reply]

I think I kinda get it now. :) What I'd still like to know is whether it will still be possible to entirely remove revisions (with no hint that they ever existed in the first place) or not. Will it be possible with the new feature? Or will it be possible because the oversight extension won't be turned off? --Conti|✉ 22:07, 2 February 2009 (UTC)[reply]

It's my understanding that the previous implementation (using Extension:Oversight to remove the edits entirely from the page) will stop completely. We will no longer "disappear" edits from the page history, in favor of marking them deleted (either at an oversight-only level or at an admin-only level).

So basically RevisionDelete will replace Extension:Oversight entirely. But this can only happen iff the extension's dependencies are completely removed. Currently there's a separate table full of revisions from the past two and a half years that have been removed from the revisions table. Those would need to be migrated first. The second catch is this. Basically for some unknown reason RevisionDelete is apparently using Extension:Oversight's log. There's a debate (on the mailing list) about whether the log of deleted log entries, etc. should be visible to all users or not (see also: this). --MZMcBride (talk) 22:25, 2 February 2009 (UTC)[reply]

Thanks for the answer. That clears a lot of things up (for me at least). --Conti|✉ 22:36, 2 February 2009 (UTC)[reply]

Just a clarification, Special:RevisionDelete uses the user's regular log if the option to hide the revision from admins is not checked. Since I haven't so far needed to, I have not used Special:RevisionDelete to hide a revision from admins (nearly equivalent to oversighting), but I assume that is what gets logged to the oversight log. But also with the current (I'm assuming temporary) configuration even when the option to hide the revision from admins is not set, admins still cannot see what was hidden, just the fact that it was. - Taxman ^Talk 23:08, 2 February 2009 (UTC)[reply]

I think you're right, you're using the 'lesser' of the two suppression levels (the level that, if the feature were made available to admins, whould be given to us). I think it would be advisable to work under the assumption that administrators will at some point be given the ability to view revisions that are "hidden" but not "suppressed". Happy‑melon 00:05, 3 February 2009 (UTC)[reply]

A few questions

Who should have access to this? Also, if we do give this to all admins will we let all admins see what was removed? JoshuaZ (talk) 22:19, 2 February 2009 (UTC)[reply]

The section directly above, specifically the links to the mailing list, give a bit more insight into these questions. --MZMcBride (talk) 22:27, 2 February 2009 (UTC)[reply]

Should admins have this ability?

I can honestly say that when I started this discussion, I was under the assumption that this 'feature' would definitely be enabled. However, after giving more consideration to its use and reading some of the comments here, I'm beginning to think that giving this ability to all admins may be a very bad idea.

Oversight (the extension) has been abused previously. This tool would likely be equally abused, however, if only oversighters had the ability, it would decrease the likelihood of abuse as oversighters are more likely to stick within the bounds of the oversight policy and oversighters are more likely to take an unbiased look at various log entries / edits. That is to say, I don't want admins trying to remove every instance of vandalism or even "extreme" vandalism simply because it's now easier to do.

The type of extreme restraint required for this power seems outside the scope of what (at least some of) our 1,000 active admins can handle. Thoughts? --MZMcBride (talk) 00:55, 3 February 2009 (UTC)[reply]

No, administrators shouldn't have this ability. -- Gurch (talk) 00:57, 3 February 2009 (UTC)[reply]

I'll second that. Synergy 00:58, 3 February 2009 (UTC)[reply]

Agreed. –Juliancolton ^Tropical ^Cyclone 01:02, 3 February 2009 (UTC)[reply]

When you say "this feature" what do you specifically mean? There are about 8 different pieces that are involved. Log hiding, which has three parts, hiding the action and target, hiding the comment, and hiding the user doing the action. There's also revision hiding which has three parts, hiding the text of the revision, hiding the edit summary, and hiding the editor's IP/username. Then for Oversighters both of these can also be hidden so that admins can't see them or reverse them. I think it's most clear that last option should not be given to all admins. I think revision hiding being given to all admins should not be so controversial since if all three options are selected it is equivalent to the current way of deleting revisions except the edit history of the page retains a record of the fact that something has been deleted. That's more transparent than the current method. Also not all three options have to be checked, just the one that is needed. If the IP or username doesn't contain the offensive information, that need not be hidden. Then we retain a record of that user doing something nefarious, but we don't have to publicize what specifically. The other more transparent part of this new system is that these changes of revision visibility can be reversed with less hassle so it is less likely to be abused. If we want the least controversial option, then consider only giving all admins the ability to delete the text of a revision. It's very hard to abuse that. - Taxman ^Talk 01:24, 3 February 2009 (UTC)[reply]

No, it's very easy to abuse it to hide evidence of comments that you later regret making. Oversighters have done so before -- Gurch (talk) 01:44, 3 February 2009 (UTC)[reply]

Unlike the current system of selective deletion or oversight, the revisions are left in the history, but with the deleted parts hidden and there's a public log of non-oversight-level deletions. It would be a lot more obvious if someone deleted their own comment from a discussion (and it would only work if they reverted it before anyone else made an edit to the page). Deletions on watched pages also show up in watchlists. It would certainly be possible to abuse, it, but ... the same is true of all the admin tools. That's why we only give them to admins. If it wasn't possible to abuse it, we could safely give it to everyone. Mr.Z-man 01:58, 3 February 2009 (UTC)[reply]

No, an oversighter hid 2 edits another editor made. Lets not bother with the reasons why—that's another discussion but claims of regret were made by third parties. Also the difference between Oversight and Special:RevisionDelete apparently needs to be pointed out again. Oversight can't be reverted without developer intervention, while Special:RevisionDelete if done by an admin can be reverted either by other admins or just oversighters depending on the config. That and what Mr.Z-man mentioned makes this significantly more resistant to abuse than you are claiming. Don't get me wrong, I think whatever the community wants is fine, but I think people should think about all facets of this before judging it. - Taxman ^Talk 03:02, 3 February 2009 (UTC)[reply]

Note to Taxman...perhaps I am understanding better. Are you saying that if we turn on the rev delete option only, then what we get is a more featured version of what we can do now (by deleting the whole page and selectively undeleting revisions) with it shown in the page history what has been done. If am reading this all correctly then I think this limited version is a good idea for admins..prevents the disruption of del+selective undel of a page with many revisions that is highly trafficed just to remove something that needs to go. - Peripitus (Talk) 03:15, 3 February 2009 (UTC)[reply]

Yes, but at the moment, the hassle and visibility of deleting and restoring the whole page (and the impossibility, for large pages) puts them off doing things like removing comments they don't like by users they've just labelled as trolls -- Gurch (talk) 03:27, 3 February 2009 (UTC)[reply]

Yeah basically. If we turn on revision delete and allowed admins to hide the revision text, username, and edit summary (all three parts for article revisions). That would be close to the old way, but the difference from the old way would be that the article's edit history would show something was missing and it would be logged publicly (separately) who hid the edit. If we turned on revision delete for all admins for hiding the text of the revision but nothing else that would actually be more limited than what the current delete and selective restore does because the username and edit summary of the original offending edit would still be shown in the article history. Either way as mentioned I think there are many levels that reduce the likelihood of what Gurch is concerned about and make it easy to discover and reverse if it did occur. - Taxman ^Talk 03:46, 3 February 2009 (UTC)[reply]