Wikipedia talk:Edit warring/Archives/2013/November

Definition of "Revert" and "Undo"
There appears to be a need for an explicit definition of the words revert and undoes in this policy.

at Administrators' noticeboard, we see an admin asserting that "Technically, any change, no matter how small to the text of an article, is a revert."

IMO this has some obvious flaws:


 * Am I really reverting someone if I add a new sentence to an article, without touching anything written by anyone else? That's "any change, no matter how small, to the text of an article".
 * What if you improve the meaning of the sentence I just wrote, so that it more correctly reflects the cited source? That's "any change, no matter how small, to the text of an article".

If we go through this process—which you and I might call collaborative editing—three or four times, should we be blocked for "reverting" each other in violation of 3RR? It just doesn't seem appropriate to me.

However, this definition was put forth by an admin and seems to be sincerely believed and accepted by some people.

Part of the cause for this confusion is that "reverting" is briefly defined in the lead here as "undoing", but "undoing" does not appear to be defined at all. Are you "undoing" my work if you improve a sentence that I just wrote? Is it only "undoing" if I disagree with your change?

What do you think would best address this problem? WhatamIdoing (talk) 19:09, 21 November 2013 (UTC)
 * I would be careful not to take one admin's definition of the terms (especially an admin who acknowledged they were under considerable stress at the time and may have misspoken) as any sort of consensus.
 * I'd like to think we don't need to actually define the word "undoing", but if the consensus is that we do, perhaps something along the lines of, "Making an edit that reverses in whole or in part changes made previously to the article by a prior editor." DonIago (talk) 19:19, 21 November 2013 (UTC)
 * FWIW, I've explained my thinking on the issue here. Mark Arsten (talk) 19:34, 21 November 2013 (UTC)
 * A "revert" is poorly defined and open to interpretation, so it's not surprising that admins respond in different ways to perceived violations. I outlined my concerns just over a month ago here. I think we should scrap the concept of a "revert" altogther, since reverting or altering another editor's edit as part of constructive editing has its place, yet it can still put you over the line as far as 3RR violations ago. The real problem edits are "redos", when an editor repeatedly reinstates an edit that has been rejected. It is a "redo" that incites edit-warring; it is a redo that breaches the spirit of the BRD cycle. Sustained "reverting" can encompass constructive editing such as refining content, but sustained "redoing" is what creates edit-war cycles. Betty Logan (talk) 19:53, 21 November 2013 (UTC)
 * That's a very interesting thought. Thank you for sharing it.  WhatamIdoing (talk) 23:17, 22 November 2013 (UTC)
 * Agree with Betty that redo-behavior is the key to edit-warring. Sometimes that is manifested as a click of the undo-button:  if every time somebody adds some phrase with a particular meaning, another editor clicks undo, then we have an edit-war.  There are cases where redo-behavior is permitted, but they are very rare:  deletion of COPYVIO, deletion of BLPTALK violations, probably some stuff related to WP:OUTING, maybe a couple other things.  But straightforward-clicking-undo isn't the only way that redo-the-revert-behavior manifests.  See my posts below for some examples.  Of course, *many* kinds of editing are redo-behavior, which don't involve reverts.  Somebody was recently blocked, for zooming through a few hundred articles, making one change to each:  replacing the flag of Hong Kong, with the flag of the PRC.  They never performed a redo-behavior, in any single article.  They never clicked undo.  But their behavior was, is, and forever-more will be, redo-behavior:  the kind that leads to edit-wars.  74.192.84.101 (talk) 14:38, 30 November 2013 (UTC)
 * I agree with Mark's comments. As a major patroller of WP:ANEW, I evaluate each report on a case-by-case basis. Just because I gave the literal interpretation of a revert doesn't mean I block someone because they made a minor change to an article.--Bbb23 (talk) 01:38, 22 November 2013 (UTC)
 * May I encourage you to check a dictionary about the literal interpretation of that word? revert, especially the most relevant transitive definition (#3), is noticeably more restrictive than "any change, no matter how small".  Revert is very much a concept of going back to something that previously existed.  Going forward—e.g., adding a new word or new sentence—is not a reversion according to any English dictionary on this planet.  WhatamIdoing (talk) 23:17, 22 November 2013 (UTC)
 * I had no idea this discussion was here. I was asking Bbb23 about this on his talk page and he was too nice to point out to me that I was missing the boat!
 * Anyway, I think I agree with WhatamIdoing. I would add one thing: The policy should clearly lay out the widest effective interpretation of the definition, not a narrower one.  I don't think it is right to lay out the definition in a way which is interpreted by a significant proportion of reasonable editors as being narrower than how some administrators interpret the definition.  This is because these reasonable editors may be breaking the rule in the judgement of an enforcing admin, even though by their reasonable reading of the policy, they do not understand why they are breaking the rule.  Imagine a hypothetical 100RR:  "User must make less than 100 reverts in a month."  If administrators were understanding "100", as many do in scientific contexts, to have only 1 significant digit, and rounding up so that any number of reverts over 50.0 would be considered infringing, then such a policy should clarify this, as many reasonable editors would understand "100" as having 3 significant digits.  Similarly here:  If administrators are understanding "revert" to mean any edit to a page which was last edited by another user, then the policy should clarify this, as many reasonable editors would not understand this. -- Atethnekos (Discussion, Contributions) 04:21, 22 November 2013 (UTC)

I made a little diagram to show the relationships between the sets associated with the different definitions given. This way any one can tell me easily how I've misunderstood what they meant. A box being surrounded by another should indicate that that box represents a proper subset of the other. So A is all edits, and B is all edits minus null edits, where some null edits exist, etc.
 * This issue was previously discussed (see archive) with a stricter definition for 1RR (which is essentially the same as 3RR but comes on a little earlier). I don't think if two editors are collaboratively editing there is any problem, when one editor is openly reverting and claiming the same then it becomes an editwar. It is true and many editors who have been editing as a team do it without discussing everything on talk page because they get each other's edits. This is not the case with a clear edit war and looking at whether you added a completely new sentence or edited something already in the article (technically a revert). Maybe the archive can add additional perspective to the existing consensus. -- lTopGunl (talk) 04:36, 22 November 2013 (UTC)
 * A = All edits
 * B = "any change, no matter how small to the text of an article"; Bbb23 technical definition, where "text" is understood as "wikitext"
 * C = Any change to an article which was last edited by another user; charitable interpretation of Bb23's technical definition by User:QTxVi4bEMRbrNqOorWBV
 * D = "undoing the effects of one or more edits"; quoted by Mark Arsten as the "technical definition"
 * E = "undoing the effects of one or more edits, which results in the page being restored to a previous version"; definition from Help:Reverting
 * F = "any edit ... that reverses the actions of other editors, in whole or in part, whether involving the same or different material." From WP:3RR; naive interpretation taken by User:QTxVi4bEMRbrNqOorWBV and others.
 * G = F minus exemptions listed at WP:3RR. -- Atethnekos (Discussion, Contributions) 05:54, 22 November 2013 (UTC)

The purpose of the 3RR is to address warring behavior
Focusing on "technical reverts" without any consideration of identifying warring behavior runs counter to the intent of the 3RR.

What is warring behavior? It involves destructively undoing another editor's changes without discussion. Why did I say "destructively" undoing? Because two editors collaborating to find better wording is not warring. Collaboration is a push to improve an article, with editors building off each other's changes. Warring is a push against another editor, with no attempt to discuss or build upon new changes. Consider:


 * 1) In an article's talk page, a user writes a voluminous explanation for a proposed change and a voluminous RfC to address conflicts. In the mainspace edit comment, the user links to the explanation and the RfC. Further, the user says "bold edit" in the edit comment, meaning, "I am inviting the WP:BRD process if you disagree."
 * 2) Editor A changes "some foo" to "some foo and bar". Editor B changes "some foo and bar" to "some". Editors A and B are now satisfied with this new wording. This is collaboration. Editors are making adjustments in response to each other's changes.
 * 3) A user does a full revert of a newcomer's edit while providing a lengthy explanation for the revert on the talk page (multiple violations of policy).

These are examples of good editing; they are the antithesis of edit warring. A user that does each of the three examples above should not be issued a warning for edit warring. This could only result in a 3RR warning if we dismiss the purpose of the 3RR, which is to address warring behavior. vzaak (talk) 18:25, 22 November 2013 (UTC)

"Technical reverts" and WP:BRD
If we're going to adopt this new "revert sense" where any edit to text, no matter how small, could be considered a technical revert, how would this affect the WP:BRD cycle? Wouldn't the B be a technical revert? Should we change it to WP:RRD? jps (talk) 19:27, 22 November 2013 (UTC)
 * Well, yes, that would be the case under that definition, unless you were the only person to have edited the article before or were boldly creating the article. But I don't think that anyone is really going to accept that definition after having the time and mental energy to think it through.  What we need is a definition that will really work for us, so that we're all on the same page.  It might even be useful to have a supplement that gives examples, e.g., I add a sentence and you remove it:  that's a revert.  I add a sentence and you move it to a different place on the page:  that's not a revert.  WhatamIdoing (talk) 23:23, 22 November 2013 (UTC)

Proposal
I would like to propose that the policy include the following two sentences as clarification:

"A revert is not simply an edit that modifies a previous editor's contribution. A revert is specifically an action that either removes or negates a previous editor's contribution."

jps (talk) 01:57, 25 November 2013 (UTC)


 * This still gives a false positive in the real-life example of "public support" -> "public and scientific support" -> "support". It takes a human to understand that the last change is a compromise rather than warring. I don't think a context-free definition of revert is possible; the focus should be on whether a change indicates warring behavior. vzaak (talk) 15:41, 25 November 2013 (UTC)
 * Well, there's another aspect to this too: "Reverting another user is not always a sign of an edit war." What I would like to encourage is a modicum of analysis. The wording as is does not imply that someone should do any analysis whatsoever. Please, propose some other wording if you can think of some. Try to keep it simple and short as that's the kind of addition likely to get more support. jps (talk) 17:33, 25 November 2013 (UTC)
 * Let's build on that. Is this helpful?


 * Does that seem like an improvement? Can anyone improve it? WhatamIdoing (talk) 16:57, 26 November 2013 (UTC)


 * That does make the case more explicit, although one could argue that any reasonable reading of WP:EW makes it unnecessary, especially when considered alongside things like WP:EDITCONSENSUS, as you cited. The mission to fill loopholes by expanding the length of the WP:EW article may be a quixotic one, and before embarking on it perhaps we should confront the matter of whether it is necessary? I am not aware of anyone who would count constructive WP:EDITCONSENSUS collaborations as reverts, save for one exception. Significant problems should be addressed by policy, while isolated, individual problems should probably be addressed individually. In 2003, the year Arnold Schwarzenegger was elected governor of California, a senator introduced a bill that would allow a naturalized citizen to become President of the United States, a move criticized for ostensibly being about the "equal right to govern" while actually just targeting one individual. vzaak (talk) 04:12, 27 November 2013 (UTC)


 * The definition on the page seems to work well: "A revert means undoing the actions of another editor." SlimVirgin (talk) 00:02, 28 November 2013 (UTC)
 * Except that some administrators are interpreting "undoing" as "modifying". This needs to be addressed. jps (talk) 05:24, 28 November 2013 (UTC)
 * Modifying counts as a revert if the effect is to undo something that has become an issue. Everything depends on context, and if it's something that goes to AN/3RR it will hopefully be dealt with by an admin who's used to interpreting these things. On the whole, admins will tend to give the benefit of the doubt. But the essential point is that a revert undoes someone's work. SlimVirgin (talk) 20:52, 28 November 2013 (UTC)
 * Unfortunately, when SlimVirgin says that everything depends on context, they are correct. What does it mean to "removes or negates a previous editor's contribution"?  There is the obvious definition:  that you deleted everything they wrote, in one of their previous edit-transactions.  Some folks see that as the *only* definition.  If you merely insert some text, especially some new sentence, Whatamidoing has opined that you cannot possible have changed what the previous editor wrote.  "The previous sentence is false."  Uh oh.  Or how about:  "But in the end, everything described in this paragraph turned out to be a dream sequence / big joke / pseudoscience /  total nonsense".  Double uh-oh.
 * In real editing, rarely will you see something so clear-cut, but it can happen. More commonly, though, you will see editors in contentious articles engaging in point-counterpoint style, where one adds a sentence pro, then the other adds an adjacent sentence con.  This is not necessarily bad, as long as *eventually* it resolves into a well-written exposition that describes all Reliably Sourced facts, giving the readership a clear picture of how the sources disagree, by neutrally describing that disagreement.  But in the short run, the prose becomes a whiplash-risk to the readership, if they're paying close attention.
 * I guess my main problem with the bare formulation as "removes or negates" is that it should not apply only to the glyphs... it should apply to the intended *meaning* to include the connotations. Edit-warring can be subtle; most isn't just is-not / is-too / is-not / is-too.  Collaborative editing, where editors work towards consensus in real-time, via mainspace revisions and edit-summaries, is a Good Thing.  But tendentious editing, where editors (alone or in groups sharing the same content-goals) gradually erode away the previously-intended meaning, gradually chip away at Reliably Sourced phrases to downplay them, bury them in their opposites, or just outright wipe them... those are all Bad Things... and there may not *be* a clear unambiguous zero-false-positives rule which can hope to distinguish the two styles, rigidly and explicitly.
 * If we give some inflexible definition of revert, and insist admins *only* follow that definition, ever, then methinks we open the doors to gaming the system. Almost any firm definition will have loopholes, subject to wikiLawyering, that permit meaning-reverts... which cannot be called reverts any longer, due to the strict definition!  But this is a double-edged sword.  If we stick with our current fuzzy definition, where a revert is context-dependent and path-dependent and the admin has to use their judgment, people will complain that admins are corrupt, and feel they were treated unfairly, because quite frankly they *will* have been treated unfairly, any fuzzy system which depends on the sheriff to Do The Right Thing is subjective by definition.  I'm not sure what reform of the system we might need... but calls for rigid definitions rigidly enforced, are directly in conflict with pillar five, and probably also pillar four.  HTH. &mdash; 74.192.84.101 (talk) 03:24, 30 November 2013 (UTC)

The essential point is that a revert undoes someone else's work
If we emphasize this in the policy, it will prevent the unfortunate run-in we had and what started this section. I suggest;

"The essential point of revert limits is that a revert undoes someone else's work."

as a modification to the definition so that admins are not tempted to interpret the policy as meaning any modification is a revert. We can discuss including a second sentence above as well.

jps (talk) 12:05, 30 November 2013 (UTC)
 * That is a firm, clear definition. If you click undo, you reverted, if you did not, you did not.  See my post above for why that seemingly-obviously-good characteristic is so dangerous to harmonious editing.  If your change undoes my intent, without clicking undo, is it a revert?  Well, that depends, doesn't it -- on your intent.  If you are working collaboratively with me, editing together in mainspace, working towards a final version we can both be satisfied with, then you did not revert, even if you repeatedly changed my meaning, deleted my phrasing, and so on.  But if you are working tendentiously against me, thwarting my edits to mainspace, working towards a final version that only you will be satisfied with, then even if you did not technically click undo, you still absolutely reverted me.

The essential point of revert-limits is that a revert thwarts someone else's intent/efforts/meaning/prose/work. Intent is key. Done in a friendly collaborative spirit this is great; done with a battlefield mentality this is awful. Admins must judge carefully which sort of event happened. Pillar five and pillar four, especially, are applicable.


 * The fuzzy definition we have now, which is something like what I've said just above methinks, gives the judgment call over which was happening to the nearest admin, one of 1433 or so. The firm rule you are suggesting, means that subtle just-short-of-war editing, simply cannot be prevented, short of the nearest admin invoking WP:IAR.  74.192.84.101 (talk) 14:22, 30 November 2013 (UTC)