Wikipedia:Village pump (proposals)/Archive 136

Survey: AWB bot ref reordering

(no threaded discussion)

• Comment - Note to closer: I strongly feel that a no-consensus result here should mean no AWB reordering; i.e. the consensus burden is on those who wish to reorder. Consensus should be required to make a controversial change, and that change occurred in 2009. ―Mandruss ☎ 21:18, 19 November 2016 (UTC)
• Comment I don't think a "let's pretend" (as in "Let's pretend it is 2009") RFC is a good idea. I would rather discuss either the substantive issue or a way forward. All the best: Rich Farmbrough, 22:47, 19 November 2016 (UTC).
  
• No. There should be no automated re-ordering of refs to retain the chronological appearance of footnote numbers, particularly not over objections. Previous discussion: Nov 2016, Nov 2016, May 2015, Feb 2015, Feb 2014, July 2010, May 2010, April 2010. The feature was added to AWB in 2009 with very little input. It should have been discussed more widely, because editors may add refs in a particular order for editorial reasons. A simple example:

Mary likes cake, but John doesn't.<ref name=Mary/><ref name=John/>

Someone will argue that we could place the Mary ref inside the sentence, but when there are lots of refs and several clauses, it looks cluttered to do that. And there may be other reasons for the order, including that the first ref is a primary source and the second a supporting secondary one; that the first discusses the issue more clearly than the second; or that the first is the source most experts in the field will expect to see.

According to WP:AWBRULES, "If challenged, the onus is on the AWB operator to demonstrate or achieve consensus for changes they wish to make on a large scale". Study 329 is one example of what happens instead: AWB reordered refs on 4 May 2015. I moved them back. A second AWB editor reordered them on 5 May. I added bots deny to stop it. The second AWB editor removed bots deny on 9 May. A third AWB editor reordered the refs again on 9 May. SarahSV ^(talk) 00:15, 20 November 2016 (UTC)

• Yes/Reforder Like I said before, refs should be in order. If you have ^{[2][1][14][4]}, you're telling the reader to jump around a list of reference for no good reason. ^{[1][2][4][14]} is the natural way of presenting things. Now if some specific article warrants deviation from this for some obscure reason, then deny AWB on that specific article, but this is a legit fix in 99% of cases. The usual workaround is to use {{R}}, which AWB will leave alone. The <1% of articles that warrants a deviation from this behaviour shouldn't block efforts to improve the presentation on the >99%. I've cleanedup thousands of articles making those fixes for the last 6 or so years, and I've had 2 complaints. A fail rate of 2/10,000+ is more than acceptable. Headbomb {talk / contribs / physics / books} 00:35, 20 November 2016 (UTC)

I wonder, Headbomb, if there is some misunderstanding here as to how cite numbers are used as cites are in order until someone using AWB comes along and changes them. The cite numbers relate to the order the cites were placed in the article as a whole, not necessarily in the sentence in question. When a cite is used more than once, the number remains the same throughout the article, so the numerical order of cites may appear disjointed in places. Sometimes there may be several cites in one sentence, and if placed next to the information they are citing (which is often done) then the sentence may read "Splot was born in Swansea,[13] on 13 March 2000,[7] to parents John Splot, an organ grinder,[5] and Mary Splot, a Suffragette.[14]" This is appropriate usage, and no AWB would change the sequence order, and the reader could check each item they were interested in by clicking on the number, and going to the relevant citation. However, in order not to inhibit reading flow unnecessarily, we encourage editors to place all the cites at the end of the sentence if none of the statements are likely to be contentious. When placing them at the end of the sentence or paragraph we keep them in order per WP:CITEFOOT: "it is usually sufficient to add the citation to the end of the clause, sentence, or paragraph, so long as it's clear which source supports which part of the text." When someone using AWB shuffles the cites, they are no longer logical or natural, and the reader may well end up confused as to which cite belongs to which part of the sentence, because the natural order has been disturbed. It seems that at the moment when cites are grouped at the end of a sentence some well-meaning AWB user will shuffle the order without realising what they are doing. We should either stop AWB reshuffling the cites, or change our guidance on grouping cites at the end of sentences. SilkTork ^{✔Tea time} 13:57, 21 November 2016 (UTC)

[Since SilkTork has repeated this in the #Clarification please section I have moved my reply there.] All the best: Rich Farmbrough, 20:24, 21 November 2016 (UTC).

• Yes As I said earlier, whatever purpose the non-chronological ref order serves may not be clear to the reader. The benefits of doing so, as stated by headbomb above, are clear. Pppery 00:54, 20 November 2016 (UTC)

Poppery, I have replied to Headbomb's comment, making reference to our existing guidance on how cites are placed at the end of a sentence. Does that make it clearer? SilkTork ^{✔Tea time} 13:57, 21 November 2016 (UTC)

• Yes per my discussion below, especially noting that article editors (assuming a consensus) can override the reordering. Ultimately there is no harm and having cites in chronological order should be seen as naturally more convenient for the reader. Stevie is the man! ^{Talk • Work} 01:01, 20 November 2016 (UTC)
• Yes - numerical ordering seems almost universal, if not universal, in modern publishing. The "John and Mary" example above is bad referencing. In the rare case that someone insists on out-of-order referencing they have the means to achieve it simply. All the best: Rich Farmbrough, 01:03, 20 November 2016 (UTC).
  

Hi Rich, see my reply to Headbomb's comment. I wonder if there is some misunderstanding on how the cite numbering system is being used in an article, because nearly every article will have cite numbers appearing in non numerical order as some are being re-used. Perhaps what is needed is a piece of software that identifies when a cite is used more than once, so people could understand that the number relates to an earlier placing in the article rather than the current one. SilkTork ^{✔Tea time} 13:57, 21 November 2016 (UTC)

It is not a question of non-numerical order, but of adjacent cites in non-numerical order. I don't think people have an issue with understanding that a footnote has been referred to again, indeed there is meta-data conveyed by seeing ^[3][43][128] - albeit unreliable and unintentional.

I suppose that the superscripts could be of the form^[2a] and^[2b], but I don't think this solves any real issues.

All the best: Rich Farmbrough, 20:34, 21 November 2016 (UTC).

• Supplementary Various editors have said that they arrange reverences in order of: i) relevance, ii) importance, iii) date and iv) the items they support - as well of course as the industry standard v) numerical order. While these schemes may have something to recommend them (except when they are WP:SYNTH) I see no way for a reader to know in advance which of the five applies, or whether it is actually vi) happenstance. Without this the value of these alternative schemes is pretty much moot. All the best: Rich Farmbrough, 19:55, 4 December 2016 (UTC).
  
• No. If an editor reviews the refs and decides they should be reordered, that's fine, but since there are occasionally reasons to have non-ascending order, this ought not to be done by bots or by semi-automated edits. Mike Christie (talk - contribs - library) 01:04, 20 November 2016 (UTC)
• No, but (see comment below) I think the wrong question is being asked here. And aside from personal preferences, why should it matter so much that ascending order should be mandatory? ~ J. Johnson (JJ) (talk) 03:04, 20 November 2016 (UTC)
• No Editors can, and do, arrange adjacent references to put to the most relevant one first, followed by more secondary ones; it is not "universal" for citation numbers to always appear in increasing order when they are adjacent. The usual rule in WP:CONTEXTBOT, part of the bot policy, is that bots should avoid making changes (such as changing spelling or rearranging references) which would require a human to review the context and determine whether a change is appropriate. That is the case here - only human review can determine the best order of the references. Given that the references are hyperlinked and numbered, there is no difficulty in finding them in the reference list regardless what order they appear in the body text. — Carl (CBM · talk) 04:34, 20 November 2016 (UTC)
• No I frequently arrange references in a particular order: sometimes most relevant first, sometimes in date order. (An example of the latter would be when writing something like: "Many studies between 2004 and 2016 have shown ..." – the references supporting this need to show clearly the time span involved.) Bots should not make context-sensitive changes, over-riding editors' choices. Peter coxhead (talk) 07:54, 20 November 2016 (UTC)
• Yes because it looks messy. In any given article, the ordering of the reference numbers could be accidental, or could be trying to signal some subtle distinction between the references. But readers cannot tell the difference. If editors intend the ordering of the reference numbers to mean something to the reader, then it would be better to find a different way to express that meaning. -- John of Reading (talk) 08:11, 20 November 2016 (UTC)
• Yes Having references appear in numerical order in the text is part of basic proofreading. Not doing it gives the appearance of a typed term paper prepared by a schoolboy for whom arranging for such niceties is real work, requiring the retyping of the entire paper. this is one of the many points of good style which were difficult to achieve with a typewriter but were the mark of professional work; they're now easy, and we should take advantage of it . DGG ( talk ) 08:24, 20 November 2016 (UTC)
• No - as per CBM and Peter's comments. Hchc2009 (talk) 08:30, 20 November 2016 (UTC)
• Yes; useful, not a problem for the vast majority of pages, and easily solvable for those which need a different order. Jc86035 (talk) Use {{re|Jc86035}}
  to reply to me 09:50, 20 November 2016 (UTC)
• No per SlimVirgin aka SarahSV. In the precursor discussion, I showed an actual real-life case where I feel my ability to control the cite order was useful to at least some readers. That was not disputed. I strongly suspect other editors could point to other such cases. In contrast, to my knowledge no one has shown actual real-life cases where readers benefit from ascending sequence, and that claim is strongly disputed. The arguments in support of reordering are therefore nothing but unproven theories, unsupported assertions. ―Mandruss ☎ 10:13, 20 November 2016 (UTC)

  "No one has shown actual real-life cases" How about nearly 100% of professional journals putting references in order? Headbomb {talk / contribs / physics / books} 10:52, 20 November 2016 (UTC)

  Taking your word for that, how does that show benefit to the reader? ―Mandruss ☎ 11:31, 20 November 2016 (UTC)

  For instance, many books that I read have endnotes divided by chapter, with numbering starting over for each chapter. When I go to look up the note, I see that here are the notes for Chapter 5 and for Chapter 6, but I have no idea what chapter I'm on, so I have to go back to the page where I saw the note, and if the chapter title isn't in the header, I have to flip back top the top of the chapter and find out what the chapter number is. Only then can I actually look up the note. That's pretty standard practice in book publication (although there are the nice books that list the name of the chapter in the notes section, or the page numbers the notes fall on), but it just because it is standard does not mean that it is beneficial to the reader. It's just the way things are done. To say that notes out of order are confusing is untrue because even when they're re-ordered, they don't go 1,2,3,4... in the body of the article, they go 1,5,8,14, because that's where the references just happen to fall. I see no benefit to 1,5,8,14 over 14,5,8,1 when 14 is the most important reference for that section and the others are less important. Give the reader what is primary first, not the reference that just happens to have a low number because it appears earlier in the article then another. There is no intrinsic quality to the number of a reference, it's all just happenstance. Beyond My Ken (talk) 23:49, 22 November 2016 (UTC)

• Reforder - reasonable compromise. Out-of-sequence limited to those cases that need it per editor discretion. The idea of the Reforder option is that intentional out-of-sequence ordering will be far more effective if not lost in a forest of unintentional out-of-sequence ordering. In that sense AWB can be the friend of editors who use cite ordering, not their adversary. ―Mandruss ☎ 12:30, 20 November 2016 (UTC)
• No. Automation should not override editorial discretion. There's no justification for the claim that refs not in chronological order would confuse readers. Few readers check references. Those that do -- the more scholarly ones, I'd say -- will be more than capable of choosing the order in which they check them. --Stfg (talk) 10:55, 20 November 2016 (UTC) Please note: I oppose the Reforder option added this morning. It would require editors who have already created articles to go back to them to add the option retrospectively. --Stfg (talk) 12:49, 20 November 2016 (UTC)
• No - It is reasonable that a more relevant source would be cited before one with less relevance when both are being cited back to back.— Godsy ^(TALK_CONT) 12:35, 20 November 2016 (UTC)
• Reorder - per Headbomb's continuous remarks. --Dirk Beetstra ^{T C} 13:22, 20 November 2016 (UTC)
• No - As stated above, editorial decision should not be overridden by automation when there is a reasonable purpose in the decision designed to aid the reader. More broadly, we should be an encyclopedia which primarily cares about content, not one which is rule-bound to the detriment of content. Beyond My Ken (talk) 18:01, 20 November 2016 (UTC)
• No — "Quotation stating idea"^{10 (source of quotation)1 (ref supporting statement)} should not be reordered. I also put the best reference for a statement first in the list. Glrx (talk) 18:35, 20 November 2016 (UTC)
• Yes – It makes no sense for ref 5 to go after ref 6 if it is invoked. Dustin (talk) 19:09, 20 November 2016 (UTC)
• No, although the damage is done. There are probably hundred thousands of articles^{[notes 1]} with references in the wrong order (in the sense of being different than any actual editor intended). I see no benefit to the reordering, and some (often slight, but occasionally serious) damage to understanding. (As an aside, a bot should tag all articles with multiple consective references/notes which were ever touched by AWB or a bot, for potential errors caused by this mistake.) — Arthur Rubin (talk) 22:01, 20 November 2016 (UTC)
• Yes because non-ordered references only are confusing to readers -FASTILY 22:49, 20 November 2016 (UTC)
• No per Sarah. Nikkimaria (talk) 00:53, 21 November 2016 (UTC)
• Yes, reorder. I wasn't aware this had been disputed for seven years, or really ever, because it seems completely obvious they should be in numerical order. I don't much care what "professional journals" do - we're the 800-pound gorilla of book-report plagiarism material and bar-bet-settling, so we can do what we want ;) But I do care that readers actually understand what they're looking at, and [15][5][2][12] is harder to read and forces people to jump around in the reference list. "But the order I wrote them in is important!!" is the kind of thing I can imagine convincing myself of while in the middle of writing an article, but it's self-delusion; the number of readers who would see that, assume the order was significant, correctly interpret why the author ordered them as they did, and actually care about the result, is indistinguishable from zero. Worse, without automatic reordering, the likelihood that any given out-of-order reference list is significant actually declines, because almost nobody is going to reorder them manually. There are workarounds for the rare circumstance where order actually does matter; use those if you must and let the rest of us sit back and let the bots fix our sloppiness. Opabinia regalis (talk) 04:10, 21 November 2016 (UTC)

"Nobody is going to manually reorder them," means that the order is not that important. I reordered several long, ref lists a few times, before stumbling upon a bot edit that was undoing my work. The edit description gave no indication of what the bot was doing. Workarounds make it more difficult for other inexperienced editors to understand and participate in updating, correcting or improving the text. Bcharles (talk) 09:57, 23 November 2016 (UTC)

• No. I complained ref reordering in gen fixes last year. The main reason is because while it might be more aesthetically pleasing to some, I order my refs by priority (best refs first, secondary or duplicate refs second). If a sentence needs multiple footnotes and the citation style doesn't suit shoving both lines into the same ref, ordering by priority makes the most sense. (This has not been contested across multiple FARs so I consider it a common sense understanding. It's also silly to call this "delusion"—readers check the refs in order.) Gen fixes are supposed to be inoffensive, which is why they minimize white space disruption and so on. Gen fixes should not impede productive work in drive-by ref reordering. czar 05:39, 21 November 2016 (UTC)
• No unless we change WP:CITEFOOT to alert editors that a bot may change the logical order of end sentence cite into a purely numerical one, so if the order if relevant then the cites should be placed right next to the relevant information, and if this inhibits reading flow (so making grouped cites preferable) then to put in a command to prevent a bot changing the order. I do think though that it would be simpler to stop AWB from reordering cites. SilkTork ^{✔Tea time} 14:09, 21 November 2016 (UTC)
• No I often will place references in a particular order, with a more specific reference first up and a more general secondary source to follow. Jumbling up will be a disservice to the reader. The foot notes are not hard to find in our online edition, since you just click on the number and the brower navigates and highlights the reference. Graeme Bartlett (talk) 22:41, 21 November 2016 (UTC)
• Yes with the understanding from several of the "no" !votes above that there is a legitimate time and place to use references in non-numerical order (several valid examples cited), and that we have at least one way, via {{R}}, to "protect" that order from bots or random gnomish editors that would otherwise come by and fix it. I understand the concern of this being added to AWB, but there are lots of wikignomes that will do this ref order adjustment to, and, unlike something like DATERET where we have templates to alert editors to a selected style, a reforder aspect will not be clear unless one uses inviscomments or a template like {{R}}. Setting up guidelines when one should force a ref order, and how to indicate that to any other editor without minimal of fuss, would be useful result from the above discussion. I think what we have to recognize is that most readers and editors appear to see the mixed numerical order of a forced reforder as "wrong" and will want to try to fix it, being just as unaware of any previous talk page or consensus discussion that AWB has, and we should find a way that doesn't affect this but does keep editorial discretion, of which there is at least one way; more options should be explored. --MASEM (t) 17:17, 22 November 2016 (UTC)

This argument amounts to saying "We won't be able to stop gnomish editors doing it anyway, so let's allow AWB to do it." It seems a very unsatisfactory reason to me. Peter coxhead (talk) 17:44, 22 November 2016 (UTC)

• No. There are good reasons for manual control of reference order, for example so that they reflect the logical sequence of the statement they substantiate, or emphasising a key reference over subsidiary ones. The burden shouldn't be placed on content editors to "protect" these decisions from AWB operators. No professional copy-editor in an academic context would arbitrarily re-order the references as written by the author! Joe Roe (talk) 18:57, 22 November 2016 (UTC)

  The idea of the Reforder option is that intentional out-of-sequence ordering will be far more effective if not lost in a forest of unintentional out-of-sequence ordering. In that sense AWB can be the friend of editors who use cite ordering, not their adversary. ―Mandruss ☎ 19:39, 22 November 2016 (UTC)

  This isn't the case at all; in my field (biochemistry/bioinformatics) I can't find any journals that would allow multiple numbered references in non-sequential order. Here's the American Chemical Society guide, for example: When citing more than one reference at one place by number... list the numbers in ascending order. Opabinia regalis (talk) 20:03, 25 November 2016 (UTC)

• No. The use of references in Wikipedia has historically been tolerant of a wide range of reference styles and philosophies. There have been arguments and controversies, but the consensus is that uniformity is not necessary. We have two major styles of templates (now wrappers of a single implementation). We use parenthetical as well as numbered references. We have references declared in the text and references instead declared in lists at the end. We have many editors who avoid templates in favor of manually formatted references. In this spirit automating a reordering of reference numbers (I assume "chronological" doesn't mean the date of the reference) in the face of even a minority opposition goes against this long standing consensus. StarryGrandma (talk) 19:20, 22 November 2016 (UTC)
• No. Wiki articles are continually changing, and the numerical order of refs will sometimes change over time. The placement of refs in a sequence is often deliberate and should not be altered without consideration given to the context (by a human editor). The diversity of contexts in which issues of ref order occur indicates that one rule will not satisfy all cases. The numerical order in text is trivial relative to other considerations (e.g. primary, secondary, tertiary importance; relevance; clause order; list item order). Bcharles (talk) 09:57, 23 November 2016 (UTC)
• Big no. In a string of multiple refs, the most relative and directly supporting should be first, even if that makes the ref numbers not be strictly increasing. This is particularly true of the first ref because in practice I often have situations where one source cites 90% of the material in a statement and the remaining refs are much less important to the overall verifiability. I don't want the highly import ref to be shuffled among those less important refs. I want it to be first. Jason Quinn (talk) 00:28, 24 November 2016 (UTC)
• No - proposal is ill-prepared as to describing the mechanism, what necessitates it, and what the other consequences might be. It seems an frivolous automation that would be restricting possible editor wishes that might cause problems. Just because one can isn't a reason to do it, it's a reason things go wrong. Markbassett (talk) 20:58, 24 November 2016 (UTC)
• No. A substantial quantity of unordered refs is intentionally "disorderly", and the vast majority of those doesn't use {{R}}. In addition, there is also likely just as large a quantity of currently ordered refs, whose order is meaningful but not protected with {{R}}, because it has not yet occured to the article editors that a future edit can mix up ref numbers. All in all, this is a job that should be left to humans who leave human edit summaries, and not to AWB users, whose edits I never check on the watchlist, because I assume that the app knows what it's doing. Daß Wölf 00:40, 25 November 2016 (UTC)
• No If an editor spends significant time fixing issues in a particular article, they might reorder or otherwise rearrange some refs as part of that work—good! However, mindless reordering by bot is not helpful because someone who actually understood the topic and who actually read the references took a decision about how the references should be arranged. Many inconsistencies are tolerated (WP:ENGVAR and WP:CITEVAR) and always displaying numbers in ascending order is trivia. Johnuniq (talk) 02:22, 25 November 2016 (UTC)

• Well said! Beyond My Ken (talk) 17:31, 25 November 2016 (UTC)

• Alternatively, the editor who has spent significant time fixing an article was concentrating on the content and didn't try to redistribute the references so they'd come out in some allegedly significant but in practice inscrutable non-sequential order, because they knew that a bot would eventually stop by to fix it ;) Opabinia regalis (talk) 20:03, 25 November 2016 (UTC)

• I frequently modify my notes so that the bot that eventually comes by won't break them by trying to merge them using named-refs. ~ J. Johnson (JJ) (talk) 22:58, 25 November 2016 (UTC)
• That is probably the case quite often, Opabinia regalis, but not always, and it isn't fair to assume it. We're talking about an automated assumption here. If I read "Smith said 'the sky is green' and Jones agreed.^[m][n]", then I'd expect m to refer to the Smith reference and n to refer to the Jones. If Jones has been cited before Smith earlier in the article, then numerical ordering won't give this. The numbers reflect the order in which sources were introduced in the article, not the order in which they apply to a specific sentence. What is done in the journals in one specific field is irrelevant here, as is any opinion about what is "better"; the latter depends on your mileage. All we're talking about here is the extent to which automation should automate one way of going about it. --Stfg (talk) 13:50, 26 November 2016 (UTC)

  Obviously, the correct way to write that in a context where someone might come along and edit the text is "Smith said 'the sky is green'^[m] and Jones agreed.^[n]" Less obviously, but importantly, if you really want your out-of-order references to communicate anything to the reader other than "this article was sloppily edited", then you should be all for automated reordering. Since most instances of non-sequential references are unintentional and meaningless, removing them makes your intentional ones more likely to stand out. The simple solution to this problem is for people who think they're communicating something significant with their reference order choice to annotate this by using one of the already-available workarounds. Opabinia regalis (talk) 04:44, 27 November 2016 (UTC)

  I used to be in favour of mid-sentence citations for precisely that reason, but turned against them after seeing an experienced FAC reviewer give them very grudging acceptance. I'll reconsider that. For the rest, if language like "obviously", "if you really want ...", "sloppily" comes into it, I fear we'll just have to agree to disagree. Language like that does little to advance a rationale; more, perhaps, to express an attitude to those who disagree with you, but that's not my problem. This is only a matter of taste after all. I remain unconvinced. --Stfg (talk) 19:17, 27 November 2016 (UTC)

  The preferences of FAC reviewers, experienced or otherwise, is very much the tail wagging the dog; the huge majority of articles where this is done are not FA candidates. I don't mean anything by the language - "sloppily" was in reference to my previous vote, where I was talking about myself - but to be honest, I do think it's obvious that people who want to use non-sequential references to signal meaning are voting against their own interests here. Opabinia regalis (talk) 22:55, 28 November 2016 (UTC)

  Thanks. Fair point about FAC, but the reviewer did offer reasons. The main one was the appearance of fussiness. From my own experience, general readers can find those little superscripted numbers burdensome and may not realise that they don't have to check out every single one. There is a significant difference between an encyclopedia and a professional or academic journal here.

  A thought that crosses my mind is that professionalism (if that is what it is) in the presentation of citations really only matters at the top quality end of the encyclopedia, where one might hope that an article would have several watchers interested in maintaining quality and would take care over this. (Or am I being naively optimistic in that? :)) At lesser quality levels, the order of citation numbers is rather the bicycle shed compared to the nuclear power station of journalistic writing, gushing fandom, misreading of sources, self-promotion, and all the other horrors we all see every day. So at one end, I suspect, this automated action may be a nuisance; elsewhere, it's almost certainly an irrelevance.

  One point that I do think matters: I'm here because articles in which I took care over this sort of thing, and that I'm watching, suddenly got hit by this automation. Yes, I now know there are workarounds, but I didn't know them then and they are not widely advertised. The automation is implementing a preference of certain editors, not something the MOS requires. I am unhappy that that was done without consensus to do it. That probably can't be undone now, but imho it shouldn't continue until the real discussion has been held: not about what automation is allowed to do, but about what the MOS should say about the presentation of citations. --Stfg (talk) 10:56, 29 November 2016 (UTC)

  Perhaps people who think cosmetic appearance should be more important than any other kind of significance could be persuaded to not use named-refs. That is the simplest solution. ~ J. Johnson (JJ) (talk) 22:35, 27 November 2016 (UTC)

• No For reasons given by many others, this is placing 'tidiness' over any editorial judgement (best source, most relevant source, most comprehensive source, before secondary or supporting sources that may merely confirm some detail). A tool capable of ordering, in the hands of an editor familiar with content, might be useful for those who wish to use it, but having bots going around re-ordering with no notion of why the refs are ordered as they are, is not a net gain IMO. OK, many articles are reffed chaotically at present, not only 'un-numerically', but because of editors putting new refs at the end of the list, or conversely in 'pole position', rather than as a result of editorial judgement, but having refs look neat, but remain 'content/relevance chaotic' is not an improvement. The practical matter of checking a footnote in a printed text is completely different from clicking a ref-number. ps Summoned by bot Pincrete (talk) 18:36, 26 November 2016 (UTC)
• No Only a human being can determine the priority order of citation placements. At the very least, I can see an automated ordering wreaking havoc at DYK when reviewers start rejecting nominations because a citation is not readily where it should be. Oh, please, let's not go down this road. — Maile (talk) 02:41, 5 December 2016 (UTC)

  Provide an instance where this was an issue (at DYK). AWB already re-orders and has done so for a long, long time, so I am skeptical that that's a realistic issue. --Izno (talk) 23:31, 12 December 2016 (UTC)

• No - there may be reasons for the order, and only a human being has the ability to determine this. עוד מישהו Od Mishehu 15:47, 7 December 2016 (UTC)
• No, and I think we are doing our readers a disservice by presuming they are likely to get significantly confused by references in non-numerical order, much less so confused that they won't figure out what's going on after the first time they click on a few of those refnumbers. While a good portion of the 'numerically misordered' references may (or may not) be so by circumstance, there is enough reason to also presume that a significant portion is ordered so on purpose. This is not something a bot can determine. The main purpose of references first and foremost is and remains to be relevant to the content, not to look tidy, and forced numerical ordering is not likely to serve that purpose well. AddWittyNameHere (talk) 23:36, 11 December 2016 (UTC)
• No Humans are smarter than bots in this case. Should we also have bots re-order sections in alphabetical order of their title? Editors have a reason for ordering references, and sections, in the way that they do. It has nothing to do with math or alphabetizing. The John/Mary example above is a good one. First Light (talk) 04:02, 12 December 2016 (UTC)
• No, especially because in a case of WP:OVERCITE, an editor who is not an expert in the topic is liable to remove the second, third, etc., references as "redundant", when it might actually be the third one that is crucial, if they've been auto-rearranged by number. It's up to editors to decide which reference has primacy for a particular fact being sourced.  — SMcCandlish ☺ ☏ ¢ ≽^ʌⱷ҅_ᴥⱷ^ʌ≼  07:01, 12 December 2016 (UTC)

  That seems like a misuse of OCITE and thus a bad example to use as a reason against cite ordering in AWB. Besides, if the third citation is crucial, why is it not alone in the list of references? Or why is the explanation not provided to the reader in a note of some sort? --Izno (talk) 23:31, 12 December 2016 (UTC)

  Editors can rightfully ask for evidence that more than one real expert supports a particular statement, especially when it's controversial. Even that overcite essay supports leaving the three most reliable (out of six) citations in the example it gives. Adding yet another rule or guideline for citations (that editors should give an explanation in an extra note that most editors don't even know how to format) is truly overkill for a problem that doesn't exist. First Light (talk) 00:58, 13 December 2016 (UTC)

• Yes, though apparently in the minority at this late date. My opinion is that we should mercilessly reduce the number of references in any given sentence, and for the remaining, ordering makes a general sense. The statement "editors order these things deliberately" surely refers to a tiny minority of all editors (and this is in the spirit of WP:IAR in that most editors probably don't care, so they won't place their references deliberately). --Izno (talk) 23:31, 12 December 2016 (UTC)

  Aside: I happen to agree with Mandruss's request regarding a NOCON close. --Izno (talk) 23:32, 12 December 2016 (UTC)

Discussion: AWB bot ref reordering

• The discussion section below was created precisely for discussion of the substantive issue, as in any RfC. This is an RfC that should have occurred in 2009 after it became clear that the issue could not be resolved outside RfC. ―Mandruss ☎ 22:53, 19 November 2016 (UTC)

  Anyone could have raised an RfC at the time. They didn't. There is no procedure for your "Let's pretend that this is 2009." idea above. This is 2016, and the status quo is established. To attempt to set up an RFC in such a way that "no consensus" means you get the change you want, contrary to normal practice, is not how we do things. All the best: Rich Farmbrough, 23:08, 19 November 2016 (UTC).
  

  No one knew about it, Rich. It was added on the say-so of two editors, one of whom now supports its removal. Every attempt to have it removed has been ignored. SarahSV ^(talk) 23:13, 19 November 2016 (UTC)'

  Not so. I removed it from my version of AWB. All the best: Rich Farmbrough, 23:41, 19 November 2016 (UTC).
  

  Rich, I'm not sure what "not so" refers to. Several people have objected to it, but it has made no difference. Magioladitis said he supported removing it last year, but seems not to be able to do it himself. I asked "how can we get rid of that AWB thing that moves references out of place?" He replied: "I support you on that. In the past I proposed that we make the ref reordering optional and disabled for bots." Can you say how you were able to remove it? SarahSV ^(talk) 13:50, 20 November 2016 (UTC)

  I downloaded the code and patched it. Later I created a patching system to help avoid regressions. Whether either have survived the intervening years, I don't know.

  I cannot speak for @Magioladitis:, but it is possible that when he said "I support you on that." he was replying to "Can you address the larger issues, M?"

  The 2009 discussion had four participants not two. three seemed to think it was a good thing, and the fourth that it was harmless. I read the discussion and did not join in as it seemed an unexceptional change. Doubtless I was not the only one.

  The Study 329 example isn't a great one, because it's a John and Mary example (and the cites, and arguably the text, should be in the body, not the caption of an image).

  All the best: Rich Farmbrough, 15:30, 20 November 2016 (UTC).
  

  Seriously? The status quo is established? Because the AWB folks who had control over the code refused to recognize the fact that they had strong opposition to the edits and seek consensus for them? Sorry man, de facto consensus is not established by obstructionism. As for my "no consensus" position, I didn't set anything up that way, I made a suggestion to the closer, clearly identified as a comment, which they can ignore if it doesn't make sense to them. Please don't misrepresent what I'm doing here. ―Mandruss ☎ 23:16, 19 November 2016 (UTC)

  The question of when the status quo has been established long enough to count as the status quo, is, of course, a vexed one. However on Wikipedia seven (7) years would generally be considered ample time. And I council against using terms like "obstructionism" when, by your own admission you aren't familiar with the history. Rjwilmsi has bent over backwards to accommodate those who have legitimate issues with AWB (and some who don't), and together with the rest of the AWB developers has put in an enormous amount of work.

  All the best: Rich Farmbrough, 23:43, 19 November 2016 (UTC).
  

  If the reordering is removed pending consensus to include it, I will retract the obstructionism claim with apology. ―Mandruss ☎ 00:44, 20 November 2016 (UTC)

• So,are we talking about AWB or bots that run under AWB? Mr Stephen (talk) 23:23, 19 November 2016 (UTC)

• The issue is AWB or any automated re-ordering of refs. SarahSV ^(talk) 23:28, 19 November 2016 (UTC)

• @SlimVirgin: I confess I wasn't clear on that point, and I should have been. Should we change the headings here and the RfC question? ―Mandruss ☎ 23:29, 19 November 2016 (UTC)

• These are two very different issues. Perhaps we want different questions? All the best: Rich Farmbrough, 23:51, 19 November 2016 (UTC).
  
• Mandruss, this is about AWB reordering refs. It is about the feature that was added in 2009. If we make it more complicated and it confuses people, we will end up with no consensus. SarahSV ^(talk) 23:55, 19 November 2016 (UTC)
• These are two very different issues. Perhaps we want different questions? All the best: Rich Farmbrough, 23:51, 19 November 2016 (UTC).
  

• Well, since this is 2016 and not 2009, I will suggest that the burden is on those who want to end the reordering, as this is a long-established process. Having the footnote numbers in chronological order makes logical sense in many cases because it generally doesn't matter what citation is shown first and having the numbers in a sequence is naturally easiest on the reader's eyes. There are of course exceptions, as editors may think a particular citation is especially strong or backs up more of the content than other cities -- in these cases, they can override AWB by placing an HTML comment between the refs. If there is any change with how this all works, I wouldn't mind AWB adding an HTML comment that explains how to avoid reordering wherever it reorders. Stevie is the man! ^{Talk • Work} 23:56, 19 November 2016 (UTC)

• as this is a long-established process - As I said above, de facto consensus arises from the facts that (1) editors have had the ability to make changes to the content in question and (2) they have chosen not to do so. That is not the case here, because the AWB developers had control of the AWB code and were not responsive to editor complaints. They abused the power they had by virtue of technical control and it would be an exceedingly bad idea to reward that behavior by asserting "a long-established process". The message would be that if you just can be obstructive long enough, you will eventually win out. ―Mandruss ☎ 00:42, 20 November 2016 (UTC)

  I'm looking forward to seeing the evidence behind these assertions that don't appear to assume good faith. At any rate, the fact is that it's a long-established process. And based on the experiences reported so far, any specific disputes over reorderings are highly infrequent. Now, if someone can show where anything other than highly infrequent disputes are occurring, of course we should consider that. Stevie is the man! ^{Talk • Work} 00:58, 20 November 2016 (UTC)

  Which of the following is incorrect? 1. Highly controversial changes require clear consensus. 2. No clear consensus has ever existed for AWB reordering. 3. A developer with the power to control bot-executed code should be expected to be aware of 1 and 2. 4. The consensus-free code still remains in AWB.
  I cannot escape the logical conclusion that (1) someone has acted in bad faith, or (2) someone should not have control of the AWB code. As I've said elsehwere here, I will gladly retract all such claims if the reordering is removed pending consensus to include it, per item 1. ―Mandruss ☎ 01:16, 20 November 2016 (UTC)

  Where is the evidence that demonstrates "highly controversial" in the first place? As I have noted above, I have done a *lot* of these reorderings, with only one complaint, which was resolved. Stevie is the man! ^{Talk • Work} 19:25, 8 December 2016 (UTC)

  How about the 24 No !votes in this RfC to date, for starters? ―Mandruss ☎ 19:28, 8 December 2016 (UTC)

  Well, that's not evidence of a problem, it's evidence of complaint. Voting in a discussion is one thing, but is there actually a problem that can be proved? Where is the evidence of harm to articles? Where is the evidence of disputes in articles, especially any lengthy ones? Prove the actual problem. Stevie is the man! ^{Talk • Work} 19:40, 8 December 2016 (UTC)

  It's evidence of "highly controversial" by definition. Whether there's a problem is a matter for discussion and consensus, which is what we're doing here, and which we should have done in 2009, which is what I said. ―Mandruss ☎ 20:22, 8 December 2016 (UTC)

• It would be useful to have some statistics:

1. How many refs have been reordered by AWB?
2. How many were out of order due to the refs renumbering, and how many were placed out of order?
3. How many cases have been documented where the re-ordering was alleged to be wrong?

Secondly it would be useful to understand if people think that, as a matter of style:

Reference numbers should be ascending:

1. Always
2. Except where there is clear priority between references
3. Let them be random

Personally I subscribe to either the first or second, tending toward the first. If the distinction is sufficiently strong to warrant out-of-order numbers then there may well be a better way of handling it.

All the best: Rich Farmbrough, 23:59, 19 November 2016 (UTC).

Like I say above, AWB ref reordering skips refs separated by an HTML comment, so editors (assuming a consensus) can decide to have the footnote numbers out of order to their heart's content. Stevie is the man! ^{Talk • Work} 00:07, 20 November 2016 (UTC)

Indeed (I edit conflicted with you). I suspect that's one of the reasons that hardly anyone is bothered about this either way. All the best: Rich Farmbrough, 00:15, 20 November 2016 (UTC).

• • Statistics I have some preliminary numbers. Approximately 100 pages every day have the references put out of order, approximately 100 pages every day have order restored. I have a sample of both over an elven day period in late July.
  • Needed Analysis of how the new out-of-order refs arise (i.e., accidentally by editing other parts of the article, deliberately by re-arranging the refs, or can't tell, by adding a ref.)
  • All the best: Rich Farmbrough, 19:06, 23 November 2016 (UTC).
    
• Mandruss, I've added a link to the 2009 discussion to your post introducing the RfC. I hope that's okay. (Sorry, I should have asked you before doing it.) SarahSV ^(talk) 00:00, 20 November 2016 (UTC)
  • The only "objector" there said "Not bothered if you continue as it isn't doing any harm, except wasting effort" - this seems like consensus was established - no reason it can't change of course. All the best: Rich Farmbrough, 00:13, 20 November 2016 (UTC).
    

• OK, once again. Where, in the real world, is the guidance that the order of references indicates the order of importance? There must be a style guideline or an 'instructions to authors' page somewhere that suggests this. I've never seen one on my travels, and nobody ever comes up with one when I ask in these discussions. Somebody must be familiar with this style and can point us (well, me) at it. Mr Stephen (talk) 00:25, 20 November 2016 (UTC)
  • I believe I saw one journal which allows this. But I could be wrong. All the best: Rich Farmbrough, 00:56, 20 November 2016 (UTC).
    
  • I just put myself in the place of a reader who only has the time and inclination to look at one source of several cited. In certain cases, which are rare but still important, I want to control which source that is, and I don't know why they wouldn't choose the one physically first in the string (Westerners tend to think left-to-right). I gave a good example of such a case in the precursor discussion. I think I as a Wikipedia editor should have that discretion, regardless of whether this is some widely recognized practice or not. Most readers won't know whether such a practice exists or not, and we are writing for them, not us. ―Mandruss ☎ 01:03, 20 November 2016 (UTC)

Mandruss; you write " I don't know why they wouldn't choose the one physically first in the string". I suggest that they would not because that is not standard practice. Mr Stephen (talk) 10:46, 20 November 2016 (UTC)

As I indicated, few Wikipedia readers know anything about standard practice regarding citation order. Do you disagree with that? I've personally been a Wikipedia editor for 3.5 years and a Wikipedia user for about 7 years, and I didn't know about any such standard practice until a few days ago. Am I some aberration? ―Mandruss ☎ 11:38, 20 November 2016 (UTC)

You are not the matter at hand. Where did you get the idea that references are ordered by importance? I've never seen guidance suggesting it; somebody must have if it's not something that's been made up. Link please. Mr Stephen (talk) 13:28, 20 November 2016 (UTC)

It is in fact something that's been made up, which is not a Bad Thing. The word is innovation, and we are not bound by conventions created by external "authorities". That kind of thinking is simply rigid pedantry. If I'm wrong, link please. ―Mandruss ☎ 04:22, 22 November 2016 (UTC)

As an aside to the main discussion, my recollection is that the Bluebook system of legal citation has some rules about ordering citations by importance. I don't know if the handful of Bluebook afficiondos here at WP observe that, but it could be a possibility.

More to the point here would is the matter of having the notes/citations immediately follow the material they apply to. But some editors believe all note-links (hyperlinks) should be at the end of a sentence. In that case the citations need to be in order as used in the text, and changing that will really confuse the readers. ~ J. Johnson (JJ) (talk) 00:24, 23 November 2016 (UTC)

@Mr Stephen and Mandruss: This study found that (student) readers assume Wikipedia refs are ordered by importance / quality. Nikkimaria (talk) 00:41, 2 December 2016 (UTC)

I found that article painful to read on a tablet, I will have a look at it on a proper screen or dead tree later. It's obviously related, but seems to a very small study (30 freshers), and doesn't really address the issue under discussion here. At first sight (please correct me) it seems to be that the participants started – reluctantly – at the top of the endnotes, rather than looking for a reference to support a particular viewpoint. Mr Stephen (talk) 18:39, 2 December 2016 (UTC)

@Mandruss and Nikkimaria: I found it an insight into the cohort's inability to read an article (though maybe they simply wanted to finish the task and get a mark) but possibly not much help to us here. Mr Stephen (talk) 20:59, 3 December 2016 (UTC)

• I'm pinging the other editors who were discussing it at CITE before Mandruss opened this: @Bcharles, CBM, WhatamIdoing, GoingBatty, B, Johnuniq, Magioladitis, Izno, Kmhkmh, Headbomb, Blue Square Thing, Arthur Rubin, Peter coxhead, Stfg, Mike Christie, and Pppery: SarahSV ^(talk) 00:40, 20 November 2016 (UTC)

• I think this discussion would go better if a couple of points are sorted out at the outset, so that we are all "on the same page".

First, let's have an understanding of the basic terminology and concepts. E.g., what (I believe) we are talking about is whether instances of (e.g.) ^[1][2][3][7] are better than ^[3][1][7][2]. (That is, I believe the issue is about the ordering of consecutive strings of numbered links, not the ordering of such links through the whole article.)

Second, we should be clear that those are not "references", they are HTML links. But not links to references, as what is contained in the the mis-named <ref>...</ref> entities is not necessarily a "reference". (Yes, I know, it seems most of you use them that way, but that is certainly not the only way. Hear me out; such distinctions are necessary to avoid ambiguity, confusion, and conflict.) What the <ref>...</ref> tags create are notes (footnotes, endnotes, whatever). What you use them for (whether "references", "citations", "explanatory notes", etc.) is largely irrelevant for this discussion, but we might avoid getting stuck in conceptual dead-ends if we use the most general term, "notes". And "note-links" (or "footnote-links") for the specific element of concern here.

Third, this RfC would be better formulated as whether having strings of note-links in ascending order is merely preferred, or should be mandatory. If that is resolved one way or the other then appropriate measures can be considered, such as adjusting AWB.

Fourth, "random" is not an alternative to "ascending". Notes, and the links to them, are numbered in the order the s/w finds them. Note-links come out of order where a named ref (the <ref=name ...> entity) has been used to point to another instance. (That is, to "re-use a reference".) Don't use named refs, and I believe the ordering problem goes away.

Which then leads us back to other potential problems that underlie the issue considered here. ~ J. Johnson (JJ) (talk) 02:54, 20 November 2016 (UTC)

J. Johnson, yes, ref name is what causes this problem. But if we don't use it, AWB editors add it. SarahSV ^(talk) 13:13, 20 November 2016 (UTC)

SV: I think you are referring to this automated merging of notes (<ref>s) deemed to be duplicates, replacing one with a named ref. I think that is a problem that needs to be dealt with. (I often modify my notes so they are not duplicates, to avoid this merging.) But behind that is the challenge of "re-using" entire biblographical entries ("full citations"). In the end I think it comes down to: how do you feel about the use of "short citations"? ~ J. Johnson (JJ) (talk) 23:04, 20 November 2016 (UTC)

• Pinging others who have discussed this: @Beyond My Ken, Ryan Norton, Beetstra, and Afterbrunel: SarahSV ^(talk) 13:10, 20 November 2016 (UTC)

• I cannot find the comment where it is most relevant, but professional journals do not reuse references/notes. They almost always use Ibid. and Loc. cit. (Discussion of Op. cit., which I wrote first, is off topic.) Discussion of what professional journals do is only relevant for the few professional journals which reuse notes.

Compromise proposal: AWB bot ref reordering

CURRENT CONSENSUS IS THAT THIS IS NOT AN APPROPRIATE PROPOSAL. PLEASE IGNORE. ―Mandruss ☎ 13:23, 20 November 2016 (UTC)

I think the simple solution here is to add something like {{AWB-no-ref-reorder}} in the reference section that'll tell AWB to leave ref order alone. This way the minority can be happy, and the majority can continue cleaning up. Headbomb {talk / contribs / physics / books} 10:49, 20 November 2016 (UTC)

PROPOSAL:

@Headbomb: Or, a new way to protect only a single ref sequence without requiring the use of list-defined references, as I've explored in the precursor discussion. - {{reforder |refs=<ref>...</ref><ref>...</ref><ref>...</ref><ref>...</ref>}} - I personally have no objection to that, once AWB gains consensus to reorder in the first place, which should have been done in 2009. First things first. I like this. ―Mandruss ☎ 11:44, 20 November 2016 (UTC)

I'd be entirely fine with that as a supplement too, but tagging the entire article is the simplest solution. Headbomb {talk / contribs / physics / books} 21:25, 20 November 2016 (UTC)

• Comment Mandruss, do you think this could feasibly be done only for articles created in the last two or three months by non-extended-confirmed editors? It seems from the discussion above that many experienced editors would rather have it at their own discretion, although IMHO it looks a little weird and unprofessional to have the citations be ordered [14][2][5][3] for no reason. Jc86035 (talk) Use {{re|Jc86035}}
  to reply to me 11:58, 20 November 2016 (UTC)
  • Sorry, I don't understand the question.
    In hindsight, maybe I should have framed this RfC as a proposal for that compromise. I am a stickler for process, for doing things the right way, and that may have affected my thinking in an unconstructive way. If so, I apologize, and the question for me is whether it's too late to change directions in this RfC. The question for others may be whether that's the best approach. ―Mandruss ☎ 12:03, 20 November 2016 (UTC)
    • • Mandruss, you say you're a stickler for process, so please let this run. The broader question is whether editors should be making context-sensitive edits with AWB. Telling us we have to use special templates or html to stop it misses that point. SarahSV ^(talk) 13:00, 20 November 2016 (UTC)
    • @Mandruss: I guess if it's possible (if it is, it would slow down AWB just a little bit because of the API querying), then there's not really anything stopping you from adding a third option. The extended-confirmed part might be unnecessary if it's for pages created in the last month. Jc86035 (talk) Use {{re|Jc86035}}
      to reply to me 12:07, 20 November 2016 (UTC)
      • @Jc86035: Considering that the {{R}} template is already being used to protect sequences from AWB reordering, I don't see why using a Reforder template instead would slow down AWB. As far as I know the protection works because AWB ignores cites inside a template transclusion. All I'm doing is eliminating LDR from the equation. I still don't understand what you're saying about extended-confirmed and pages created in the last month. ―Mandruss ☎ 12:13, 20 November 2016 (UTC)
        • @Mandruss: Sorry for the confusion. I just realised I put this in the wrong subsection; my idea was to limit reordering of references to pages created recently by inexperienced editors, regardless of whether the blocking template was used. Jc86035 (talk) Use {{re|Jc86035}}
          to reply to me 12:33, 20 November 2016 (UTC)
        • Added the third option, Reforder. ―Mandruss ☎ 12:34, 20 November 2016 (UTC)
• How is one making a reasoned choice to order back to back refs un-sequentially expected to know about the proposed "{{AWB|no-reorder}}" template? Though I fail to see a majority in support of reordering references in this manner at this time, and would oppose this template as such; if it were implemented it should be required in the edit summary of bots reordering back to back references into ascending number sequence, so users would know how to prevent it from happening again.— Godsy ^(TALK_CONT) 12:47, 20 November 2016 (UTC)
  • Sorry, I included the {{AWB|no-reorder}} comment only for context, the proposal in this subsection is for Reforder. I have attempted to clarify that by adding PROPOSAL above. ―Mandruss ☎ 13:03, 20 November 2016 (UTC)

• Mandruss, please don't change the question mid-RfC. The compromise is another version of yes. First you were no in the discussion at CITE, then yes, then no again, then you opened a parallel discussion here, and no, and now yes again (sort of) and changing the question. Please just let it run. SarahSV ^(talk) 12:51, 20 November 2016 (UTC)

  Let it never be said that I am inflexible! :) Your reasoning is like accusing a politician of being a "waffler" because they changed their mind. Sorry I failed to get it right the first time, or the second (or the third? I don't know, I'm not counting). RfCs add options all the time, because it's difficult to predict what options will be necessary in advance. Wikipedia decision-making is messy. As for "the compromise is another version of yes", I think that's a matter of perspective, and I disagree. ―Mandruss ☎ 12:56, 20 November 2016 (UTC)

  It's another workaround, another version of yes. SarahSV ^(talk) 13:05, 20 November 2016 (UTC)

  @SlimVirgin: I would like to voice my strong objection to your removal of the third option, but I am not going to edit war this with you. You are now being disruptive and exhibiting WP:OWN of this process in my view. Editors who agree with you may !vote No. ―Mandruss ☎ 13:08, 20 November 2016 (UTC)

  @Mandruss:, I'm sorry, but you've caused a lot of extra work here. We had an ongoing discussion at CITE. You flipflopped back and forth there, then opened this before that had ended, without any warning. Now you're doing the same here, and trying to change the question. What you're calling a third option is really option yes(b). SarahSV ^(talk) 13:19, 20 November 2016 (UTC)

  Deferring to this two-to-one opinion, I have retracted the proposal for now by adding a comment at the top of this subsection. Thank you. Sorry for causing a lot of extra work by trying to help move this issue to a resolution, after seven years. In hindsight I should have ignored your first ping at CITE and stayed the hell out of it. ―Mandruss ☎ 13:23, 20 November 2016 (UTC)

  Thank you for fixing it. SarahSV ^(talk) 13:52, 20 November 2016 (UTC)

I've restored that 'proposal'. It's a perfectly reasonable option, and one that's in my opinion ideal. Headbomb {talk / contribs / physics / books} 21:37, 20 November 2016 (UTC)

• @Mandruss: Your third option is, really, just "yes", because it was implicit that the minority of editors who wanted deliberately to put them out of order would have to do something like that anyway. I still think it'd be best if we only reordered the refs for pages recently created by non-extended-confirmed editors (I doubt most newcomers would find out that they're supposed to do that). Would you add that to the RfC? Jc86035 (talk) Use {{re|Jc86035}}
  to reply to me 13:11, 20 November 2016 (UTC)

  Fine. I added a third option because you said, "there's not really anything stopping you from adding a third option", later realizing that you were in the wrong place. Sheesh. You guys sort it out. ―Mandruss ☎ 13:17, 20 November 2016 (UTC)

  Nothing needs to be sorted out. Let it run, then a closer will summarize the consensus. SarahSV ^(talk) 13:20, 20 November 2016 (UTC)

• I suggest to the closer that there is no consensus here, unless it be that: everything is confused, the question is poorly put, we haven't worked out what the problem is, and – there is no consensus. ~ J. Johnson (JJ) (talk) 23:21, 20 November 2016 (UTC)

  Just 26 hours after the RfC was created seems a little soon to be drawing conclusions about the level of consensus, don't you think? Why do you think the closer needs the advice of commenters on how to do their job anyway? --Stfg (talk) 00:44, 21 November 2016 (UTC)

  Yeah we should have had a separate pre-RfC RfC to decide what question to ask in this one, as I'm certain there is wide disagreement about how to frame this issue. Or, we can all learn how to give a little for the sake of moving on, and that none of this is all that fucking important in the end. This has become obsessive on the parts of many. High intelligence is counterproductive in the absence of wisdom. ―Mandruss ☎ 08:43, 21 November 2016 (UTC)

I don't know that a pre-RfC RfC is needed, but as a lot of these kinds of questions arise where various editors are running on different interpretations, I think it is generally good to work out a mutually acceptable formulation of the question beforehand. Of course, to get the question right you pretty much have to be half-way to the answer. ~ J. Johnson (JJ) (talk) 00:24, 22 November 2016 (UTC)

Based on what I know about the history of this issue, the personalities involved, and Wikipedia culture in general, I strongly suspect any attempts to "work out a mutually acceptable formulation of the question beforehand", outside RfC, would have been fruitless. ―Mandruss ☎ 04:30, 22 November 2016 (UTC)

The first time you tried to walk it probably didn't work so well, either. I agree that success in these matters is difficult, rare, and even fleeting. But I hope that eventually we'll get the hang of it, and the rock will crack. ~ J. Johnson (JJ) (talk) 02:29, 24 November 2016 (UTC)

It is not appropriate to need to add templates to articles to prevent bots from making changes; bots are meant to avoid inappropriate changes on their own. The characterization as "majority"/"minority" reveals a kind of bias, as does the description of rearranging references as "cleaning up". Remember there is no guidance at all that reference numbers should be in increasing order; that idea is something that a small number of AWB operators invented. The solution is for AWB to back out that change until WP:CITE actually requires a particular ordering of references. — Carl (CBM · talk) 13:47, 21 November 2016 (UTC)

Reorder end-notes, not the references

If it is deemed desirable for references to refer to end-notes in increasing order, then the MediaWiki software should assign numbers to the end-notes accordingly, rather than simply in sequential order of first reference. It can't resolve all cases (if a pair of end-notes are referred to in opposite order in different places in the main text, for example), but I imagine most cases would be dealt with. Reordering the references is just a work-around for the underlying issue. isaacl (talk) 20:52, 20 November 2016 (UTC)

It seems to me you have an idea of possible interest, but I can't quite make it out because of various ambiguities. E.g., what exactly do mean by "references: the numbered-links (e.g.: ^[1]) in the text? The actual end-notes the link goes to? Or the full bibliographic citation of the source that is frequently the content of the note? And are you assuming the bibliographic citations themselves are scattered through the text, or gathered in a "References" section? Are you considering the use of LDR? ~ J. Johnson (JJ) (talk) 23:15, 20 November 2016 (UTC)

To clarify: if it is desirable for a sequence of adjacent superscripts to be in increasing numerical order, the numbering of the corresponding end-notes can be selected appropriately to maximize this occurring. (*) It doesn't matter where the contents of the end-notes are defined; the software can reorder the final output as desired.

(*) As noted, there are situations where preserving increasing numerical order would require re-ordering or relocating one or more superscripts. For example, an article using a couple of authoritative sources throughout may have occasion to refer to the sources in opposite order. isaacl (talk) 05:48, 21 November 2016 (UTC)

I think this would likely cause even more confusion. It is not true that every style makes the first endnote be number 1 (e.g. endnotes might be sorted by author name, so the first endnote to appear in the text could be number 14). But many styles do make endnotes sorted by number, so the first to appear is 1, etc., and more people would likely be confused if this did not happen.

The underlying cause of the issue on Wikipedia is that our reference notes are not purely endnotes or footnotes (in which case no number would be repeated) but they are also not purely numerical pointers to a list of references (because references lists would not include comments like endnotes and footnotes do). So our reference numbers are somewhat ambiguous, leading different people to interpret them differently. Which is fine, but changing the system more would be confusing. — Carl (CBM · talk) 13:44, 21 November 2016 (UTC)

Strictly speaking, no "style" (or method) of citation "makes" the notes (end-notes) be "sorted by number"; that's just the order the wikimedia s/w finds them in the text. If an article's first note-link is ^[14] it is not because the end-notes were sorted by author name (presumably using LDR); it is because that link results from a slave named-ref (e.g.: <ref name=xxxx \>) that points to the fourteenth actual note.

The underlying cause is not that "reference notes are not purely endnotes or footnotes", but that named refs are being used to make a note-link appear in more than one place. Don't use named-refs, and this problem won't happen. It is not so much that your so-called "reference numbers" are ambiguous, but that a list of notes referenced by number is confused with a list of sources referenced by an editor.

Which is why I oppose this bot-reordering: changing a basically okay system (aside from named-refs!) to conform to ambiguous and confused understandings of what it should do makes the system more confusing, and the editors more confused. ~ J. Johnson (JJ) (talk) 00:30, 22 November 2016 (UTC)

With hypertext links, I don't think readers notice or care if the first superscript is a "1" or "14". In theory, the MediaWiki software could turn named references into a list of references and generate individual end notes pointing to the same reference, instead of reusing end notes. I suspect it may be challenging to gain consensus support for this approach, though. isaacl (talk) 01:39, 22 November 2016 (UTC)

I think you have missed the main point: some editors care. Also, you seem confused as to what constitues a "list of references" and how that differs from "individual end notes pointing to the same reference". (Particularly: how does re-use of an end-note differ from re-use of a reference?) Depending on how you understand this, I would say that the mediawiki software already does what is necessary. It's just a matter of getting editors to use the tools available, rather than inventing new tools that don't get to the heart of the problem. Which is, indeed, challenging. ~ J. Johnson (JJ) (talk) 00:39, 23 November 2016 (UTC)

There was a suggestion that named references not be used in order to avoid reusing superscript numbers. The MediaWiki software can still avoid reusing superscript numbers with named references: the resulting end notes can either replicate the citation, or point to a citation list that is automatically generated from the named references. isaacl (talk) 03:14, 23 November 2016 (UTC)

Not using named references is not a viable option in long articles with hundreds of refs, some used in 10 or 15 different places in the article. Both keeping ref list bloat down, and managing corrections or refinements to refs mean using named refs is essential. Defining named refs in the reflist section is one way to manage them, but that means new editors will need to learn that system. Bcharles (talk) 10:42, 23 November 2016 (UTC)

If the MediaWiki software were enhanced to create a reference list automatically out of the named references (or even automatically grouping identical un-named references), as I suggested, then editors would no longer need to do it manually. isaacl (talk) 02:05, 24 November 2016 (UTC)

You are probably thinking that named-refs are your only option, but that is incorrect. What you really want is probably 1) to cite ("re-use") your sources multiple times, but 2) without replicating the "reference" (the end-note containing the full citation). You do not need named-refs to do this. Another option (and vastly superior, I think) is short citations (or "short cites"). The full citations get collected on any convenient area (or not), where they can be put in any order you want; in your notes (in the <ref>...</ref> tags) are the short cites (like "Smith, 2004", as created by the {{sfn}} or {{harv}} templates) which link to the full citations. For an example see Hockey stick controversy#Notes. (The big red error messages are where someone hasn't figured out how to use named-refs properly.)

Isaacl: we already have bots "automatically grouping [merging] identical un-named references." That only creates more problems. ~ J. Johnson (JJ) (talk) 02:34, 24 November 2016 (UTC)

All I'm saying is the software could automatically gather up the citations and produce appropriate end notes in a manner such that any consecutive associated superscripts will be in increasing order. In this way, editors will still retain the ability to control the order of the hyperlinks in the superscripts. isaacl (talk) 03:59, 24 November 2016 (UTC)

And what I am saying is that we already have that. Where it breaks is when named-refs are used. (Don't use named-refs, and this entire RfC is moot.) But this is perhaps unclear to you because you are not entirely clear on the difference, and the significance of the difference, between notes and citations, and their use. ~ J. Johnson (JJ) (talk) 19:39, 24 November 2016 (UTC)

My apologies for failing to communicate clearly. I understand there are ways for editors to manually create citation lists and refer to them in end notes. I think that it would be nice if the software could take of the grunt work and automatically create a citation list. With this approach, it would be irrelevant if a named reference were used to refer to a citation. isaacl (talk) 03:12, 25 November 2016 (UTC)

I wish I had time to write some code that would remove named-refs, and even extract all full citations from the notes and put them into a list. Until then, the fix to the problems created by use of named-refs is not more s/w functionality to paper over the problems of the last fix, but to just not use named-refs to start with. As to the "grunt work" of creating citations (manually?), it is not any more work to put them in a special section than into <ref> tags. (Unless VE adamantly insists on doing things its way.) Use of named-refs provides no savings of time or effort over the use of short cites; the only difficulty is that having already learned how to use a poor method of handing citations editors balk at investing any effort in learning a better method. ~ J. Johnson (JJ) (talk) 22:50, 25 November 2016 (UTC)

• I think this suggestion shows that the original statement of the RFC was not sufficiently clear. All the best: Rich Farmbrough, 19:45, 4 December 2016 (UTC).
  

Clarification please

I thought when I came here that this was a discussion about should AWB reorganise grouped cite links at the end of a sentence or paragraph into order according to the number assigned to the cite when first placed in the article. From comments above, I am unsure if that is what people are discussing, as it seems people are discussing the order of cites in the article as a whole, or how they are presented in the Ref section at the end of the article.

My understanding of how the system works. See if others agree with this, so we are discussing the same thing.

We use several sources to write an article, and we cite the sources at the point in the article where the information is used with an inline marker. The software identifies the cite with either a number or a letter or both and links the reader to information on the cite which is collected and placed in a section at the bottom of the article. The cite information at the bottom of the article has the same number or letter as was used in the inline marker. The numbers are purely for identification, and do not relate to sequence, as some cites are used in more than one place, so the number marker may appear several times. See Covent Garden. In the reference section the first source listed is Christopher Hibbert; Ben Weinreb (2008). The London Encyclopaedia. Pan Macmillan. pp. 213–214. This is listed first as it is the first used in the article. It has been given the cite number 1. Because that source is used three times, the places it is used are identified as a, b, c. If anyone clicks on 1 a, they will be taken to the first time the cite is used (in the opening sentence). If anyone clicks on 1 b they will be taken to the second time the cite is used (in the first sentence of the geography section, where it comes after cite 25 used in the previous section and before cite 26 which is used in the following sentence. Numerically it is out of sequence here, but that doesn't matter because the number is purely an identifier, and is not meant to relate to any number sequence). If anyone clicks on 1 c they will find the third use which is at the end of the Covent Garden square section, where it comes between cite 59 which has been used in the sentence before and the sentence after. So that cite appears three times, and two of those appearances are out of number sequence. But no editor and no bot would attempt to change the number of the cite or its position. So, no worries.

In some cases, though, we may use several cites in one sentence, and some of those cites may have been previously used, and so their numbers are not in numerical order. Such a case is in The Kinks article where we have this sentence: "The group's fourth single, "All Day and All of the Night", another Ray Davies hard rock tune, was released three weeks later, reaching number two in the United Kingdom, and number seven in the United States.[4][33][36]". Now, note that there are three numbered cites, and they are not consecutive numbers, which means at least two of the sources have been used elsewhere in the article. The sentence contains three pieces of information that have been cited: the single was the group's fourth, it reached number 2 in the UK charts, and it reached number 7 in the USA charts. According to WP:CITEFOOT instead of putting a cite next to each piece of information, we should group them at the end "so long as it's clear which source supports which part of the text." The order of the cites was originally 36, 4, 33 - 36 citing 4th single, 4 citing No 2 in the UK, and 33 citing 7 in USA, in the same order as those statements are given in the sentence. This AWB edit changed the order so the cite numbers are in order. So someone checking that the single was indeed the band's fourth would assume that would be the first cite listed, and would click on cite number 4, which tells us that the single reached 2 in the UK. That person may assume that we haven't cited that it was the band's fourth single, and may either go look for one to place in the article, or may tag that piece of information as uncited.

The numbers are to identify the cites, they are not to indicate any kind of sequencing. That's my understanding. SilkTork ^{✔Tea time} 15:01, 21 November 2016 (UTC)

Except this example violates WP:INTEGRITY. CITEFOOT doesn't say anything that the preferred place to put footnotes at the end of a sentence, but at the end of the phrase that the ref supports, as to support text-source integrity. The three cites given for the Kinks example should be placed throughout the sentence instead of grouped at the end. --MASEM (t) 15:09, 21 November 2016 (UTC)

The Kink's sentence currently violates WP:INTEGRITY because the cites were rearranged into number rather than text order by an AWB edit. As regards the wording of WP:CITEFOOT, in full it says "The citation should be added close to the material it supports, offering text–source integrity. If a word or phrase is particularly contentious, an inline citation may be added next to that word or phrase within the sentence, but it is usually sufficient to add the citation to the end of the clause, sentence, or paragraph, so long as it's clear which source supports which part of the text." My understanding is that if the information is likely to be challenged, put the cite next to the word or phrase, but if it's not contentious the cite can be placed at the end of the sentence. We generally don't have three cites for the same piece of information, so when there are three cites at the end of a sentence they will generally be supporting three different pieces of information - and they will generally be placed in the order in which the information appears in the sentence. Perhaps the wording of WP:CITEFOOT needs to be tightened to avoid misunderstanding? SilkTork ^{✔Tea time} 17:18, 21 November 2016 (UTC)

The keyword is "clause", which means, to me, where you have a comma is a very valid place to drop in a reference to support the clause prior to the comma. And I have done cases where I've done on individual works such as "Smith[1] and Jones[2] both agree that the sky is blue." so that I'm assured that reference won't move than rather playing with reference order issues.

At least to me, when you do multiple refs at the same point, the weight of all those refs relative to the text in front of them and to each should be equal that reference re-ordering (moving refs around at that point to have increasing ref numbing) should not affect the understanding of the article in anyway. If the reference order is important, you've structured something wrong, and it is nearly always possible to rewrite a sentence to have refs appropriate distributed throughout so that you eliminate the reordering problem. --MASEM (t) 00:53, 22 November 2016 (UTC)

You're probably right that it is nearly always possible to do this. To me, and perhaps to some other editors !voting "no" above, the issue is not that there might not be some way to finesse the issue; it's that bots shouldn't be doing work that requires me to finesse an issue. I wouldn't mind a human reordering the refs; I can talk to a human and discuss the issue. Having to format refs in a certain way to keep bots off is like having to put up a deer fence around the roses. I'd rather have deer that don't eat roses. Mike Christie (talk - contribs - library) 01:25, 22 November 2016 (UTC)

"[W]e encourage editors to place all the cites at the end of the sentence" - I don't think that is true. I would only encourage it if the cites broadly supported the sentence as a whole. Where I am relying on a specific cite for a specific fact I will cite at the phrase or word level. For example "Aethelwulf's kingdom was claimed by Aethelhnoth,^[1] Aelfric and Cynewulf.^[2]" Here it is clear that any reader specifically interested in Aethelhnoth's claim should look at ref. 1, while ref 2 will cover at least the two other claimants.

I read the guidance you quote as encouraging precisely the opposite behaviour, if placing the references together makes it unclear what they support, then they are best placed at the end a smaller semantic unit.

Moreover your "logical and natural" system makes absolutely no sense. For example:

"Aethelwulf's kingdom was claimed by Aethelhnoth, Aelfric and Cynewulf.^[1][2][3]"

could mean all three cites apply to the whole sentence, [1] applies to Aethelhnoth, [2] to Aelfric and [3] to Cynewulf, or that [1] demonstrates it was a kingdom, [2] that Aethelhnoth and Aelfric claim it, and [3] that Cynewulf also claimed it, or about thirty other possible interpretations.

All the best: Rich Farmbrough, 20:11, 21 November 2016 (UTC).

I think what you and Masem say makes sense. My understanding from the wording of WP:CITEFOOT and other discussions on this matter, is that the preference has been to reduce as much as possible the number of times that a cite marker is used in a sentence because it inhibits reading flow. We are encouraged to remove cite markers from the lead (even though the lead is the most read part of an article) because any information in there will be cited somewhere in the body of the article. I personally like having useful access to the sources. I dislike it when the double citation method is used when the cite marker directs me to a short citation rather than the full one, and then I have to go to the book details separately which don't tell me the page number, and now I can't go back automatically to the short cite with the page number or to my place in the article without rolling back my browser. I don't know why folks do that - it makes work for me when checking details, and simply wastes time. Anyway. Rant over. Perhaps what is needed is clearer wording in WP:CITEFOOT so we are not having discussions over its interpretation. I particularly like your Aethelwulf's kingdom example, and something like that in WP:CITEFOOT would clear up any potential confusion. That said, I still don't see why we should have a bot changing the order of the cite markers simply because they use numbers - if they used symbols we wouldn't even be having this discussion. Well, they do use symbols, it's just that the symbols are numbers which can be arranged in a sequence. SilkTork ^{✔Tea time} 00:14, 26 November 2016 (UTC)

I've been pinged in many reviews for footnotes out of numeric order. I thought there was something in the MOS about this. Can anyone confirm or refute this? I agree with Rich in that in SilkTork's example I would have taken advantage of the commas to insert the footnotes in mid-sentence, but this is not always possible. And Rich is quite right; without checking the references, you cannot know what parts of the sentence they support. Hawkeye7 (talk) 20:49, 21 November 2016 (UTC)

I agree that it may not always be possible. But it seems to me that if the citations order is so important, we should at least be able to agree on what the order in a specific case should be. Silk Torq claims that it is in semantic order. Others claim that it is in "importance order". If either were the case then every consecutive use of cites following the other model, and every use which follows neither may very well be "wrong". (Or perhaps we should use ^[2][1] for importance ordered ^[4][3] for semantically distributed and ^[6][5] for the fortuitous occasions when both are valid?)

Again, if they are that important, the citations could well be referred to in the text, thus:

Norden ^[1] says the name is derived from a wood called Hitch, but Ekwall believes this is a conflation Wychwood, for which the Old English is Hwiccawudu.^[2]

All the best: Rich Farmbrough, 21:02, 21 November 2016 (UTC).

It's trickier than that. You get sentences like:

John Norden says the name is derived from a wood called Hitch.^[1][2]

Where the first reference is about the river, and the second merely supplies his first name. Hawkeye7 (talk) 11:25, 22 November 2016 (UTC)

And again if that is sufficiently important, it can be set up accordingly.

Norden (first name John^[1]) says....

All the best: Rich Farmbrough, 00:05, 24 November 2016 (UTC).