Wikipedia talk:Rollback

RfC: Restricting high-speed reversion software

Should the following text be inserted in the Additional tools section? – Teratix ₵ 09:41, 26 January 2021 (UTC)

Certain tools that permit extremely fast reversion, such as Huggle, must only be used by rollbackers.

Background

Certain reversion software, such as Huggle, prevents editors without the rollback permission from operating the program. An editor, who had thrice been denied rollback permissions (1 2 3) by multiple administrators, opted to ignore these decisions by modifying Huggle in order to bypass its rollback requirement. There does not appear to be a policy or guideline that prohibits this.

A discussion about inserting the wording petered out without a definite conclusion. See also VPP and BOTPOL discussions (both inactive).

Survey

• Support as proposer. This insertion merely installs current practice into a guideline. Experienced Huggle users have strongly affirmed that the software should not be used by inexperienced editors, due to the higher risk of problematic reverts arising from its speed.

To me, a lack of consensus on this matter would be the worst possible outcome – if Huggle and similar tools are to be available to all users regardless of experience, they should be available to all users – not merely those with the technical savvy to modify Huggle and knowledge of this loophole (which is the current situation). Hence this RfC. – Teratix ₵ 09:41, 26 January 2021 (UTC)

some of the Huggle tools are easily usable by anyone; some require skill and experience. DGG ( talk ) 10:35, 26 January 2021 (UTC)

• Support. I find this an interesting topic matter. Recently a user modified the counter-vandalism tool RedWarn to bypass the requirements of extended-confirmed, as determined by consensus. However tools such as Huggle are much more powerful than diff tools such as Twinkle and RedWarn, as someone could just hold down the “Q” key and wreak mass havoc, even if they were blocked within a minute or two. Rollback is not only a useright for first reversions, it also shows that an administrator has asessed a user's contributions, and deemed it sutiable for them to have the rollback permission to use high-revert tools such as Huggle. Rarely are there cases of “rouge Huggle users”, as an sysop has given them a check, approving them for the userright. There is a reason tools like Huggle are locked to rollbackers, it means the chances of the counter-vandal becoming the vandal are much lower. There is however a broader question as well, should modifying tools to bypass restrictions be allowed? Maybe a decision for another RfC. — Yours, Berrely • ^Talk∕_Contribs 11:14, 26 January 2021 (UTC)

  It's essentially the same question: if reverting rapidly with a tool is prohibited unless you have rollback privileges, it doesn't matter if you wrote the tool from scratch, modified an existing tool, or used an existing tool that relies on the honour system. isaacl (talk) 19:21, 26 January 2021 (UTC)

  @Isaacl: If so, the text should be changed to something like

  Certain tools that permit extremely fast reversion, such as Huggle, must only be used by rollbackers. Modifying a tool to bypass restrictions must not be done against community consensus.

  ―sportzpikachu _{my talkcontribs} 08:14, 28 January 2021 (UTC)

• I'm not against this change, but as I mentioned before, I'd prefer to clarify that the restriction applies to the action (such as fast reversions), regardless of how it is carried out, whether it is through tools or someone staging a lot of edits and then executing them manually. isaacl (talk) 19:16, 26 January 2021 (UTC)
• I agree that modifying tools to bypass restrictions should not be allowed and that the restriction should apply to the action (such as fast reversions), regardless of how it is carried out. If we want to avoid damage/vandalism, it shouldn't matter how the damage/vandalism is done. ---Avatar317^(talk) 20:43, 27 January 2021 (UTC)
• Support as a first step. However, I believe that adjusting the API throttle limits would be a more efficient method. It appears to me that all nonadmin users are currently limited to a rate limit of 100 rollbacks and 90 edits in a 60 second time frame. Couldn't it be adjusted downward to some appropriate number for nonrollbackers like 10 or 15? -- Dolotta (talk) 23:36, 27 January 2021 (UTC)
• Support, per Berrely, and also my opinions above ―sportzpikachu _{my talkcontribs} 04:29, 28 January 2021 (UTC)
• Support per the reasons I stated in the #Explicitly restricting high-speed reversion software to rollbackers section above. Mz7 (talk) 05:47, 7 February 2021 (UTC)
• oppose I don't think it's a good idea to generalize few incidents. I can see many good contributions from users who use fast reverting tool. One such tool (maybe? I am not even sure it's a fast reverting tool) is SWViewer which requires you to have rollback on a wiki but you can also patrol other wiki if you have a certain amount of global activity and there are many trusted users from other wikis who use this feature and help patrolling English Wikipedia too. So by this we are basically saying no to them. ‐‐1997kB (talk) 04:10, 21 February 2021 (UTC)
• Support per Teratix. As suggested by Dolotta, API throttle limits are also worth exploring. 207.161.86.162 (talk) 08:18, 25 February 2021 (UTC)

Speed of rollback vs Twinkle

Is rollback proper faster at reverting a string of edits than Twinkle's rollback? Wikinights (talk) 02:49, 30 July 2021 (UTC)

You might ask this at WP:VPT. My guess is that in theory rollback would have a performance advantage since it's built-in to MediaWiki, but it is unlikely that the difference would be sufficient to show up in measurements. Johnuniq (talk) 05:05, 30 July 2021 (UTC)

Wikinights, Johnuniq, rollback is particularly faster for reverting multiple edits. You can open a dozen rollback links in tabs and without further interactions you'd perform a dozen reverts. For individual reverts the difference isn't as much. Wikinights, if you want to try it out, register on https://commons.wikimedia.beta.wmflabs.org/wiki/ and tell me your username, I can give you the right there. — Alexis Jazz (talk or ping me) 06:52, 1 August 2021 (UTC)

Why is this a granted user right?

Simple question, why is the rollback user right even a thing when anyone can just install Twinkle and use its rollback features? What am I missing? Throast (talk | contribs) 15:30, 21 November 2021 (UTC)

@Throast: Rollback is much faster than Twinkle's pseudo-rollback features. The "rollback" API call is just one request instead of undo, which takes two (one to get revision information, another to undo). Likewise, some of the more powerful reversion features (e.g. mass rollback, Huggle's "one key to revert and warn" system) are most commonly hidden behind rollback, so as to prevent abuse. When getting rollback, you usually get a message saying that it's not as momentous as installing Twinkle, which is true for the most part. However, rollback can be a useful tool (and a key to unlock even more tools) for dedicated counter-vandalism editors who regularly deal with massive swaths of vandalism. Chlod (say hi!) 16:08, 21 November 2021 (UTC)

An addition

I'm noting that I've made an addition for something I see far too often. I thought it would go without saying, but obviously not: If you see a user with a username like "John Smith is a rapist", don't use a default rollback summary which adds the text, "John Smith is a rapist". Write something else instead. Duh. Thanks. -- zzuuzz ^(talk) 16:05, 11 December 2021 (UTC)

@Zzuuzz: This usually occurs to me about 1/2 second after I click the rollback button. Suffusion of Yellow (talk) 22:06, 15 December 2021 (UTC)

Indeed. And then I'll delete your revision summary, describing it as "Grossly insulting, degrading, or offensive material", which will make it look like you added something really grossly offensive - which of course you did :/ Hey sometimes it takes many years for policies to seep into the collective consciousness. Maybe in a few years a few more people will think twice. One day some of you rollbackers may become an admin and start deleting stuff. You'll soon realise how annoying it can be to have to clean up after at least twice the number of vandals. -- zzuuzz ^(talk) 22:29, 15 December 2021 (UTC)

Rewriting the guideline

As recently discussed at WP:XRV, the rollback guideline looks out of date and needs a rewrite. In particular, "It is used to undo problematic edits such as vandalism." doesn't appear to be the main use anymore; indeed, the primary use is to be able to revert quickly, and there are now examples of people using rollback in good faith on edits which are also good faith, none of which causes an actual problem with the encyclopedia except for anyone looking at the text on this guideline page and walking away with the wrong impression. So the guideline has to change, because it doesn't reflect reality.

I'm just wondering if anyone has got an ideas of what specifically needs changing. Or shall I just be bold and go forth and copyedit myself? Ritchie333 ^{(talk) (cont)} 15:30, 16 December 2021 (UTC)

I've been using rollback more frequently lately (examples: diff1 + diff2) due to the unending stream of nonsense but I do not think any rewrite should suggest that rollback is permitted on reasonable edits. Once upon a time I might have crafted a message for the IP and the user in those examples but such cases are too frequent now. To me, "rollback" means there is no way this edit is reasonable and onlookers who might generally trust my judgment needn't bother examining the diff. Or (as explained at WP:ROLLBACKUSE #5), rollback is ok when reverting lots of edits provided an explanation is made elsewhere. The discussion at WP:XRV is too long for me to want to fully understand but a tool or rollback should not be used to revert reasonable edits without explanation. I think examples of the issue in that discussion are diff3 and diff4 which both include an explanation in the edit summary. That makes them ok as far as I am concerned. Johnuniq (talk) 22:41, 16 December 2021 (UTC)

Revdelled usernames

@Zzuuzz: Regarding [1], I just tested it out here, and I think the behavior for revdel and oversight is the same. I made an edit to my sandbox with a test account, revdelled the username, then used rollback on the edit, and the auto-generated edit summary was Reverted edits by a hidden user to last version by Mz7. Mz7 (talk) 01:03, 17 December 2021 (UTC)

Hi Mz7. I accept your test, however, I could point to numerous examples where this did not happen. From recently:[2][3][4][5] I would need to do more testing to see what's going on here, but I think I'd have a valid point if I said we can't rely on this behaviour. -- zzuuzz ^(talk) 05:30, 17 December 2021 (UTC)

@Zzuuzz: Hmm, in each of the four examples you linked to, it looks like the revision deletion of the username happened after the rollback. In order for the a hidden user text to be generated, the revision deletion of the username needs to happen before the rollback occurs. Mz7 (talk) 06:05, 17 December 2021 (UTC)

That would probably explain a few things. It's a very rare workflow, eh? -- zzuuzz ^(talk) 06:19, 17 December 2021 (UTC)

Right, yeah, exactly. Mz7 (talk) 07:35, 17 December 2021 (UTC)

Thanks for helping clear that up. I think we can extend the same principle to the existing statement about oversight/suppression, and also extend it to global hidden locks (which aren't and don't need to be mentioned). It's probably useful to know, but it's so rare I've only seen it a handful of times. I think as it's framed it provides a bad example, giving the impression that it doesn't matter what the edit summary contains. I hope to return to this page to sort this out some time. -- zzuuzz ^(talk) 11:45, 17 December 2021 (UTC)

Misuse

What should one do, when one's edits have been reverted with rollback in contravention of the stated criteria for its use? Lognomm (talk) 11:59, 10 March 2022 (UTC)

Wikipedia:Sockpuppet investigations/Best known for IP. Mako001 (C)  (T)  🇺🇦 12:30, 10 March 2022 (UTC)

Strange answer. Why exactly did you roll back my edit? Lognomm (talk) 12:50, 10 March 2022 (UTC)