Wikipedia talk:Revision deletion/Archive 1

Title
Rename to "Criteria for Redaction" as a commonsense title and more in line with other page titles? FT2 (Talk 11:56, 25 May 2009 (UTC)
 * Seems fair, CRD may be a good acronym ;). I removed the "using RevisionDelete" in the title, seems a bit redundant. ~ fl 09:15, 26 May 2009 (UTC)
 * I agree, nice acronym, and CamelCase is not a Good Idea in page titles (consider Flagged revisions vs mw:Extension:FlaggedRevs). Happy‑melon 09:50, 14 June 2009 (UTC)
 * I've moved the page to "Revision deletion". ~ fl 11:13, 14 June 2009 (UTC)

Advertising the proposal
These are some places where the proposal could be advertised:
 * WP:VPP
 * WP:VPT
 * WP:AN
 * WT:OVERSIGHT
 * Requests for comment/Policies
 * Wikipedia talk:Selective deletion

"Blocks declared invalid by the Arbitration Committee"

 * 1) Blocks declared invalid by the arbitration committee. This should also include the removal of the corresponding edit summary included with the block and ensure that the hiding summary includes reference to the block being invalid.

Absolutely not. --MZMcBride (talk) 00:08, 10 June 2009 (UTC)
 * I'm restoring it. I would like to just clean it up then get the community views on the policy itself when it's ready. ~ fl 00:16, 10 June 2009 (UTC)
 * What? There is absolutely no consensus for this anywhere. Why would you put it in the middle of a mostly benign document? --MZMcBride (talk) 00:18, 10 June 2009 (UTC)
 * Equally, there is no consensus anywhere for removing it. This policy hasn't yet been presented to the community for any consensus gathering (there is no consensus at all), so that argument is invalid. ~ fl 04:01, 10 June 2009 (UTC)
 * You've lost it. --MZMcBride (talk) 04:15, 10 June 2009 (UTC)
 * Block log clean up is a disaster waiting to happen honestly. I imagine much drama spilt over this, and it makes it more difficult to track admin behavioral patterns. --Tznkai (talk) 06:07, 26 September 2009 (UTC)


 * 1)  Purely disruptive material, for example where the entry contains (and is apparently posted to promote) blatant provocation, warring, or perpetuation of an off-wiki battleground within the project, including allegations, grossly inappropriate threats or attacks, browser-crashing or malicious HTML, and the like, being of little or no relevance or merit within the project. This is distinguished from G2 in that although the material is not necessarily grossly offensive, it is clearly harmful to the project, has little project benefit, is usually not posted by established respected editors, and merely encourages the hostile use of the project as a battleground.

This also needs discussion. --MZMcBride (talk) 00:39, 10 June 2009 (UTC)
 * Again, all in due time. The policy will soon be presented to the community (after it has been brought up to a grammatical and style standard suitable for policy, not before), so they can decide on the validity of each of the criteria, until then, extra discussion is pointless. ~ fl 04:01, 10 June 2009 (UTC)
 * It's called anticipating problems and working to address them proactively. --MZMcBride (talk) 04:15, 10 June 2009 (UTC)

What, and deprive us of Jimbo's glorious rapsheet? But seriously, I agree with Tznkai, deleting block logs is a bad idea. Administrative pardon will delete important fingerprints that may make the future more difficult. Valley2 city ‽ 06:52, 9 October 2009 (UTC)

Administrators
There's a theme throughout the current version of this page that administrators will eventually be getting this ability. Is there evidence (links, etc.) to support this anywhere? --MZMcBride (talk) 00:23, 10 June 2009 (UTC)
 * Well the entire point of the policy is to govern administrative use of such a tool. The idea is, if there is consensus for the policy, then there is defacto consensus for the feature itself. So if the policy is adopted, a bugzilla request will be make to enable to tool for administrators. ~ fl 04:10, 10 June 2009 (UTC)
 * Right. Did that already. There's uncertainty about whether or not to give this to admins. --MZMcBride (talk) 04:14, 10 June 2009 (UTC)
 * What were the main objections? ~ fl 04:16, 10 June 2009 (UTC)
 * See Wikipedia talk:Selective deletion. The discussions there were inconclusive. The community will have to decide if it should be enabled for admins or not. This part is not settled. Cenarium (talk) 13:45, 13 June 2009 (UTC)
 * It also requires the sysadmins agreeing to follow community consensus.... --MZMcBride (talk) 14:24, 13 June 2009 (UTC)
 * Based on the design of the system, I thought it was understood that it would be. If it was only intended to replace oversight, there wouldn't be 2 different levels of deletion (which are almost equivalent when admins don't have the necessary right). One level of deletion was intended to replace the oversight extension and the other to replace the current revision deletion hack. Mr.Z-man 15:18, 13 June 2009 (UTC)
 * There's a possibility that administrators could have the ability to see suppressed content, but not the ability to suppress content. Personally, that's the system I would prefer given the past experiences we've had with administrators. --MZMcBride (talk) 19:47, 13 June 2009 (UTC)
 * I think that possibility has come and gone; oversighters now use the supression option of RevDel instead of the old oversight interface, and no one will ever agree to let admins view the nasties that the oversighters are hiding these days. ~ fl 09:07, 14 June 2009 (UTC)
 * No, my point is that RevDel has 2 levels of deletion. Oversighters have access to both and admins would theoretially have access to only the "lower" level. Therefore it can replace both oversight and normal revision deletion. Mr.Z-man 14:56, 14 June 2009 (UTC)

There is no reason why Administrators' should not be given access to this tool if there is the appropriate consensus to do so. It is no more controversial than other configuration requests; the issue of "sysadmins agreeing to follow community consensus" is neither new nor restricted to this issue. Happy‑melon 09:48, 14 June 2009 (UTC)
 * We have users arguing that "blocks declared invalid" can simply be erased. I'm a bit hesitant. --MZMcBride (talk) 16:09, 14 June 2009 (UTC)
 * Now why shouldn't blocks declared invalid be erased? The whole definition of invalid is that they maintain no merit besides a permanent and unjust stain on someone's record. ~ fl 09:36, 15 June 2009 (UTC)
 * I agree that there are issues to be worked out, and they certainly must be worked out prior to deployment. But the presence of one or many issues requiring further discussion is not a reason, as you seem to be suggesting, that activating the feature is Not Possible. You know our development process better than this, MZ: we don't get notifications "feature X will be activated in three months, write appropriate policy". To suggest that we need such notification is disingenuous. Happy‑melon 17:46, 15 June 2009 (UTC)
 * There are serious suggestions that this tool should not be given to admins at all. Search around Bugzilla, you'll find it. Perhaps "don't shoot the messenger" is the appropriate phrase here? (Though I will admit that I mostly agree with the argument to keep it in limited hands, so perhaps my bias is shining through. Who knows.) --MZMcBride (talk) 04:06, 23 June 2009 (UTC)
 * I have to say I'm not seeing them; maybe I'm searching for the wrong thing, could you point me in the right direction? Certainly I haven't heard anything to suggest that the insurmountable obstacle to enabling this feature for admins, whatever it might be, exists. Happy‑<b style="color:darkorange;">melon</b> 21:36, 26 June 2009 (UTC)

Edits
I've picked up the drafting work on this page, quick summary below:
 * 1) Mention "oversight"; even if the policy isn't about oversighting, it's still about the tool and saying "the tool can be used for oversighting (which is separately covered)" will be helpful to a reader.
 * 2) Update text representations of RevDelete to actual screenshots.
 * 3) Added back (and edited) "current status" section as a notice - while RevDel is being made ready for admin use, it is still a new function and many users reading the page will want to know the current position with it. This would be deleted when RevDel goes "live" for admin use so it's a temporary notice only, but useful.
 * 4) The header text for "Criteria for redaction" was very overweighted; it could be said much simpler. It also covered references to "multiple criteria" which is now redundant. Redrafted.
 * 5) Moved guidance notes for G2 and G4 to their own section, "Notes on use". (As footnotes they would not usually get read and people would miss them, included in the criteria they would be too wordy.)
 * 6) Clarified G2 a bit.
 * 7) Created section "other uses" to separate usual use from exceptional uses, and amended "Arbcom may hide block log entries" to the more general wording at #S1 ("Deletion mandated by a decision of the Arbitration Committee") which covers the matter better.
 * 8) removed a lot of wordiness that was still there from the first draft - Editing logged out is covered under privacy (since inadvertant disclosure of IPs is oversightable), and "hiding edit summaries" and "hiding log entries" can mostly be removed too as being redundant.

I have left the text at "Hiding revision text" and "Hiding usernames" for now, since these might need other eyeballs to look at them first. The section "General guidance for administrator use of RevisionDelete" could also probably be handled better - the wording is fine, but the positioning and handling might be capable of being better.

FT2 (Talk 13:21, 13 June 2009 (UTC)


 * Update:
 * I have reorganized the page from above, with almost no change of text: Status at the top in a noticebox (keeps main page simpler), and technical notes after the actual criteria, not at the top (works a lot better that way). FT2 (Talk 15:09, 13 June 2009 (UTC)

Possible G7
Would anyone object to adding a general criteria G7 into the draft:
 * Denial of recognition of certain banned users (subject to review by the community). A minority of banned users persist in editing in a problematic manner. RevisionDelete may be used to deny recognition by redacting their sock-puppet edits and actions. An administrator wishing to use RevisionDelete for this purpose should open a section on WP:ANI stating they have begun denying recognition to the user, so that other users may give their views.

The aim of this would be that WP:DENY via RevisionDelete can be enacted effectively, but not without careful thought. We usually trust admins, and if an admin feels there is a good case for denying recognition to a repeat-offending banned user (per usual deletion policy or WP:RBI norms) then this allows them to do so - but they must note the treatment on ANI so other admins can review their decision and that users' conduct if they wish. FT2 (Talk 17:29, 13 June 2009 (UTC)


 * I certainly object. There needs to be a re-examination of why this tool exists and why it has had such limited parameters previously. It isn't appropriate to use it "pretty up" history pages or hide every instance of vandalism (not that that's what you're suggesting necessarily). Attempting to remove edits by "banned users" is a fruitless endeavor that will ultimately do far more harm than good. Egregious vandalism should be removed. Non-public personal information should be removed. Any and all productive edits should never be obfuscated, no matter who is making them. --MZMcBride (talk) 19:44, 13 June 2009 (UTC)


 * If it were not being used in accordance with communal wishes, then the wordings "open a section on WP:ANI" and "other users" would rapidly say so. The target here would be "certain banned users", not "all banned users". There are certain banned users where Wikipedia is used as a trophy board or means of disruption, rather than productive use, and redaction may better persuade them to cease. FT2 (Talk 21:48, 13 June 2009 (UTC)


 * Acting differently with "certain banned editors" and codifying a special case for them in policies / guidelines is giving them far more recognition than anything else. How does "deny recognition" apply when you're actively promoting the idea that we give them (and their edits) special status? --MZMcBride (talk) 22:22, 13 June 2009 (UTC)


 * Starting a thread on ANI is pretty much the opposite of denying recognition. A thread on ANI is pretty much the most visible place someone could post without editing an interface message or a protected page. Application of WP:DENY should be discreet. Mr.Z-man 22:31, 13 June 2009 (UTC)
 * I agree with Mr.Z-man, if we're going to list it anywhere, it needs to be a more discrete location. ~ fl 02:07, 14 June 2009 (UTC)


 * How about this approach - an admin who feels a specific banned user should have recognition denied this way, should privately notify the functionaries list (or Arbcom) of their view, and if others there agree then they would be considered to have made adequate disclosure. The point being we don't want this used on every case; equally it is appropriate for a few cases and the comments above are right insofar as it is best not asked on a public wiki page:
 * Denial of recognition of certain banned users (2). A minority of banned users persist in editing in a problematic manner. RevisionDelete may be used to deny recognition by redacting their sock-puppet edits and actions. An administrator wishing to use RevisionDelete for this purpose should check with [the functionaries list | the arbitration committee] and should state "WP:DENY as discussed with (list)" or "...per Arbcom" in the reason.
 * FT2 (Talk 11:08, 14 June 2009 (UTC)


 * WP:CSD is "user request". I assume you mean G5? <b style="color:forestgreen;">Happy</b>‑<b style="color:darkorange;">melon</b> 17:42, 15 June 2009 (UTC)
 * No, I mean "Revdel draft#G7", not "CSD#G7" :) FT2 (Talk 11:59, 16 June 2009 (UTC)
 * Eek. That's certain to be a source of much confusion.  Is there a leter series we can justifiably use that doesn't conflict with the CSDs? <b style="color:forestgreen;">Happy</b>‑<b style="color:darkorange;">melon</b> 13:01, 17 June 2009 (UTC)
 * R? ~ fl 23:07, 17 June 2009 (UTC)
 * Taken. The CSDs have claimed series A, F, G, I, P, R, T, and U. Leaving us to choose from BCDEHJKLMNOQSVWXYZ.  <b style="color:forestgreen;">Happy</b>‑<b style="color:darkorange;">melon</b> 08:10, 18 June 2009 (UTC)
 * S? For selective deletion? ~ fl 10:03, 18 June 2009 (UTC)

ArbCom formulating their own policy?
I was talking to Prodego (I think it was him) on IRC, and he thought that ArbCom was actively developing their own policy for RevDelete for admins. Anyone know anything about this? ~ fl 02:09, 14 June 2009 (UTC)
 * Why should they be? RevDelete for admins is not a privacy tool; it is not within the mandate of the Arbitration Committee to control its use.  They need to be consulted to the extent that they need to ensure that any admin use of the feature does not conflict with Oversighters' use of it for privacy issues, but that's the limit. <b style="color:forestgreen;">Happy</b>‑<b style="color:darkorange;">melon</b> 09:46, 14 June 2009 (UTC)
 * Suppression that is equivalent to Oversight is completely within ArbCom's scope (or the Audit Committee's). For what it's worth, I think Prodego is a bit mistaken. I don't know of any real efforts on the part of the Committee to draft a policy in this area currently. --MZMcBride (talk) 16:28, 14 June 2009 (UTC)
 * I don't either, for what its worth. FT2 (Talk 17:07, 14 June 2009 (UTC)
 * I agree, MZ, but that is not what is being considered. The level of suppression that <tt>hiderevision</tt> provides is exactly equivalent to <tt>delete</tt> and <tt>viewdeleted</tt>; both of which are very clearly outside ArbCom's remit, as regular cleanup activity. I'm sure you're aware of the distinction. <b style="color:forestgreen;">Happy</b>‑<b style="color:darkorange;">melon</b> 17:41, 15 June 2009 (UTC)
 * /me blinks. Did I suggest otherwise? --MZMcBride (talk) 04:03, 23 June 2009 (UTC)

Ugh, a bit late now. I mentioned that Arbcom was looking into the creation of a policy, not that they were going to create one themselves or anything like that. Prodego <sup style="color:darkgreen;">talk  19:02, 3 November 2009 (UTC)

Bloat?
I don't really disagree with much on this page, but I think most of it is unnecessary. First, there is clearly a new interface, so the overview and "Technical details" sections are useful, but probably belong on the how-to. I would note that selective deletion has already been in use for years, so the functionality being added here is mostly something we can already do and have done without extra policy. I'm not sure why we would need seven new deletion criteria for this, when we seem to have few problems with current practice. The criteria do seem like mostly restatements of some speedy deletion criteria. Before reading over this page, my idea was more that we would just add a note to the deletion policy saying like "any revision content, edit summary, or username that fits the current criteria for speedy deletion but is so offensive or harmful that reversion is insufficient may be deleted." This would cover outing, libel, copyright infringement, and so on. In fact, we could already add something like that to the existing policy, since selective deletion doesn't seem to be addressed in policy yet, anyway. Dominic·t 03:56, 23 June 2009 (UTC)

Community consultation
I would invite all users to read the suggested revision deletion policy, my statement and FAQs on this talk page and then discuss on this talk page the merits of ratifying the corresponding policy, and subsequently enabling this feature as suggested. Thanks. ~ fl 06:22, 26 September 2009 (UTC)

The revision deletion functionality has been enabled on the English Wikipedia for those users in the oversight group for some time now, and is now used primarily instead of the original oversight extension for hiding edits containing violating material (a full comparison between the two different tools can be seen here). Currently, the administrative community has to resort to the complex and destructive (not to mention relatively undetectable) action of deleting and then subsequently undeleting a page if they wish to hide a revision. This method has a many problems associated with it, such as; These three problems above, and many more, are fixed by the use of the new revision deletion function instead of the old one. In summary, revision deletion provides: over the normal delete/undelete method
 * In the interval between the deletion of the page and the subsequent undeletion (which are separate and non-connected actions from the software's point of view), the page that is having a revision hidden does not exist. If the administrator is not very quick when performing the delete/undelete, a user who visits the page will be presented with the "This page does not exist" error.
 * After the deletion is performed, there is no record in the history log of the revision existing. Furthermore, people performing diff comparisons on the revisions can be given the impression that users performed edits that they in fact did not perform, see here for an example.
 * Because of some limitations built into MediaWiki, pages with more than 5000 revisions cannot be deleted by administrators. Because full page deletion is required to hide one revision, pages with many revisions cannot have their revisions hidden by administrators with the delete/undelete method. The Catch-22 here is that the pages with the most revisions are the pages most likely to require revision deletion, and currently administrators cannot hide revisions on the biggest pages. Pages such as Jesus, World War II and Barack Obama cannot have their revisions deleted.
 * More transparency
 * More flexibility
 * More efficiency
 * More oversight
 * Easier reversibility
 * Less server strain
 * Faster deletion times

Frequently asked questions
Q. If admins get revision delete, does this mean they will be able to see revisions hidden by users with the oversight permission?
 * A: No. The revision deletion system has two tiers, one for oversighters, one for administrators. Oversighters can choose, when hiding an edit, if they also want to hide the edit from administrators. Even though the revision deletion permission is not enabled for administrators, the oversights have ensured that offending edits are hidden from those users with the lesser revision delete permission (administrators) in case the feature is enabled.

Q. Is it easy to unhide a revision once hidden?
 * A: Yes. Unlike the Oversight extension, it is trivial to unhide any revision and vice versa.

Q. The revision deletion permission isn't even enabled for admins yet, why are we discussing how it should be used?
 * A: In short, permission changes should be driven by community consensus, not the other way round. In long, by creating (and agreeing on) a policy that defines the administrative use of revision deletion, we are de facto declaring to the developer community that we wish this function to be enabled for administrators.

Q. If admins get revision delete they will abuse the permission / This will cause less oversight etc.
 * A: While technically not questions, this type of statement seems to make its rounds in debates such as this. What people fail to realize about the revision deletion permission is that it is hugely more transparent compared to the normal deletion function. The current revision deletion method for admins leaves no trace in the history page that a deletion has occurred (much like old oversight), and the only way to determine if a deletion has occurred is to check the deletion log for the page. The new revision deletion function, on the other hand, clearly displays an entry in the history log, showing where the revision was, and what parts of it are now missing.

Discussion

 * What about attaching it to a separate user group? — V = I * R  (talk to Ω) 05:57, 26 September 2009 (UTC)
 * Adding more layers to the bureaucracy would likely make this highly inefficient. Plus, administrators have a bad, hacky, opaque version of the tool already, why would we force then to resort to using it when they have to, instead of providing them with the better version, ~ fl
 * I figured that removal of the existing tool was basically a given already, but what do I know? :) Assuming that the existing functionality is unavailable, creating a new usergroup and assigning this tools access through that seems to be an obvious answer to... well, all of the below. — V = I * R  (talk to Ω) 09:15, 26 September 2009 (UTC)
 * The existing tool is actually just the regular deletion mechanism built into MediaWiki. So unless we want to remove the admin right of being able to delete an entire page, we cannot remove this "hack" without fundamentally redesigning the way that page deletion works in the software. Our only option is to provide a better option (RevDel) and encourage admins to use it (which I'm sure they will, the current method is such a pain). ~ fl 10:23, 26 September 2009 (UTC)
 * Its too powerful a tool to be given to the whole admin corps. Hidden revisions cannot be revealed later easily so this needs to stay in a small group who will use the tool judiciously and only when there is no doubt about it needing to be used. Far too many bad guys have got adminship for us to trust this widely. Imagine if Archtransit had been able to hide revisions and the damage that could cause.  Spartaz Humbug! 06:03, 26 September 2009 (UTC) I'm obviously confused about what im on about but if you can reverse the deletions then I can't see the real harm. Spartaz Humbug! 10:43, 26 September 2009 (UTC)
 * The new tool has more oversight, not less that the current revision delete functionality of admins. Archtransit would have done less damage if he had had this tool and hypothetically used it. ~ fl 06:07, 26 September 2009 (UTC)
 * I'm not exactly sure what you mean by "cannot be revealed later easily" - revisions hidden using this can be trivially unhidden. Mr.Z-man 06:16, 26 September 2009 (UTC)


 * I like the idea of rolling out admin level RevDelete to administrators at large, especially since on normal pages, its similar in effect to delete en mass and selectively restore except for more convenient for both little-o-oversight and deleter. Proper application of RevDelete can reduce the strain on oversight as well. I am however, considerably more reticent about revdelete and block logs. That seems like a brave new world that we are not yet prepared for.--Tznkai (talk) 06:05, 26 September 2009 (UTC)
 * That was considered, and removed from the policy. If this policy was ratified, that would be prohibited by merit of not being in the the permitted uses. ~ fl 06:10, 26 September 2009 (UTC)
 * Not the way it works sadly. Aside from the likely IAR insanity, any admin thats human is going to see the shiny new button while viewing a block log and go "huh. What does THAT do? We do not, and cannot, expect all 1000ish active admins to somehow wander their way across the policy discussions that prohibit block log tampering. I would be far happier to support of block logs were taken off the table at the technical level.--Tznkai (talk) 06:37, 26 September 2009 (UTC)
 * While we cannot expect policy to be read in advance, we CAN expect it to be read before the action is committed if we warn about it clearly. Modify the MediaWiki text presented after you press show/hide to say words to the effect of
 * "Here's a link to the policy. If you are at all unsure what you are about to do, DON'T!!! ...go read it first. You CAN do X Y and Z. You CANNOT do P Q and R, and if you do any of the prohibited things you may summarily be desysopped. We mean it. Sincerely, your pals at ArbCom. PS: Ixnay on editing this text so you can claim you didn't know."
 * where P Q and R are block log hiding or whatever. As policy changes, change the displayed text. Use giant red blinking text or whatever necessary. Note: that's probably not the exact wording I'd advocate using, it may need minor (!!) tweaking. This is why they don't let me write documentation for end users. ++Lar: t/c 13:52, 26 September 2009 (UTC)

I think it is fundamentally a bad idea to allow admins to delete revisions like this. I think the current user rights surrounding data suppression should be granularized (bug 19199), giving admins the ability to see deleted content, but not delete it. While it's certainly true that admins can currently use hacks to "selectively delete" revisions from a page history, that is not an excuse to encourage the behavior. Admins cannot be trusted with this ability, period. They will misuse it (intentionally or not), just as admins have misused countless other features (e.g., "move without leaving a redirect"). --MZMcBride (talk) 06:15, 26 September 2009 (UTC)
 * As the functionality of this extension can be achieved by any admin right now, I see this as an increase in transparency and accountability, and is easier for admins. I agree with Tznkai that block logs should not be included for admins. Kevin (talk) 06:20, 26 September 2009 (UTC)
 * Will there be any way to check which admin deleted a revision? This seems essential for accountability. --<b style="color:#3773A5;">Cyber</b> cobra (talk) 06:51, 26 September 2009 (UTC)
 * Revision deletion is logged just like normal deletion. Since it has already been in use by oversighters for some time, you can see how the log looks at . Dominic·t 07:29, 26 September 2009 (UTC)
 * I quite often do history purges for articles with copyright violations. Revision deletion would make that job a lot easier for me. So big support here. Garion96 (talk) 08:00, 26 September 2009 (UTC)
 * Strong support for making admin-level revdelete available to administrators. Note that this is NOT the same as making the suppression tool available - that would stay "oversighter only" whether or not this goes ahead. My rationale:
 * Due to a lot of work by the devs, RevDelete is now very stable, in use for oversight for over 6 months, and the major bugs now seem to be fully addressed.
 * It allows administrators to handle matters that are within the spirit of current deletion policy, but which they cannot handle at present (due to limitations of their current deletion tools).
 * It is more transparent, communally reviewable, and easier to reverse than current deletion tools, and there is no significant disadvantage, nor need for a separate usergroup. (I'll expand on that if asked, for those who don't know RevDelete well).
 * There will need to be a very strict norm on some things (block log redaction by arbcom agreement only? tool only to be used in certain very well defined circumstances?) but that's something we already have for admins in other tool areas; it's a point to add into any draft policy.
 * The project will benefit from it, in my view. FT2 (Talk 09:07, 26 September 2009 (UTC)
 * Yeah, suppression is for oversighters. This is about giving admins the ability to zap a single rev without doing a huge and tricky job of delete-all/restore-all-but-one - David Gerard (talk) 09:13, 26 September 2009 (UTC)


 * May I just say, as an admin who's had to do the delete-all/restore-all-but-one dance to remove a single problematic revision from public view that's nevertheless not oversightable ... HELL YES. PLEASE. It'll be a function admins have begged for for years, it'll be logged and accountable, it'll be the removal of a major headache. Doesn't need a separate group, admins are already trusted with the power to delete all revs after all - David Gerard (talk) 09:13, 26 September 2009 (UTC)

Please enable it. I thought the discussion here would be the policy for when administrators would use it. For what I see there would be a subset of the speedy delete criteria, the extra things not in the policy already:
 * delete redirect page after history merge G6
 * Sole contributor request - G7
 * perhaps remove id of some one operating while not logged in G7
 * delete illegal material (not just copyright violations)
 * things like office actions that are not eg OTRS reqs (seems to be no speedy tag for this) G7
 * ?Admin does something embarrassing and wants to cover the mistake and correction, could be a user request G7
 * delete newest revision of a file without a revert. Already old revision of files can be deleted, but for some reason not the latest one. This may have to happen if one bad version updates an otherwise good file. F5 or F1
 * remove banned user edit G5

This gives us a horizontal cut through time, but not a vertical one, so if there was a piece of content that needed to be removed from 100 revisions this could not be cone with out removing all those revisions. This would be an attribution nightmare. So I suppose we need a policy to cover how attribution is done for all the deleted revisions if something of consequence to the article happened in that interval. I am unconvinced about its use for removal of disruptive or vandalism edits. Graeme Bartlett (talk) 09:56, 26 September 2009 (UTC)
 * This sort-of is supposed to be a discussion on the policy as well as the feature, the policy that would be adopted is something similar to that at Revision deletion. ~ fl 10:48, 26 September 2009 (UTC)


 * I've oversight elsewhere so I've seen how this feature works and am a long time user now. It's nifty, a vast improvement over mass deletion (as well as a mass improvement over the old oversight). It'd be swell if it was turned on here. ++Lar: t/c 13:52, 26 September 2009 (UTC)

I strongly object the proposed 2nd, 3rd, and (to a lesser degree) 5th General Criteria. I know that this is an American viewpoint but those all strike me as being too close to censorship and suppression of ideas. Genuinely harmful material should be removed but not material that may merely offend. Removing such material is a genuine waste of our time and energy. RBI. --ElKevbo (talk) 15:32, 26 September 2009 (UTC)


 * I completely agree, and yours is not an "American viewpoint" at all. Much of the criteria seems to be against the openness that Wikipedia stands for.  When we decide that offensive material, for example, is so horrible that it can't be reviewed by looking at history, we are treating ourselves like children, and giving the admins an old fashioned parental role.  I often look at the history of revisions and reverts, and when material has been removed, I can't make head nor tail of what's going on, or why there was a concern.  A major feature of Wikipedia is that all users can see what's happening in the background, and feel invited to get involved.  Some of these proposed rules go against that principle.


 * To comment on each criteria:
 * copyright violations: Is there a (new) concern that, for legal reasons, CV can't be allowed to exist on history pages? It has always been regarded as legally acceptable to leave this on history pages, as long as it's not on the live article.  Is this proposal in response to a changing legal opinion from the Foundation?  Or are we trying to anticipate such a change before it happens?  If the answer is the latter, it is not appropriate, and reminds me of many rejected good faith attempts to make fair-use rules for images more restrictive, in an attempt to second-guess Foundation requirements.
 * offensive material: I strongly disagree with this item, unless an INDIVIDUAL who is the subject of an attack requests removal of material, which I believe is covered by an existing process.
 * disruptive material: This item covers several things that shouldn't be lumped together. Some of it is a repeat of C2 (despite the clarification).  Classification of material as "of little or no relevance or merit within the project" is no reason whatsoever for revision deletion.  Browser-crashing and malicious HTML: That part I agree with! - if it refers to material actually added to WP, and not on an external page linked from WP.  In the latter case, see my comment about C5.
 * personal info: I agree with this criteria.
 * links: The problem of external links to problematic sites could be resolved by unlinking the text, which I realize is not currently possible in preserved history pages, but IMO is the only right way to address this problem. So we would need to develop a feature to do this.  We've lived without it up to now, so I see no justification to implement an inappropriate interim admin privilege to deal with this.
 * speedy deletion: Don't we already do this? I believe we remove deleted pages to prevent WP from being cluttered with a gazillion rejected articles, and that's fine, but it's not new.
 * arbcom decision: Of course this is necessary, and I would like to see this as C7 of the main list, instead of an "other criteria" section.
 * Extra item! Nothing in this list seems to cover the recent controversies over removal of content that may interfere with attempts to protect the lives of kidnapped journalists. (And by the way, I agree with what was done.)  There will probably be a need for such removals in future, so we should probably own up to it, and put it in the policy.   I'm surprised that these proposed rules, coming at this time, make no mention of this.  One would think that these recent events are what led to this proposal in the first place!
 * --A Knight Who Says Ni (talk) 18:26, 26 September 2009 (UTC)
 * WRT the journalist, there is/was News suppression, which has so far not gotten consensus; so it seems a bit premature adding a related criterion here. --<b style="color:#3773A5;">Cyber</b> cobra (talk) 21:08, 26 September 2009 (UTC)
 * The proposal grew during March - June 2009, from the experience of some oversighters that admin RevDel would be a useful admin tool, rather than due to any incident. WMF norms are that a tool will be released when the community reaches a stable agreement on the matter - ie a policy and consensus.


 * RevDel should be used on mostly the same issues that speedy deletion would be. It's more transparent than deletion, it can also hide things that ordinary deletion can't, but unlike normal deletion it can't hide that something was there.


 * We balance between "everything public" and "here to build an encyclopedia" by allowing most insults and offensive matters to stand, but some matters are so gross that we do endorse removal (Revert, block, ignore, WP:DENY and so on). Running an open encyclopedia does not mean tolerating some kinds of posts that are likely to actively harm the editorial environment and serve no good purpose. That's an old well-established principle. Some editors who are good to have on board can ignore gross insults and attacks, but others cannot, and some comments or links are routinely seen as appropriate to delete using traditional tools too. This codifies that they are appropriate to be deleted by the new tool where traditional delete can't. Does an edit summary "X licks Y's cunt and wanks off over her nude pictures" in some article history (offensive but not defamatory) really need great thought to decide it's best removed? The aim is to benefit the project and its ability to smoothly develop good content and some posts that do nothing but egregiously disrupt or slur, sitting in a user space or article's page history eternally every time some new person looks at that page, raising stress levels.... that's not what we're here for. FT2 (Talk 23:04, 26 September 2009 (UTC)


 * Support rolling this out to all admins. We can delete revisions already, only much more cumbersomely and much less transparently. Thanks for drafting this!  Sandstein   21:41, 26 September 2009 (UTC)


 * Support as a general policy. But the individual reasons as discussed by A Knight require discussion.
 * for the addition he suggested,  deletions on the basis of "do not harm" in cases like the journalists should be done only by OFFICE or perhaps by the oversighters--letting admins dfo this would be chaos and counterproductive,   considering the different views of admins on the matter.    DGG ( talk ) 23:06, 26 September 2009 (UTC)

I would support the general concept, but I think the criteria need some tightening. The criteria should only cover cases where any admin can just delete it unilaterally. Deletions following a consensus from a discussion would presumably be exempt from this (though we might want to specify this in the policy, including the approximate "level of consensus" necessary). I agree with Tznkai about block logs, but I would take it one step further and say any "admin log" should not be deleted by admins. That is, any log that requires admin or higher privileges to use - blocks, protection, user rights, etc - but not user creation, move, upload, etc. Perhaps an exemption for deleting the title of attack pages from the deletion log, but any other case should be handled by ArbCom or the oversight team. Mr.Z-man 23:50, 26 September 2009 (UTC)
 * 1) The wording for this is odd. "by users not associated with the copyright violator" - we're using guilt by association now? If the revision doesn't contain any significant non-copyvio content, it could be deleted. There's no reason to have all sorts of extra fluff here.
 * 2) I would restrict this to only BLP vios. After that it starts to get too subjective. If someone makes a comment where they say nothing but "John Doe is a stupid asshole" we can delete it because it has no value, but if they add some useful commentary with the same insult we can't? That sounds somewhat awkward to enforce.
 * 3) perpetuation of an off-wiki battleground - Are we going to have a bot delete every revision of the talk pages of eastern European subjects? Things that are actually harmful should be removed, but this criterion needs a lot of trimming.
 * 4) It would probably be easier to say that this is a superset of the oversight criteria. Anything that should be oversighted could also be deleted.
 * 5) Links to shock/malware sites should fall under 2 and 3. Do we really have a problem with people posting links to porn sites? And if so, do we really need such a heavy handed approach?
 * 6) This is asking for trouble. CSD covers plain vandalism, author requests, all spam, etc. We need to be specific as to which CSD criteria can apply at the revision level.
 * 7) I don't think we need to specify ArbCom rules here.
 * Question: I thought that administrators already had the option to select only some individual revisions to delete when they click the button to delete a page?  If this is the case, then the only change would be to make the deleted revisions appear as unclickable entries in the history (instead of hiding), correct? Bwrs (talk) 01:47, 4 October 2009 (UTC)


 * Administrators currently have access to an undeletion workaround. They can undelete selective revisions that have already been deleted. So if a page has 200 revisions of which 10 are deleted, they can undelete selectively from those 10. To selectively delete any of the other 190 revisions, an admin would have to delete the whole page, then memorize and undelete the 189 that are okay. Which is very slow, messy and easy to make mistakes, because there isn't a visible indicator which were the existing 10. The admin has to try and get that bit right from memory. So although it works, it's very disruptive to the project, cumbersome, and very easy to delete/undelete the wrong ones. There's also an technical limit; the operation's blocked on any page with 5000+ revisions as it grinds the server down badly (several thousand revisions all to be deleted then a list of all many thousand shown to the user to click next to the ones that should be undeleted.....)


 * As far as the effects and differences in actual use go -- yes, you're right. The main difference to a reader would be that the deleted material will appear as unclickable history and contribs, rather than being silently vanished.


 * Other positive side effects: attribution errors are less likely, admins only have one page to look for deleted revisions, much less disruption and easier to use, much less chance of delete/undelete errors, no restriction on page history size.


 * FT2 (Talk 03:24, 4 October 2009 (UTC)


 * Obvious support and I'm still not sure why this hasn't been given to us yet. –<b style="font-family:verdana; color:black;">xeno</b><sup style="color:black;">talk 19:42, 5 October 2009 (UTC)


 * Support adding the functionality, unsure about the wording. I am particularly concerned about CFR#6. <span style="font-family:monospace, monospace;">decltype (talk) 08:57, 6 October 2009 (UTC)


 * Support. Case is sound.  No to confusing CSD.  --SmokeyJoe (talk) 09:07, 7 October 2009 (UTC)


 * Qualified support. Resounding Yes Please to the principle. Strong No to Criteria For Redaction 1, 2, and 3, except where very problematic material appears in log entries; and for CfR5, delete reference to "links to web pages that disparage or threaten some person or entity and serve no other valid purpose, but not mere spam links.", which is too open-ended and not necessary. Rd232 talk 10:24, 7 October 2009 (UTC)


 * Support but would like a rewording of criteria 1: Blatant copyright violations that can be redacted without removing attribution to non-infringing contributors. What I mean by this: Contributions on wikipedia must be attributed (that's the BY in CC-BY-SA). If a copyvio is introduced in eg revision 10, the article is then improved by others in revisions 11 to 15, and the copyvio removed in revision 16 (retaining the other additions), redacting revisions 10 to 15 becomes a copyvio in itself, introduced by the redacting admin. A null edit crediting the contributors after the redaction would however be acceptable, wherever practical. This must be made clear in the policy, would appreciate help with the language. MLauba (talk) 10:47, 7 October 2009 (UTC)
 * If there is a copyvio, the offending revisions currently vanish from public view by deletion. The deletion log is full of "delete + restore minus copyvio". That's what currently happens, right or wrong. Attribution's a known issue with that too. Introducing a new tool, while at the same time debating a change to an existing established norm that affects all kinds of deletion, is going to be much harder than doing one, then the other.


 * Criterion 1 mirrors current practice for revision copyvios. If we need to change that, we need to change it for both this and selective traditional deletion. That's a big discussion in its own right, focused purely on the narrow question "what's best practice for removing copyvio revisions that have been perpetuated into other users' edits?" It's separable. Let's for the moment use the new tool the same way as the existing one. The same attribution issues apply to both, and are best discussed as a separate question for both. Criterion 1 is an improvement on the status quo, because copyvios in individual revisions currently get deleted using traditional delete, but at least this way if not all subsequent edits are removed, one can see when it's been done and whose revisions were removed, which is a "plus".


 * FT2 (Talk 12:51, 7 October 2009 (UTC)
 * I'd dispute that it mirrors current best practice for revision copyvios. The best practice is that one followed by the admins active at WP:CP, which is to ensure that attribution for non-infringing copyvios is maintained. Perpetuating bad practice in a new policy is not a good idea. If the balance is between "use revision deletion to remove third party infringment even if that infringes on our own contributors" and "remove Criterion 1 because it's too hot a potato for 98% of all admins", you can read my support as changed to exclude the copyright criterion.
 * Doing the right thing for one party while at the same time wronging other parties is not something I'd like to endorse. MLauba (talk) 13:42, 7 October 2009 (UTC)


 * I agree with MLauba's proposed change. If admins are doing this, then they are contributing to a violation of copyright themselves, since reuse of material without attribution is outside the allowance of our license. Copyright problems/Advice for admins (which I've just noticed hasn't been updated to reflect the licensing change!) explains that history should not be deleted where it breaks attribution chain. If there's already confusion on this issue, there's no reason to perpetuate it. --Moonriddengirl (talk) 15:36, 7 October 2009 (UTC)
 * In that case, the issue needs addressing/educating for existing deletion tools too. I don't see anything here that says RevDelete and traditional "delete+selective undelete" would have very different uses or need different policies. How would you feel about a general guideline/policy that covered selective deletion generally (whichever tool it used), made clear when it was/wasn't appropriate to delete selectively from history, and educated on best practice (eg for copyvio and attribution issue avoidance)? From what I'm seeing the issue is much more about selective deletion generally than about the specific RevDelete tool, and if it's already an issue with the existing tool, then address it for both. I might do something on this (see below). Sounds ok? FT2 (Talk 06:20, 8 October 2009 (UTC)
 * You're right, the issue is common to three scenarios: recreation of previously deleted material off one of our mirrors (eg wikibin) by another contrib that is kept, deletion + selective restore, and the future revdel. And I agree with you, addressing this is a separate matter. My concern here is simply to find a language in this proposed guideline that encourages the best practice rather than the mistaken one. That being said, I take your point, I'll look into tightening the other deletion guideline pages at the same time. MLauba (talk) 07:43, 9 October 2009 (UTC)
 * Rethinking this after going through deletion guidelines. In reality, the deletion guidelines AND the specific guideline on selective deletion already make it abundantly clear that that procedure shouldn't be used at all if valid contribs are included in the deletion. Selective deletion as is done currently is a hack that is difficult to revert, whereas if I understand it properly, version redaction can be undone simply by unticking checkboxes. Once the feature is live, I believe selective deletion should be prohibited, the new feature OR oversight remaining the only two valid options. Which reinforces my point: the language in the proposed guideline has to be clarified in order to make sure that the admins using the feature do not infringe on the right of the legit contributors to the article. Makes sense? MLauba (talk) 08:00, 9 October 2009 (UTC)
 * I'm all for clarifying common points of confusion. :) Selective deletion already indicates that attribution needs to be retained, although I suppose it needs to be made stronger (and, like the Copyvio advice for admins, I see that it, too, had slipped out of date before MLauaba caught it up) and better publicized. If this is widely misunderstood, then the inclusion of something like the six words MLauba proposes here ("without removing attribution to non-infringing contributors") would probably also be helpful, and I think it's an important enough issue that it isn't a major concern under bloat. Of course, there's also a lot of conversation further down, so I don't know if the direction has changed. Going to catch up. :) --Moonriddengirl (talk) 10:49, 9 October 2009 (UTC)
 * I've started to collate some data for this discussion. I do agree, once RevDelete is live there will be only a tiny number of issues where traditional selective deletion (delete+partial restore) might still be needed or desirable. That would need discussing though. FT2 (Talk 15:02, 9 October 2009 (UTC)


 * Strong support - Primarily for it's usefulness in copyvio work, where I spend time, and probably applications in BLP, where I do not. The issue above must be addressed but this should be worked on at the copyvio pages, not here.  Deletions of copyvios are often done incorrectly  by well intentioned admins (How many non-copyvio admins have deleted the talk pages without looking for a Talk:__/Temp? or had no idea what to do with it if they found it?) but we don't try to explain the entire copyvio policy at WP:DEL.  Best practices should be developed where the experts are after the basic policy is outlined here.  People will always screw it up but at least we'll be able to fix it easily now.--Doug.(talk • contribs) 19:44, 7 October 2009 (UTC)
 * Strong support per David Gerard. —Preceding unsigned comment added by Keegan (talk • contribs) 20:03, 7 October 2009 (UTC)
 * Support. This would be a great thing for dealing with copyvios and attacks that don't require locating an elusive oversighter and will make the administrative job a lot simpler. The transparency's a bonus too. <b style="background:blue; color:white; font-family:Comic Sans MS;">Valley</b>2 city ‽ 06:44, 9 October 2009 (UTC)

Add to CSD?

 * On reflection, there's a second approach to administrator use of RevDelete and it might actually be a better one.


 * At present we're discussing a separate policy. But we could simply recognize RevDelete as similar in form to speedy deletion (same spirit, different tool) in that it allows single revision deletion (but without full delete+reinstate) and it can do the same job in fields that usual delete + reinstate can't. If so, we could simply add RevDelete into WP:CSD as a second tool one can use for the same kinds of material, and check if any of the listed items should be added to CSD. You might end up with one policy not two, and the following addition to CSD:
 * {| style="border:black solid 1px;font-size:90%" width="90%"


 * '''Addition to WP:CSD:

== Revision deletion ==

The speedy deletion criteria are usually applied to entire articles, but they can also be used to delete specific revisions from articles, either by deletion and reinstatement, or by using the RevisionDelete tool. The same criteria apply to both.

As RevisionDelete is less disruptive and more transparent than full deletion, it will usually be the preferred tool when specific revisions must be removed or redacted. The following additional deletion criteria usually apply to individual revisions of pages or log entries:


 * CSD#RD1 Copyright violations that do not affect the entire page history...
 * CSD#RD2 Grossly offensive material...
 * CSD#RD3 Grossly disruptive material...
 * CSD#RD4 Link to a web page of a purely malicious or grossly inappropriate nature...


 * }


 * This approach might be better for keeping down policy bloat and ensuring our "on the spot deletion" policies and tool use stay consistent and simple. FT2 (Talk 07:21, 27 September 2009 (UTC)
 * Interesting approach, we have way too many policies now, it's a rather confusing thicket. ++Lar: t/c 18:27, 27 September 2009 (UTC)


 * Support I really think this is a good idea in general.  Selective removal of copyright violations from history is a big boon.  However, I do not like the idea of a link to a shock site or normal offensive vandalism being hidden through this tool.  Transparency is one of the core values, and it shouldn't be given up for aesthetic reasons.  Gigs (talk) 19:30, 27 September 2009 (UTC)
 * Strong support. We can already do it, just awkwardly, so this isn't giving admins any new powers (except on articles that are too big to delete, I suppose). Stifle (talk) 21:14, 28 September 2009 (UTC)
 * Unfortunately, not quite true. RevDelete reaches into various logs.--Tznkai (talk) 22:45, 28 September 2009 (UTC)
 * Admins can already delete specific revisions using the existing tool. This tool extends that existing ability into logs, and more openly. Not really much of a new thing, rather, a limiting problem with the existing tool (some actions that admins would take over page revision contents, they can't, and the only reason they can't is if it's in a log not a page) and a transparency improvement. FT2 (Talk 09:45, 29 September 2009 (UTC)
 * I don't like that it extends to logs. Certainly not when we're proposing rolling it out to what, 1000 additional users? We don't have a policy, or a common practice, or even principles to guide admins on how to use that aspect of the tool. I don't want to throw out the baby with the bathwater here, but deletion of block logs alone is a recipe for chaos.--Tznkai (talk) 14:35, 29 September 2009 (UTC)
 * Doing stuff to block logs will usually be a direct route to RFAR for gross abuse. It works on wheel warring and self-unblocking, all 1700 admins know there are a few first-time desysoppable actions. But also, block log entries being posted by admins only (like several other logs) should almost never need redaction anyway (suppression yes but never unilateral deletion). So one rule solves all of these issues: Administrators must never redact any log entry that was part of an admin tool action without discussion or strong prior community consensus, or Arbcom consent. Deleting admin tool log entries may lead to loss of adminship even for a first incident. FT2 (Talk 18:26, 29 September 2009 (UTC)
 * Assuming the actions are still fully visible to admins, I think that "never" is a bit of an overstatement. If someone creates an account or page which is highly insulting, it may be appropriate to hide the deletion/block title first and consult with other admins afterwards. עוד מישהו Od Mishehu 09:47, 4 October 2009 (UTC)


 * String support - as an admin who has had to hide moves of semi-protected pages (frequently user pages) to insulting names, when there were previously deleted revisions in the histories, this method is both more transparent and less mistake prone. עוד מישהו Od Mishehu 09:21, 29 September 2009 (UTC)


 * Conditional support I have no problem with adding it to CSD but we need to be careful about it. Take FT2's examples above. The first criterion proposed is G12, so we don't need it. The other three are vague and likely covered by G3, G10 and G11. Since revision deletion does not remove the whole article, some admins are likely to use it much more often to clean the page history, thus removing material that should not be deleted or breaking the attribution chain. Thus we need to be very careful when creating new criteria to avoid revision deletion whenever it's not absolutely crucial (for the same reason we currently blank sensitive AFDs or RFAs or user talk pages). Maybe we should rather try to use the existing criteria for speedy deletion and just expand them to allow revision deletion (e.g. add something like "This criterion can also be used to remove certain revisions from the page history using Revision Deletion") instead of creating new ones. I cannot think of any situation where a revision could be deleted that could not be speedy deleted as a whole article. Regards  So Why  15:05, 29 September 2009 (UTC)
 * I mentioned several situations above. We shouldn't be using this for ordinary vandalism, ordinary spam, author requests, test edits, no assertion of notability, etc. Mr.Z-man 16:23, 29 September 2009 (UTC)
 * Strong support - as have others, I have had to use the cumbersome delete-all-revisions-restore-most process. This would be a great improvement, and reduce the need for oversighters to intervene. The fact that revisions still can be hidden (by an oversighter) from those without the oversight flag as needed should allay other concerns. This is simply a better way of doing what we all ready do. Xymmax So let it be written   So let it be done  15:15, 29 September 2009 (UTC)
 * Support iff there are further restrictions added on deleting logs, similar to what I suggested above. We've been able to live without anyone but the sysadmins being able to delete logs for years until oversights got the ability, I believe admins deleting admin logs is likely to cause more drama than it will be worth. Mr.Z-man 16:23, 29 September 2009 (UTC)
 * Agree on this qualifier. Please don't open that can of worms. Xymmax So let it be written   So let it be done  17:23, 29 September 2009 (UTC)
 * I proposed a stronger and very easily drawn line above for this concern (which I agree with): Doing stuff to block logs will usually be a direct route to RFAR for gross abuse. It works on wheel warring and self-unblocking, all 1700 admins know there are a few first-time desysoppable actions. But also, block log entries being posted by admins only (like several other logs) should almost never need redaction anyway (suppression yes but never unilateral deletion). So one rule solves all of these issues: Administrators must never redact any log entry that was part of an admin tool action without discussion or strong prior community consensus, or Arbcom consent. Deleting admin tool log entries may lead to loss of adminship even for a first incident.
 * FT2 (Talk 18:31, 29 September 2009 (UTC)
 * If it were used to delete logs, wouldn't it still show up as a crossed out revision? Gigs (talk) 22:54, 29 September 2009 (UTC)
 * Yes - RevisionDelete cannot be used at all by any admin to hide the existence of the log entry/revision. It doesn't have that capability. If RevisionDelete was enabled for admins, then any admin would always be able to see any log entry revdeleted by another admin, and unrevdelete it if improper, and any user or IP editor would always be able to see there was a log entry there (and could ask an admin if it was salient or proper), even if they couldn't see some of the details. It doesn't "delete" in the sense that existing admin and oversight tools do, by removing log entries completely. That capability doesn't exist.


 * What RevDelete does (at admin level), is if you're an admin you'd see the normal visible data and links (+ change of font color), if you aren't an admin you get a crossed out placeholder for that field, not "vanishing" as at present. So any editor can at least tell there is something there, eg for attribution purposes and even admins benefit (they don't have to separately check deleted history as well, to see if there's missing edits), which is a problem with the current tool. Everyone can see there is something there. Same whether the item is a log or a page revision. FT2 (Talk 02:04, 30 September 2009 (UTC)
 * Well, there are a couple valid reasons for deleting parts of admin logs, deleting the title of an attack page from the deletion log for example. Though such instances probably come up so rarely that they could be handled by oversighters. Though that raises another issue; if an oversighter deletes a revision/log entry at admin-level (not suppressed), are they bound by these rules the same as all admins, or are they afforded some extra latitude for dealing with special cases? Mr.Z-man 03:23, 30 September 2009 (UTC)
 * RevDelete at admin level can't presently be viewed by non-oversighters so right now its effect is temporarily similar to oversight (current status at bugzilla). If proper community review were possible, there'd be norms for suppression mode RevDelete based on oversight and norms for admin mode RevDelete based on deletion and communal consensus. An oversighter who admin-level deleted an item would be doing it like any admin. And yes, there's quite a fair amount of gross attack page titles and other serious (but not quite oversightable) stuff - see WP:SIGHT policy #4 and associated footnote. FT2 (Talk 04:42, 30 September 2009 (UTC)


 * Support per Sandstein and Mr.Z-man. – Juliancolton  &#124; Talk 18:10, 29 September 2009 (UTC)
 * Question How much extra work would it take for the devs to spin off log deletion from page revision deletion?--Tznkai (talk) 18:13, 29 September 2009 (UTC)


 * (Note -- anyone interested in trialling admin level RevisionDelete for themselves, it's included in the admin tools on the flagged revisions test wiki for any reasonably plausible user. The purpose of that wiki is to test Flagged Revisions, but it's unlikely anyone will object very much to seasoned users looking at RevisionDelete as well. - FT2)


 * Strong Support for the revision delete feature. I was trying it out earlier today on the test wiki, and I have to say that the ease with which one can do the tasks - and the flexibility - is impressive.Equally impressive is the fact that my tests were clearly logged, and thus available for review if need be. There were concerns expressed about adding 1700 new bodies with this "power"; another way of looking at it is that there would be 1699 additional bodies to review an admin's actions. --Ckatz <sup style="color:green;">chat <sub style="color:red;">spy  08:01, 30 September 2009 (UTC)


 * Strong Support. Among copyvio cleaners, article history deletions are already a common practice, as the only real alternative to G12 deletion to new articles in violation that could otherwise be salvaged. Tools that simplify the current G6 - delete all then selectively restore are definitely a good thing in my book. Whether in practice we create a new RD CSD group or expand G12 doesn't matter too much in principle. One drawback to expanding G12 is purely practical: a today's G12 is handled as a high priority item to delete, along with G10. A revision deletion is less urgent, as the article on display is supposed not to be in vio anymore. In terms of impact, I'm estimating that about 20% of all articles checked on WP:SCV lead to history deletions, and a sound number of items cleared at WP:CP also end up with a selective deletion. MLauba (talk) 09:39, 1 October 2009 (UTC)
 * From my experiments on the test wiki, it doesn't seem like the interface allows one to delete more than one revision at a time, which would actually make the old method more efficient in many cases. <span style="font-family:monospace, monospace;">decltype (talk) 09:12, 6 October 2009 (UTC)
 * There is such an option on page history. That functionality hasn't yet been added to user contributions and log pages though. See Bugzilla request. FT2 (Talk 09:30, 6 October 2009 (UTC)
 * There is? How do you go about doing it? <span style="font-family:monospace, monospace;">decltype (talk) 09:42, 6 October 2009 (UTC) Adding multiple revids to the ids parameter seemed to work, but the checkboxes didn't.
 * You should see checkboxes and a button to use them. See screenshot I've added (on the right). FT2 (Talk 14:18, 6 October 2009 (UTC)
 * Thanks for the pic. It works in Firefox but not IE7 ("The action specified by the URL is invalid"). <span style="font-family:monospace, monospace;">decltype (talk) 05:57, 7 October 2009 (UTC)
 * Check with someone else who uses IE7 on the same platform, if it's just you, or them as well. if it's them as well, consider reporting it at Bugzilla. FT2 (Talk 13:09, 7 October 2009 (UTC)


 * Support. I'm sure we're grown up enough to be trusted with this feature to make our work much easier. -- &oelig; &trade; 20:52, 3 October 2009 (UTC)
 * Strong support. Per David Gerard. --Closedmouth (talk) 07:06, 5 October 2009 (UTC)
 * Not sure where such a comment would fit into the threading on this page, but I strongly oppose making this just a section of CSD. It has too many caveats and too much importance to be 'relgated' thus, at this stage.  Suggestions of "we have too many policies" fundamentally misunderstand what a Wikipedia "policy" is: a documentation of a community norm or best practice.  How and where such a thing is documented is an entirely trivial detail.  Happy ‑ melon  19:14, 6 October 2009 (UTC)
 * I agree. Whilst minimising policy complexity is a worthwhile objective, that isn't necessarily achieved by putting things together on a single page, if they're too dissimilar. I think RevDel and CSD are (or should be) too dissimilar. Take out CFR1 (which I object to), and there really isn't that much overlap with CSD. Rd232 talk 10:32, 7 October 2009 (UTC)
 * I can't think of a single criterion where it would be valid to use selective deletion (delete all and restore most) but more harmful to use RevDelete (redact just the offending fields from the problem revisions, leaving the rest publicly visible). I also can't think of a case where RevDelete would be valid that we don't at present find widespread use of selective deletion. Both tools/methods serve the same purpose, removing selected information from history without formal discussion (xFD/Prod), by an admin's on-the-spot decision. Common criteria/guidance for when an admin may remove material on the spot from public history - whatever tool they do it with - would be useful. FT2 (Talk 13:07, 7 October 2009 (UTC)
 * I agree fully with your first two sentences, but that's where the similarity ends. CSD does not consider selective deletion.  CSD is just a set of pressure relief valves to stop XfD getting overloaded; they're supposed to be a set of largely-uncontroversial sitautions for page deletion where a somewhat 'cavalier' attitude can be tolerated, if not encouraged.  Do we want to encourage that same trigger-happy attitude towards RevDelete?  That is inevitable if the policy is merged there.  There is no need for that to happen, WP:NOTPAPER etc: it makes no sense to cram these policies together when they are at best awkward bedfellows, and where the last thing we want is for the ethos of CSD to rub off on CRD.   Happy ‑ melon  14:01, 7 October 2009 (UTC)
 * I wasn't aware that CSD was used in a "cavalier" sense. A user who repeatedly deleted pages or revisions in a problematic manner would probably be raised at ANI, like most questionable tool uses are. I've been looking at our use of selective deletion with the current tool, and what I'm seeing so far is widespread unspoken agreement on what it's used for - the same basic set of uses across the board with not a lot of exceptions. It might help to put up some stats. I'll compile them in the next few days if people would like. (I'm away a bit today/tomorrow). FT2 (Talk 14:24, 7 October 2009 (UTC)
 * Do you read WT:CSD? Admins using CAT:CSD as a whack-a-mole opportunity to delete pages as fast as possible are so common that people have given up taking them to ANI.  Take any brief look through the archives.  Admins being too trigger-happy at CSD is an endemic problem, and the last thing we want to do is spread that attitude to CRD.  Because I agree, there is currently widespread agreement over what comprises acceptable use of this feature.  I see no reason to risk losing that.  Happy ‑ melon  07:34, 8 October 2009 (UTC)
 * Noting that Happy-melon and I caught up by email on the above comment, I think we're pretty much in agreement now... if not, I hope HM will let me know. FT2 (Talk 15:02, 9 October 2009 (UTC)


 * Strong Support we already have the ability to do this, just through an extremely lengthy process. This is a much easier way, and I don't know why we never had this before. I also support adding this to CSD as it adds even further functionality to both the CSD process and the RevDel process. --<small style="color:#999;white-space:nowrap"><big style="color:#ffa439">Coffee //  have a cup  //  ark  // 09:39, 10 October 2009 (UTC)

Conclusion
So it looks like this has enough support, doesn't it? What now? When will this happen? Rd232 talk 21:08, 16 October 2009 (UTC)
 * I'll tweak what needs to be tweaked, reword what needs to be reworded, then file a request on bugzilla to have the feature enabled. ~ fl 07:03, 17 October 2009 (UTC)
 * Policy reworded and clarified, Mediawiki namespace messages edited accordingly, bug 21165 filed to have the feature enabled. ~ fl 08:41, 17 October 2009 (UTC)

I'd be relieved if were resolved before or soon after the implementation, in the mean time we can use Database reports/Atypical deletion log actions. Cenarium (talk) 22:16, 18 October 2009 (UTC)


 * Stupid question, is this live now or are the entries we're seeing in the deletion log with the criteria done by oversighters? MLauba (talk) 07:37, 20 October 2009 (UTC)
 * That's the oversighters. ~ fl 09:07, 20 October 2009 (UTC)

Some questions
First, I'd like to say this is a pretty cool feature and I'm glad its being enabled. Unfortunately, the current instructions aren't clear enough (for me anyway) from a policy stand point.


 * 1) Let's start with the most unambiguous example: Someone adds copyright violating text to an article with no other text.  I assume the proper thing to do here is revert & then delete the offending revision.
 * 2) Now let's assume someone adds some copyvio text, but also some original text.  I assume the correct thing to do here is remove the violation and nothing else.
 * 3) As to attacks, what constitutes "grossly offensive?" A good chunk of vandalism takes the form "so and so is a [insert favorite insult here]."  If it is just one sentence like that I view it as ordinary vandalism, although it is technically an attack.  Does the attack have to go beyond this (i.e. be a clear attack and not just especially bad vandalism) to be deleted?
 * 4) Point 5 ("Valid deletion under Deletion Policy, executed using RevisionDelete") is a little vague not in its meaning, but in what situations the new tool would actually be helpful.
 * 5) I don't really see anything indicating when log entries should be deleted. I can make a good guess - the user name/edit summary itself is an attack - but maybe this could be more clear.  I assume the username shouldn't be redacted if the edit is live (and non-meaningless) as that would violate attribution.  What about the edit summary - should this be redacted just because it is an attack or does it have to be a "gross violation"?  I imagine if we start redacting attacks, it will become harder to determine a pattern of abuse among non-admins.
 * 6) Partially duplicate to previous comment... There was some (wise) discussion above that admin logs should never be tampered with. I see this hasn't made it into the document, which has very little about logs in general.  As such, some clarification on use of the tool for log entries is likely warranted.

Thanks, ThaddeusB (talk) 02:48, 18 October 2009 (UTC)


 * Vis 6, at least, if you look at Special:RevisionDelete (which I know you can't yet, but trust me, it's there) you will be left in no doubt as to the Do Not Want nature of tampering with admin logs. I wouldn't be at all averse to seeing it in the policy, though.  Happy ‑ melon  08:20, 18 October 2009 (UTC)
 * It is in the policy actually, though not very prominently, in the "Log redaction and improper use" section. Mr.Z-man 22:23, 18 October 2009 (UTC)
 * In regards to 1 and 2, User:Flatscan has started drafting an Attribution guideline that also covers these points explicitly, but to keep this short and to the point, the safest route is what you describe, but you can also satisfy the attribution requirement by creating a dummy edit crediting the user eg. "Revisions removed for copyright reasons. Text retains content contributed by User:Example" still satisfies the attribution requirements. MLauba (talk) 22:05, 18 October 2009 (UTC)

Dropdown
There is a page providing options for a dropdown list of reasons used in the revdelete interface in the same way we have one for regular deletion and block reasons, found here. One of the options at the moment is "RD5: Valid deletion under Deletion Policy". This makes little sense to me as it provides almost no information. Can we remove it? — Jake   Wartenberg  21:58, 19 October 2009 (UTC)
 * Not right now. That is the exact wording of the policy, as I noted in my edit summary when I placed it back on the page when NuclearWarfare removed it. If you want to remove the option, you'd need to get consensus to remove that entry from the policy. ~ fl 23:52, 19 October 2009 (UTC)
 * I don't see why. There isn't an entry for RD6.  We really don't want people's deletion reasons to consist of this; it is about as bad as deleting a page "per the deletion policy".  —  Jake   Wartenberg  00:46, 20 October 2009 (UTC)
 * There is no entry for RD6 solely because that criteria is only available to the smallest group of people, not the general admin population in most circumstances. If you can think of better wording for the policy that captures the essence of what RD6 is trying to say, you are most welcome to change it. The only thing removing the option really achieves is inconveniencing people now have to type out that text every time they wish to use this option. ~ fl 08:51, 20 October 2009 (UTC)
 * Reworded slightly to "RD5: Other valid deletion under Deletion Policy (detail required)". May help. FT2 (Talk 15:46, 20 October 2009 (UTC)

Intro and misuse wording
I have added a strong wording to the introduction, covering the "misuse" issue.

We will have a thousand or so admins reading this page shortly, and several thousand non-admins or future RFA nominees reading it. A part of the consensus above was a very strong feeling that the tool should not be abused. Highlighting this in the intro, and that abuse may lead to desysopping (which is true of any tool abuse), seems desirable. It may help to establish the point firmly as RevDel gets activated.

FT2 (Talk 07:44, 20 October 2009 (UTC)
 * I love the wording, but thought that a paragraph of bold disrupted the page's "flow", so I moved it to a "very serous warning" box instead. ~ fl 09:04, 20 October 2009 (UTC)

(Hopefully) non-contentious criterion RD#7
It's likely there may be some pages where it'll be beneficial to undelete old revisions and redact the problem information instead, either for history's sake, for ease of admin reference, to improve attribution, or simply for transparency. The current criteria don't include a "housekeeping" category. Suggested addition:


 * 7. Non-contentious housekeeping including correction of clear and obvious unintended mistakes in previous redactions, changes to redaction based upon communal discussion and clear consensus, adding information to the delete logs, and converting traditional selective deleted edits to RevisionDelete. (The action must not be likely to be contentious or controversial, consult if needed)


 * FT2 (Talk 11:23, 22 October 2009 (UTC)
 * Seems OK to me. ~ fl 12:22, 22 October 2009 (UTC)
 * It needs to be more clear what "correction of clear and obvious mistakes" refers to. As worded someone might think it means removing editorial mistakes from the history, but I'm not sure if that is your intention. --ThaddeusB (talk) 23:55, 22 October 2009 (UTC)
 * Clarified, thanks. FT2 (Talk 23:59, 23 October 2009 (UTC)
 * Much better. --ThaddeusB (talk) 00:00, 25 October 2009 (UTC)

And possibly one other last one:
 * Non-public material not meeting Suppression criteria BLP violations and non-public material that should be deleted but do not meet the criteria for suppression.

FT2 (Talk 11:39, 25 October 2009 (UTC)

Review of deleted vs suppressed revisions
I remember some findings indicating that oversighters occasionally used revision deletion without restricting visibility to oversighters for oversightable items, such as personal info. Before admins get the rights, it should probably be made sure that those revisions have their visibility restricted to oversighters - and that oversighters well know about this. Has this been dealt with, or could AUSC take care of this ? Cenarium (talk) 14:56, 24 October 2009 (UTC)
 * It should have been notified and dealt with - oversighters have been aware and should have been checking it for a long time now. "Yet another reminder" probably won't go amiss though, although I'd be surprised if they weren't all up to date by now. I'll write one. FT2 (Talk 15:19, 24 October 2009 (UTC)
 * I have checked Database reports/Atypical deletion log actions and found what I think are examples as recent as a few days ago. It could be an immediate source of concern for a reason I forwarded by email (to FT2 and AUSC), I don't know if it's a known problem and prefer to be careful. Cenarium (talk) 15:49, 24 October 2009 (UTC)
 * I have flagged the issue twice to other oversighters, everyone should be responsible for their own fixes. I have a list of unfixed issues that need to be checked, many of which will need to be re-suppressed.  In the mean time, admins who see things that should have been fully suppressed can notify the oversight mailing list. Thatcher 17:12, 24 October 2009 (UTC)

In any case, admins have been able to see admin-level revdeleted material for about a week. — Jake   Wartenberg  01:49, 25 October 2009 (UTC)

A use for "delete some edits"
Steps 1 and 2 could be replaced by "delete some edits" of P, if "revision deletion" can be undone by anyone who had done it. Anthony Appleyard (talk) 23:02, 25 October 2009 (UTC)
 * A use that I would have for "delete some edits" is in histmerging. Say that someone cut-and-paste moved page P to page Q, replacing P by a redirect to Q. People afterwards edit P variously, creating new edits which are parallel to the history of Q; sometimes those late edits are only a bunch of redirects, sometimes they are a useful article. So this must be done, as it is now:
 * Delete P.
 * Undelete P, only the edits before the cut-and-paste.
 * Delete Q.
 * Move P to Q :: this usually automatically does the previous step, but not when move-an-article also moves its talk page.
 * Undelete Q.
 * Undelete P :: this restores the late edits of P.
 * Revert any junk last edit left in Q's history by the move from P to Q. In the process, remove any tag used to call for the histmerge.
 * No, revision deletion doesn't work that way. Complex history merges like this are pretty much the only remaining use for selective deletion like you note. Mr.Z-man 23:56, 25 October 2009 (UTC)
 * But when history-merging I often have to temporarily "delete some edits", and I would prefer to delete some edits directly, than to have to delete all then undelete some every time. Anthony Appleyard (talk) 07:09, 27 October 2009 (UTC)
 * The way I understand it, when you use revdel the record of the edits aren't actually removed from the article's history. Thus, it would be useless for history merge purposes because if the edits aren't removed from the history they will won't be left behind when you move the page. --ThaddeusB (talk) 14:31, 27 October 2009 (UTC)
 * Correct. (Which in turn suggests this idea.) FT2 (Talk 17:13, 27 October 2009 (UTC)