Wikipedia talk:Date formatting and linking poll/Archive 4

Raw Results

I believe these may be the final tallies:

Date autoformatting

Sup: 209 : 40.1%

Opp: 287 : 55.1%

Neu: 25 : 4.8%

TOT: 521

Day-month linking

Op1: 256 : 79.0%

Op2: 18 : 5.6%

Op3: 8 : 2.5%

Op4: 42 : 13.0%

TOT: 324

Year linking

Op1: 208 : 71.2%

Op2: 41 : 14.0%

Op3: 6 : 2.1%

Op4: 37 : 12.7%

TOT: 292

--RexxS (talk) 00:33, 14 April 2009 (UTC)

Partly implemented

I've partly implemented the results of the poll. Option 1 for month-day and year linking had by far the greatest support so I've added it into Wikipedia:Linking and Wikipedia:Manual of Style (dates and numbers)/Datestempprotectedsection. I think this should be fairly uncontroversial - the results aren't ambiguous at all. There's still the autoformatting issue and I'm thinking how to handle it. Ryan Postlethwaite^{See the mess I've created or let's have banter} 10:33, 14 April 2009 (UTC)

(edit conflict) Good. Personally, rather than duplicating the text, in WP:MOSNUM I would just write:

==== Linking and autoformatting of dates ====
Dates should not be linked purely for the purpose of autoformatting (even though in the past this was considered desirable).^[1] They should only be linked when the linked article is germane and topical to the subject, per Wikipedia:Linking#Chronological items.

So, as far as linking is concerned, the second stage of the RfC should just ask how excessive links should removed (essentially, whether to use bots to do that). As for autoformatting, since many people supporting the "general concept" appear to oppose its current implementation, question #1 should read:

Do you support the de-activation of the Dynamic Dates feature, which autoformats linked dates? If accepted, this will be implemented by adding $wgDynamicDates = false to the configuration file, and will have the effect that date links will be rendered as any other link, regardless of user preferences.

A. di M. (formerly Army1987) — Deeds, not words. 11:30, 14 April 2009 (UTC)

Now, when will we have a final decision and the injunction against date delinking be lifted? RainbowOfLight ^Talk 07:18, 15 April 2009 (UTC)

Actually, I don't agree with Ryan's analysis. Option 1's subtitle should be added; which would make the second sentence "They should only be linked when the subject of the linked article is germane and topical to the subject." I find less than 50% of the year voters have a clear vote for option 1 as written. — Arthur Rubin (talk) 01:19, 17 April 2009 (UTC)

Autoformatting

I've just been having a discussion with a sysadmin regarding autoformatting and the poll. There is no consensus to use the {{#format}} autoformatting style (consensus against it infact) and the current method of autoformatting by looking for linked dates can no longer continue because dates aren't going to be linked. We're therefore left with no other options for autoformatting at present. Would everyone agree that autoformatting is not viable at this stage in time given the result of the poll? I'm willing at this stage to file a bug to have it turned off completely (i.e. removed from preferences). Ryan Postlethwaite^{See the mess I've created or let's have banter} 11:20, 14 April 2009 (UTC)

Ryan, please implement Sapphic's proposal, which amounts to disabling DynamicDates (setting $wgUseDynamicDates = false in LocalSettings.php, not disabling date preferences entirely, as they're used for things other than DynamicDates) and barring the opponents of autoformatting that are named in the ArbCom case from interfering in any future discussions to develop a new software replacement. --169.229.149.174 (talk) 15:15, 14 April 2009 (UTC)

Interesting post above. Nnow it seems there is a distinction between dynamic dates and user preferences not widely known about, which could/should be elaborated upon. Ohconfucius (talk) 15:27, 14 April 2009 (UTC)

Date preferences apply to the "last modified" date at the bottom of pages, as well as other places. I'm not sure where the other places are, honestly, and don't particularly care. Other people brought it up below. It doesn't have anything to do with dates in articles, doesn't impact editing of articles in any way, and doesn't have anything to do with DynamicDates. —Preceding unsigned comment added by 169.229.149.174 (talk • contribs)

Be careful of wording such as "interfering in any future discussions to develop a new software replacement". Opponents of a system have a right to a seat at the table in discussions - perhaps some of their objections could be satisfied; if not, they can still work as a devil's advocate to make sure that nothing is proposed that could cause a greater burden on editors. It's always wise to get a wide variety of opinions to create the best possible specification. Otherwise, it is more likely that any proposal will get derailed quickly in the first poll. Karanacs (talk) 15:38, 14 April 2009 (UTC)

Opponents of autoformatting have no reason to participate in developing new software, though if they want to contribute constructively they're perfectly welcome (any such development discussions will occur on this site anyway) although the named parties in the ArbCom case have actively disrupted the development process in the past and there is every reason to believe they'll do so in the future. Those particular people should be banned from any discussion on date autoformatting software. Anyone who disrupts the process should be similarly warned by ArbCom, or blocked. —Preceding unsigned comment added by 169.229.149.174 (talk • contribs)

I strongly oppose removing the date format option from preferences. It is necessary to get readable dates in page histories, watchlists, edit diffs and so forth, and it was never disclosed during this poll that the absence of autoformatting in articles would also result in date preferences disappearing from the rest of the user interface. You cannot use the people who complained about the existence of article markup for dates (let alone those many who based their opposition on not wanting dates to be linked) as support for not letting us have YYYY-MM-DD dates in lists. What will be next, forcing times in watchlists to be displayed in Florida time because timezones are confusing? –Henning Makholm (talk) 11:42, 14 April 2009 (UTC)

You know, I give up. Go for it. Never mind that there's already evidence that canvassing has occurred, let's just use these (now flawed) results anyways. You have my blessing Ryan, auto formatting should never have been turned on in the first place. Clearly Wikipedia is also flawed, and we should take the entire site down and use print instead. I'm sure we'll be there soon enough if we keep removing features only possible online. —Locke Cole • t • c 11:28, 14 April 2009 (UTC)

Sorry, I should have elaborated. I'm looking into the canvassing privately - At this stage, I don't think the results would have been too different but I'm not doing anything until I know for sure how many emails were sent and who by. Ryan Postlethwaite^{See the mess I've created or let's have banter} 11:33, 14 April 2009 (UTC)

I doubt you'll honestly ever know how many were sent, it's unlikely those who canvassed will be forthcoming with exact details (and even if they are, given the secrecy used, is it really likely they'd tell you the truth; it's not to their benefit at this stage). As to recipients, again, depending on how important this is to them, it's unlikely they'd be willing to "out" the person who canvassed them. My faith in Wikipedia as a project is severely shaken by this, and I don't believe good faith efforts are at all possible in an environment where this conduct is rewarded (as you are doing here). —Locke Cole • t • c 11:53, 14 April 2009 (UTC)

You seem to be shooting the messenger here. I think it is the community you should 'blame' for not delivering what you wanted. Ohconfucius (talk) 12:03, 14 April 2009 (UTC)

Not at all. The community that showed up could be made up of a large number of people contacted by Tony, Lightmouse or even you. I was interested in what a real sampling of the community thought on these issues, not what a group of people selected and contacted by Tony/Lightmouse thought. —Locke Cole • t • c 12:06, 14 April 2009 (UTC)

I've written a short essay on this here. —Locke Cole • t • c 12:06, 14 April 2009 (UTC)

I'm glad common sense has finally prevailed, and I am sure the community will heave a huge sigh of relief that this is not the start of more RfCs. As to Locke's reply, it's a shame it has come to this. I would say it has been clear to me since before December the consensus view about not wanting a technical solution to a problem which few believe exists, and which benefits few. Nevertheless, I would salute Locke's tenacity. Ohconfucius (talk) 11:45, 14 April 2009 (UTC)

To Locke Cole. If you were truly interested what a real sampling of the community thought on these issues, neither yourself nor Tony should have been involved. A random sample of all registered editors, as well as of regular IP editors should have been invited. That is the problem and the reason why Wikipedia is not a democracy and why polling is evil. You are not likely to get a true sample of the community ever, as you are most likely to get a sample of editors who are more than averagely interested in the topic in hand. In this case a rather technical solution, so you were more likely to get an overrepresentation of editors interested in technical solutions and thus the likelihood you would get a "support" in your sample while the idea was not supported in the larger community at all. (Trust me, I know what I am talking about, I am statistics and methods teacher at university)

If you would use this as a vote, we are no longer talking (random) sampling but a true representation by the Wiki population (community). In that case the turnout of this poll (approx 500 on 7 Million registered editors) is lower than 1 in 10,000. Any democratic election with this kind of turnout would be called ridiculous. Hence the vote can at best be an indication of the feelings of the community and should not be seen as a vote of the community.

I also disagree with Ryan Postlethwaites original assessment that there is consensus against adoption as there is a rather modest majority against, which is something else entirely than consensus.

So in that light, the (supposed) canvassing is only one problem with the sampling. Arnoutf (talk) 12:16, 14 April 2009 (UTC)

I agree that neither Tony or myself (or anyone else closely involved with the subject) should have been allowed to be involved, hence why I suggested this long before the RFC opened. Unfortunately my request was not heeded, and here we are... —Locke Cole • t • c 12:25, 14 April 2009 (UTC)

I think getting in a representative and unbiased sample of all Wikipedia editors would be the most difficult issue here. But some of the problems might have been prevented if your suggestions had been followed. There is nothing in the Wikipedia project however that could have made that so. Arnoutf (talk) 12:52, 14 April 2009 (UTC)

The simple truth is that people who don't care about the issue don't vote. Therefore, we basically miss the opinions of the most moderate of our editors, and get those who feel more strongly about the issue and tend to polarise a discussion. In the case of a poll, of course, polarisation is encouraged; look at how few the neutral votes are. It's inevitable—the sample will never be representative unless the subject is so important as to convince a large chunk of Wikipedians to participate. I doubt this is such a subject. Waltham, The Duke of 14:49, 14 April 2009 (UTC)

Only 157,493 users have made any edit in the last 30 days (see Special:Statistics). And there aren't anywhere near 9,424,769 distinct persons registered to Wikipedia. There are people having used thousands of sockpuppets, people who left Wikipedia and then came back with another username (including me), etc. So the "lower than 1 in 10,000" is irrelevant. --A. di M. (formerly Army1987) — Deeds, not words. 13:26, 14 April 2009 (UTC)

• Indeed, that would put participation rate at a more respectable 1:300. Ohconfucius (talk) 13:34, 14 April 2009 (UTC)

• Bit cynical Ohconfucius to call a turnout rate of 0.3% respectable. (BTW why a 30 day limit?) Arnoutf (talk) 13:40, 14 April 2009 (UTC)

• I suggest that you read what I wrote carefully. Li'l ol' me a bit cynical? It beats your previous attempt to mislead by a factor of 30. Pray, remind me who said "there are lies, damn lies and statistics"? Ohconfucius (talk) 14:37, 14 April 2009 (UTC)

• Nobody knows for sure, but the statement is most frequenty attributed to Disraeli. But you do have a point I only took the number provided somewhere above in the discussion without checking them myself, or even looking whehter they were at all likely; not very methodologically sound, I agree; as for many things statistics lives and dies with the GIGO principle. Arnoutf (talk) 14:44, 14 April 2009 (UTC)

• Same reply I guess goes for the 'why 30 days' question - its what numbers the query generated. Possibly 90 days would be more indicative as to the "real" level of editor activity... Ohconfucius (talk) 14:51, 14 April 2009 (UTC)

• The date formatting is visible to all Wikipedia readers. Depending on which professional web measurement is used, English Wikipedia has between 50 and 150 million unique visitors each month. What samples size is needed for 100 million readers? -- SWTPC6800 (talk) 16:22, 14 April 2009 (UTC)

And only 286 viewed my article on Model Rocketry (magazine).[1] -- SWTPC6800 (talk) 16:36, 14 April 2009 (UTC)

• It is not as much the sample size as the sample composition that is the problem. In this case people with a strong opinion about the topic are overrepresented (by many magnitudes). My statement is that such a sample is likely to overrepresent supporters (as they are most likely to have been involved and most likely to know of this debate). Locke Cole states that, within the already biased towards strong opinion sample, the opponents are overrepresented because of the supposed canvassing. Both claims cannot be supported by facts as we don't know all Wikipedians. It is in my opinion very likely though that the sample is not representative of the larger community. Arnoutf (talk) 16:45, 14 April 2009 (UTC)

The overall majority against auto-formatting may not have been overwhelming enough for all to comfortably interpret as consensus, but this is only half the picture. Not all of those supporting the principle are in favour of the proposed new style, and indeed some are against it, while by definition all who are against the principle are also rejecting said new style. Therefore, the adoption of the style has been arguably defeated with a much greater majority than the principle of auto-formatting, easily large enough to be considered a consensus. In conjunction with the second and third questions, which are clearly against universal linking of dates and years, Mr Postlethwaite has correctly stated that there is only one viable course of action right now. I am not sure about special pages, logs, and statistics (I'd never thought of that until now), but on regular Wikipedia pages auto-formatting should be turned off.

Balancing different needs is an integral element of any large and complex system involving great numbers of people, and we all know Wikipedia is no exception. The debate on date auto-formatting has shown how hard this thing can sometimes be. No matter how great or small the majority of one side, some people will invariably be disappointed. Many fine colleagues have spent significant time and effort and made great arguments to support their case. I know how disappointing it is to see such efforts failing to achieve the desired effect, but my honourable colleagues can take heart at the fact that no matter what the poll's outcome, many of those arguments continue to ring true, and indeed, there is always the possibility to re-examine the matter in the future, if and when sufficiently sophisticated technical means are developed to address the objections now raised against auto-formatting in general.

I haven't participated nearly as much as I had wanted to in the long debate that preceded this poll, but that's just as well—the auto-formatting issue may have been over-analysed, and I don't think I am the only one to see the end of it with relief. It has generated much tension and drained significant resources, and although this is to be expected of such discussions, their duration should not be over-extended. Now we can put this behind us and close the cycle that started six years ago. We have seen how it is to have the feature; let us now see how it will be the other way. And with this in mind, take a breath and move on towards other endeavours. There are always vandals to fight, statements to source, and typos to correct, the familiar tasks Wikipedia's continued improvement depends on. Little will change now. Uninvolved editors may notice a few differences. Most readers will never know. Waltham, The Duke of 14:49, 14 April 2009 (UTC)

• Yes, Ryan. Your and the ArbCom’s decision is the correct one. Greg L (talk) 19:09, 14 April 2009 (UTC)

Action plan

Ryan, whatever the interpretation of the result, the better solution lies in encouraging or discouraging behaviour, rather than in software changes. First off, any software changes would have to be confined to EnWiki, not across all Wikimedia sites. Second, disabling the "Preferences" option would affect much more than just autoformatting, as it also controls how dates and times are presented in article histories, watchlists, and other lists. Finally, based on the responses, there appears to be stronger opposition to the method of autoformatting, rather than the concept of it. As such, any move now should not unduly compromise the possibility of introducing a different system should the community support it. Thoughts? --Ckatz^chat_spy 20:07, 14 April 2009 (UTC)

That's an important point that I was unaware of until today. I had a suggestion on my talk that disabling DynamicDates (setting $wgUseDynamicDates = false in LocalSettings.php) would have the affect of disabling autoformatting of articles without affecting users changing their preferences for article histories e.t.c. Would that be the case? Ryan Postlethwaite^{See the mess I've created or let's have banter} 21:52, 14 April 2009 (UTC)

Ryan, the config would appear to enable/disable the feature (per this Mediawiki page). I've no idea how or if it will affect the "formatdate" function. A question, though - is it even necessary to disable the feature? If Arbcom decides that the RfC has not shown support for the use of autoformatting at this point in time, and if the links are proposed to be stripped out anyway, there is no need to actually disable the software feature. This would allow the use of autoformatting (if desired) on personal pages, or perhaps in guidelines or other such pages that are part of the "infrastructure", rather than article space. Thoughts? --Ckatz^chat_spy 00:33, 15 April 2009 (UTC)

It doesn't affect the {{#formatdate}} parser function. Werdna's patch does make use of the same code, but they're triggered in different ways, and disabling DynamicDates would have no impact on the parser function. There are two good reasons to disable DynamicDates entirely: 1) It would immediately cause all editors to see dates the way anons do, including the inconsistency of format on some pages (which everybody has always acknowledged to be a genuine problem) and 2) It would allow links to dates to work like any normal link, without having to use a strange syntax to avoid triggering the autoformatting. --Sapphic (talk) 00:38, 15 April 2009 (UTC)

• We should just let the bot go back to delinking. The advantage of this is dates like 1941‑12‑07 can be converted into human-readable form by a bot. If we just turn off autoformatting, it would leave these abominations. That’s been the trouble all along with autoformatting: it made *pretty* results for 0.03% of our readership (Locke and a few others) like this:

(Fixed-text preview of what a U.S.-style preference produces)

---

• After the March 30, 1791 order by the French government instructing the Academy to…

• After Japanese planes attacked Pearl Harbor on December 7, 1941 the U.S. Congress on…

…but 99.97% of our readership (these are authenticated numbers) have long been looking this crap:

(Fixed-text preview of what all regular I.P. users see)

---

• After the March 30, 1791 order by the French government instructing the Academy to…

• After Japanese planes attacked Pearl Harbor on 1941-12-07 the U.S. Congress on…

Back in December, we started fixing all of this with Lightbot but Locke put a halt to it. We had it all figured out. We weren’t going to fool around providing custom content for registered editors; they could look at what everyone else sees. The community consensus in December was clear enough for this and the injunction against Lightbot needs to be lifted so it can continue its magic. Even though these proceedings have taken on an apparent importance of letting North Korea have a nuclear reactor, it’s not. Just let us get back to doing what we were doing before Locke claimed that all past RfCs were horribly flawed. The only RfC he will ever think is correct is the one that gives him what he wants. I think bot delinking is far preferable to just turning off date autoformatting. It is most beneficial to those for whom we have really been authoring Wikipedia: the rest of the world, not use privileged few.

The solution for the ArbCom is simple here:

1. Rescind the injunction on Lightbot activity,
2. Enjoin the pro-autoformatting crowd from pushing autoformatting for one year,
3. And when they do come back, have their ideas vetted by Wikipeida’s CTO before advancing their ideas to the community for consideration.

Greg L (talk) 23:48, 14 April 2009 (UTC)

But if Dynamic Dates is disabled, then everybody will see this crap:

• After the March 30, 1791 order by the French government instructing the Academy to…

• After Japanese planes attacked Pearl Harbor on 1941-12-07 the U.S. Congress on…

Many fewer people will oppose the unlinking of dates, then. As a bonus, these few dates which will be linked won't need to be coded as [[January 1|January 1]] in order for all readers to see them the same way. --A. di M. (formerly Army1987) — Deeds, not words. 00:26, 15 April 2009 (UTC)

• I thought that was the point I was trying to make, A. di M. Greg L (talk) 01:59, 15 April 2009 (UTC)

(ec) Greg: Nobody is suggesting that DynamicDates be turned off instead of delinking, they're suggesting that both things should be done (if either is.) Leaving DynamicDates turned on will require editors to jump through extra hoops to link dates that should be linked, because they'll have to write dates as either [[:25 December]] or [[25 December|25 December]] instead of the simpler [[25 December]]. That doesn't benefit anybody. DynamicDates can be turned off right away, which will let all editors see dates the way anonymous users currently do — which will help make more editors aware of the inconsistent format in many articles. Please stop fighting every single thing your "opponents" suggest, simply for the sake of fighting it — in this case, the people who have opposed your plans are actually helping to implement them in an orderly fashion, since that's better than the alternative. --Sapphic (talk) 00:28, 15 April 2009 (UTC)

• Like I often say: Lead, follow, or get out of the way. If you know how to clearly help Ryan and the rest of the ArbCom to carry out the wishes of the community, then I accede to your expertise. Lead away. Greg L (talk) 01:25, 15 April 2009 (UTC)

• P.S. As A. di M. wrote above (and the point I was trying to make), is that if DynamicDates is turned off, there could be a period of real crap on Wikipedia. Really, Lightmouse and his bots have been doing an extraordinary job keeping Wikipedia a much better place. Lightbot does so much housekeeping, it’s nearly impossible to describe it all. The guy has been dumped on a bunch throughout this ArbCom and many an editor would have said “you can take this (volunteer) job and shove it.” We are lucky he didn’t just bail. We really need his input to coordinate on how long it would take for Lightbot to go through Wikipedia and deprecate the links. We just don’t want to have DynamicDates turned off if it’s going to take a week or two for Lightbot (operated by the man) to do its thing. As I’ve mentioned before, the error rate on Lightbot, on a sample of 40 articles was 0% false positives and 0% false negatives. If I am seeing this issue correctly, Sapphic, then please coordinate with Lightmouse. Greg L (talk) 01:59, 15 April 2009 (UTC)

• Ryan: It would be nice if we didn’t have to start watching Sapphic’s and Lightmouse’s talk pages to see what is being discussed on these techno‑issues. Others, such as A. di M. have a hell of a lot better insight into the technobabble than you or I do and it will be a lot easier on all of us if we can all participate at one venue. I ask that you please formally “unbanish” Lightmouse and invite him to speak to the technical issues here on this page. Does that sound reasonable, given the task ahead? Greg L (talk) 02:09, 15 April 2009 (UTC)

Greg, A. di M.'s point was that if DynamicDates is turned off, registered users will see things the same way as anons do now. It won't affect anons (or anyone with "no preference" set) one way or the other if it's turned off. There's no "mess" created by turning it off now. ..and the error rate for Lightbot was a lot higher than 0% — there were numerous complaints of delinking things like image filenames that contained a date, etc. that were all errors. I tend to think bots should really have a 0% error rate before they're allowed.. but since I hope to be using a bot to put markup back around the dates at some later point anyway, I'd rather have that bot judged by the same (more lenient) standards that Lightbot is now. Anyway, the real issue with using a bot to delink isn't going to be mistakes, it's going to be disagreements over which dates are relevant to the article. --Sapphic (talk) 02:17, 15 April 2009 (UTC)

• Yes. Brain fart. It just looks like crap for registered editors, which is no worse than what I see all the time since I leave my date preferences turned off to ensure I see what I.P. users see.

  I didn’t say that Lightbot’s error rate was 0% across a wider population of articles; I wrote that “on a sample of 40 articles was 0% false positives and 0% false negatives.” That is a true fact. My point is that there are literally millions of dates to be addressed. A bot is the only practical way to address this. Our tests showed that it left links alone in intrinsically chronological articles. It’s not like there is going to be a complete disaster if there are some false positives; the date would be there but no longer linked. It’s not the end of the world when this happens and one can always hit the [undo] link. Still, we need to keep the number of occurrences as low as possible.

  I would suggest that we just ask Lightmouse to delink, say, a hundred articles and wait for feedback. Tweak his bot if necessary, and then try, say, a thousand articles. Once it’s error rate is as low as the technology allows or is acceptable by normal community standards, then let it rip. Greg L (talk) 02:29, 15 April 2009 (UTC)

• Someone please explain this to me: What would turning off DynamicDates do to this: September 1. Would it leave the link but not autoformat it? Would turning off DynamicDates unlink it too? My reading above is that it would be that latter one. How can that be good? Articles like 1985 contain hundreds of linked dates and those are, by agreement, supposed to stay linked. What am I missing here? Greg L (talk) 02:44, 15 April 2009 (UTC)

• As I understand it, turning off dynamic dates would stop the use of date links for autoformatting purposes. The square brackets would then function as they do on any piece of text. Ohconfucius (talk) 03:26, 15 April 2009 (UTC)

• That makes sense, Ohconfucius. In which case, like Sapphic said, we clearly need to do both: 1) Turn off DynamicDates, which handily addresses the autoformatting issue. And, 2) to implement the community’s wishes on the linking of dates (icky poo in most cases), we simply must get Lightbot back into the saddle; there are far too many links to manually shake a stick at. I think I’ve captured the logical approach here. Yes?

  As I wrote above, we don’t have to have Lightbot soar through Wikipedia like an A‑10 through a tank column; we just ask Lightmouse to delink, say, a hundred articles and wait for feedback. Tweak his bot if necessary, and then try, say, a thousand articles. Once it’s error rate is as low as the technology allows or is acceptable by normal community standards, then let it rip. Greg L (talk) 03:46, 15 April 2009 (UTC)

• I completely agree with Greg that Lightmouse has been shat on from a great height throughout this whole affair, and we really should give him a break. I am not saying that we should let Lightbot completely loose, but a limited trial on a clearly defined sample would be a good way to go. As sapphic is interested in the potential for bot action, perhaps she could help improve Lightbot's functionality and eliminate false positives/negatives - that way, a small modification of the code may be sufficient to swing the function into a different gear to apply necessary markup when the time comes. Also, Lightbot could be used on categories of articles to unify all date formats and insert {{use dmy dates}} or {{use mdy dates}} tags on each article it has passed through for future default display formats. Ohconfucius (talk) 03:36, 15 April 2009 (UTC)

• Your suggestion that Sapphic and Lightmouse could collaborate sounds interesting. Greg L (talk) 03:46, 15 April 2009 (UTC)

I just realized I'm not at all familiar with what Lightbot had been doing, and was referring to the script which was (I think) written by Lightmouse. Maybe Lightbot really was operating with a 0% error rate.. I don't know. In any event, I don't think the error rate is going to be at all significant in comparison to the disputes over relevancy. A lot of people (though still a minority) consider birth/death dates to be relevant, and so I think you're going to encounter a significant amount of opposition if those are delinked en-masse. There will probably be opposition all across the board by people who think dates are relevant.

I would suggest letting Lightbot (or any bot for that matter, if somebody else wants to write one) start delinking all linked dates (except ones that are piped and/or colon-prefixed, since somebody presumably did that on purpose and thus the linking wasn't there to trigger autoformatting) and then be very conservative in re-de-linking any dates that people re-link. In fact, I'd suggest that if somebody re-links a date that has been delinked, and they do so based on the argument that the link is relevant (and not just because they're trying to cling to the — now genuinely disabled — autoformatting) then the link should be left linked (for now.) It's just not worth the arguments. The questionable links can be revisited after the "heavy lifting" of mass delinking has been completed. (Note that this is not a call for the pro-linkers to go ahead and undo all the delinking; obvious gaming of the system should be met with administrative action.)

I'd be happy to help with any bot programming, although I'd probably be of more use in generating "work lists" from analysis of Wikipedia database (XML) dumpfiles, since that's more my area of expertise. I do know regular expressions pretty well though, so if the bot needs any debugging in that area, I could certainly help. --Sapphic (talk) 04:00, 15 April 2009 (UTC)

• • I think we would be looking to completely delink on a first pass, combined with a no-go list for Lightbot, especially where notifications/complaints have been received. Also, there was a discussion a while back on WT:MOSNUM which was adjourned to Lightmouse's talk page where we discussed how Lightbot could potentially work by crawling already tagged articles and recording in a dump file the number of inconsistent date instances within an article requiring manual treatment. These are ideas which have started, and were stopped because of the Arbcom case. Now, perhaps some of these can be resumed. Ohconfucius (talk) 04:22, 15 April 2009 (UTC)
• Hopefully, Ryan will invite Lightmouse to weigh in here now. It seems there is a lot of programming smarts that, if combined into a team effort, could benefit Wikipedia. Greg L (talk) 04:08, 15 April 2009 (UTC)
• "There will probably be opposition all across the board by people who think dates are relevant"—remember that over 100,000 pages were delinked during Nov. 2008 alone—with proportionally trivial residual issues. An updated bot-comment will help explain the process, and will (hopefully) provide people with a link supplying a deeper explanation of what is happening.  HWV258  04:27, 15 April 2009 (UTC)

I would have thought that those who specifically wanted birth/death dates linked would have voted for "Year option 2"[14%], but maybe not. Nevertheless, I still see people confusing the year of a person's birth with the wikipedia article bearing the title of that year. The actual year is relevant to the person; but the article with that title is almost invariably irrelevant for the purposes of option 1. Until we clear this piece of obfuscation out of the debate, this will remain a bone of contention. --RexxS (talk) 16:51, 15 April 2009 (UTC)

Please don't start in with the annoying rhetoric again. The "proportionally trivial residual issues" include the past five months of argument, blocks, and ArbCom case, and several polls — in which a significant portion (over 40%) of respondents indicated that they'd rather keep autoformatting in its entirety. A good real life friend of mine (that many of you also liked or at least respected, given some of the comments left on his talk page) also got so fed up with the "process" here as a result that he's most likely left the project for good. So stop trivializing the number or importance of the people who disagree with you. I'm not claiming that it's going to be some overwhelming number — but if even 1% of the active editors disagree over whether a date link is relevant or not, that's still much higher than the probable error rate of a well-written bot (which I'm assuming Lightbot is.) With regards to the relevancy of birth/death dates, the percentage is even higher (though still very much in the minority.) People are going to complain about the delinking; the issue is finding a reasonable way of dealing with that situation. Being dismissive is not the way to handle it. --Sapphic (talk) 04:40, 15 April 2009 (UTC)

• Let's not get too worked up, I believe HWV was only referring to the incidence of specific individual complaints purely on the articles delinked. Of course it's always a shame that good editors leave the project, but I do not believe the comments were to denigrate or justify the departures or label them as "trivial", as you seem to have read it.Ohconfucius (talk) 04:55, 15 April 2009 (UTC)

• I have little idea of what you're talking about, and the facts are plain for everyone to see. In regards date-delinking, everything was going exceedingly smoothly before one editor (who pretended to speak for an entire sub-community) dragged us all through this mess. It has been a draining and harrowing experience for us who have been involved in the entire journey of this debacle. But the community has vindicated our stance. If you spend the time to investigate the results of the enormous bot activity last year you will understand that there are two distinct phases at play: the process that happened before the RfCs, and the process during the RfCs. Your comments above clearly refer to the RfC debates, whereas my post refers to the pre-RfC process. In no way was I "dismissive"—quite the contrary in that I'm pointing out how the bot edit comment needs to be adapted to point out the support of overwhelming community consensus. Also, could you please not order people around with bold text—I'm surprised you haven't realised by now that it has the exact opposite effect you're after. Greg_L is quite correct: "Lead, follow, or get out of the way" (bolding tactfully removed).  HWV258  05:01, 15 April 2009 (UTC)

Reply to Sapphic (talk) 02:17, 15 April 2009 (UTC) above: Actually, it's not as simple as "if DynamicDates is turned off, registered users will see things the same way as anons do now": DynamicDates also affects anons and "no preference" users: for example, it adds a comma and a space in April 152009 or 200915 April, removes the comma from 15 April, 2009, makes the single pair-of-double-brackets 2009-04-15 into two links instead than the single 2009-04-15, and maybe something else I'm not aware of. So the short term conseguences would be crappy (April 152009), but I don't consider that a downside. If an article contains some crap like April 152009, then as soon as someone with more than about fifteen spare seconds reads it, they will fix it, and I don't think anyone will oppose that. --A. di M. (formerly Army1987) — Deeds, not words. 09:36, 15 April 2009 (UTC)

• Yes, that was one of the advantages we were talking about if DA was switched off. We exlist a potential 7 million pairs of hands to sort out the mis-formatted dates. Ohconfucius (talk) 09:56, 15 April 2009 (UTC)

The other thing that Dynamic Dates seems to do, in HTML terms, is to wrap the <a href > link in a <span class="mw-formatted-date" title= > tag. I don't know if that is used for anything else, so I'd suggest taking a short period where all the ramifications can be anticipated before the switch-off. --RexxS (talk) 16:51, 15 April 2009 (UTC)

• Clearly, we shouldn’t be turning DynamicDates off; it will instantly turn Wikipedia’s dates into a mass of ungodly ugliness. The proper way to do this is to go back to what we were doing before we got dragged here kicking and screaming into ArbCom and yet another RfC: let Lightbot unlink dates, which takes care of the overlinking and the attendant autoformatting in one step. The bot is smart enough to not unlink a single date in articles like 1985 and to de‑link the “15 April” from here in Politics of Turkmenistan.

  As I wrote above, we don’t have to have Lightbot soar through Wikipedia like an A‑10 through a tank column; we just ask Lightmouse to have Lightbot delink, say, a hundred articles and wait for feedback. He can then do any necessary tweaks and then try, say, a thousand articles. Once it’s error rate is as low as the technology allows or is acceptable by normal community standards, then let it rip. Greg L (talk) 16:56, 15 April 2009 (UTC)

Greg L, i'm struggling to understand what's brought you back to "we shouldn’t be turning Dynamic Dates off". what seems clearly sensible is letting Lightbot get back to work and turning off Dynamic Dates (naturally as RexxS notes all the ramifications of turning it off need to be considered first). if delinking is done without turning off Dynamic Dates, the dates that do remain linked (in Lightbot's "no-go zone", and dates re-linked on purpose after the bot run) will be autoformatted, and why should they be, when there's no consensus for it?? meanwhile, if the "ungodly ugliness" you anticipate becomes visible only when Lightbot removes the links, i fear the long-suffering Lightmouse will be fielding miles of unfounded complaints that his bot is causing it. if the way dates are entered is really "ungodly ugly" i say let that be visible before Lightbot does its rounds. as you've often noted, we need to see it in order to fix it. Sssoul (talk) 17:51, 15 April 2009 (UTC)

• Sssoul, please read A. di M.’s 09:36, 15 April post, above. If what he says is true, DynamicDates also affects anons and “no preference” users (in addition to “Registered & Preferenced” editors). DynamicDates does more than just flip things around; it adds commas and bifurcates the month-day and the year into two separate links. So until Lightbot can catch up, this code: [[2009-04-15]], would be seen by everyone like this April 152009. Logged‑in, Preferences‑set editors would no longer see either April 15, 2009 or 15 April 2009 they used to see. Even I.P. users (99.97% of our readership) would no longer see what we used to consider as ugly: 2009‑04‑15; they too (everyone) would see April 152009. This is clearly no good at all.

  The proper thing to do is just (carefully) go back to what we were doing before: delink dates. The community is quite clear that unless one is in an intrinsically chronological article, such as April or 1985, which may contain lots of linked dates, date are to be very rarely linked. WP:MOSNUM is now clear on this issue.

  By leaving DynamicDates turned on, we won’t be upsetting the apple cart in articles like April or 1985. As I wrote above, we don’t have to have Lightbot soar through Wikipedia like an A‑10 through a tank column; we just ask Lightmouse to have Lightbot delink, say, a hundred articles and wait for feedback. He can then do any necessary tweaks and then try, say, a thousand articles. Once it’s error rate is as low as the technology allows or is acceptable by normal community standards, then Lightmouse can quicken the pace.

  What Lightbot was doing was somewhat controversial before, because MOSNUM wasn’t definitive and clear on the issue. Now what is and is not to be linked is a settled matter. Moreover, in a test of 40 articles, Lightbot demonstrated that it knew to leave articles like 1985 completely alone, yet it caught every single instance of things like “15 April” here in Politics of Turkmenistan. Lightbot may not yet be perfect, but Lightmouse has always been responsive to community feedback and has quickly tweaked his bot. So it’s just a matter of going slow at first, ensuring Lightbot is as compliant with MOSNUM as Lightmouse can make it, and letting technology go to work. There are far, far too many linked dates to address without using a bot. Greg L (talk) 18:41, 15 April 2009 (UTC)

(edit conflict) thanks Greg L, but i feel like you've misunderstood what i wrote. i did read A di M's very interesting post about the "extra" things DD does; and i am thoroughly in favour of bot assistance with delinking. what i'm talking about is turning DD off in addition to getting Lightbot back on the job.

maybe this question will clarify what tree i'm barking up: even if DD is left "on", when Lightbot removes the date links, that will eliminate the DD-related "extras" that A di M pointed out. will Lightbot correct things like April 152009 at the same time as it removes the links?

if it will make that kind of correction, it's easier to see the point in leaving DD on until Lightbot has done its thing. Sssoul (talk) 19:22, 15 April 2009 (UTC)

Removing the link markup while retaining an appropriate date format is a goal while it's delinking. Someone can double check with Lightmouse, but I'm almost positive that it makes these corrections as it delinks. To prevent ugly dates in articles site-wide, I think it would be best to let Lightbot delink and clean first, before turning off DD. —Ost (talk) 20:09, 15 April 2009 (UTC)

Trying to sort out the facts, ma’am

• P.S. If we want to turn DynamicDates off, we should do so after Lightbot has delinked the vast majority of dates and turned them into plain text so the number of dates that might be affected will be greatly minimized. After the major delinking, Lightbot could even go back for a second pass through Wikipedia and correct the date‑formatting syntax so that when DynamicDates is turned off, there won’t be any junk like I wrote of, above. I should think it would be a trivial task for Lightbot to translate syntax like [[2009-04-15]] into syntaxes that will look proper after DynamicDates is turned off.

  I see no need whatsoever to rush into turning off DynamicDates at this juncture. I think some editors here are over‑minimizing the consequences of having so many dates to correct. I see no reason to have some of Wikipedia’s articles look poor for months and months as humans chase this stuff down and correct it by hand. That makes no sense to me at all. You don’t put a shotgun in your mouth to blow out a bad tooth; you go from bad to real bad in a hurry. Whatever we do, the objective should be to have the most benign impact for our I.P. users during the transition. Greg L (talk) 19:04, 15 April 2009 (UTC)

← I think I wasn't clear enough. With Dynamic Dates turned off, date links will work exactly like links to any other article, much like they did before Dynamic Dates was implemented in the first place. [[April 15]], [[2008]] will look like April 15, 2008, [[15 April]] [[2008]] will look like 15 April 2008, [[2008-04-15]] will look like 2008-04-15, [[April 15]][[2008]] will look like April 152008, and so on. --A. di M. (formerly Army1987) — Deeds, not words. 19:34, 15 April 2009 (UTC)

The date display option in "Preferences"

• Are you sure of this A. di M. (Army)? And were you mistaken when you wrote that [[2008-04-15]] will look like 2008-04-15, which is a broken link? If it is a broken link, that doesn’t sound good. And why did you write this(?):

“ So the short term conseguences would be crappy (April 152009), but I don't consider that a downside. ”

If true, I see that as unnecessarily disruptive to our I.P. users. Note that I see that date example as April 152009 (linking disabled in this example). Is your date pref turned off? Mine is. Maybe you and I seeing different things here in these posts. Greg L (talk) 19:59, 15 April 2009 (UTC)

Yes I have set "No preferences". As for the "So the short term ..." sentence, I meant that an ugly coding such as [[April 15]][[2009]] would appear just as the code suggests, namely two links with no space between them. It's very rare compared to [[April 15]], [[2009]], but I have seen it on a couple articles. --A. di M. (formerly Army1987) — Deeds, not words. 21:19, 15 April 2009 (UTC)

(ec) ... Greg L wrote "After the major delinking, Lightbot could even go back for a second pass through Wikipedia and correct the date-formatting syntax so that when DynamicDates is turned off, there won’t be any junk like I wrote of, above." but DD doesn't work on dates that aren't linked - when Lightbot removes the links, the "junk" will be exposed. my point is that i'd rather have the "junk" exposed before Lightbot does its thing, so that there's no risk of people misperceiving Lightbot as *creating* the "junk".

meanwhile i'm surprised if you expect typos like April 152009 to be frequent enough to qualify as "disruptive". you may be right, but they may also not be that common; and in any case they can be corrected by anyone who sees them until Lightbot or its relatives gets around to it. Sssoul (talk) 20:31, 15 April 2009 (UTC) ps: for the record, i've had my date preferences turned off for months.

←

• Quoting you: …but DD doesn't work on dates that aren't linked - when Lightbot removes the links…: Yeah, I know. What I am talking about is that some date forms are particularly nasty for I.P. users. Examine this table:

What you type	What logged-in registered users see (settings on first row)					What others will see [A]	DynamicDates disabled
	May 15, 2001	15 May 2001	2001 May 15	2001-05-15	No preference
[[May 15]]	May 15	15 May	May 15	May 15	May 15	May 15	May 15
[[15 May]]	May 15	15 May	15 May	15 May	15 May	15 May	15 May
[[May 15]], [[2005]]	May 15, 2005	15 May 2005	2005 May 15	2005-05-15	May 15, 2005	May 15, 2005	May 15, 2005
[[May 15]] [[2005]]	May 15, 2005	15 May 2005	2005 May 15	2005-05-15	May 15, 2005	May 15, 2005	May 15 2005
[[May 15]][[2005]]	May 15, 2005	15 May 2005	2005 May 15	2005-05-15	May 15, 2005	May 15, 2005	May 152005
[[15 May]] [[2005]]	May 15, 2005	15 May 2005	2005 May 15	2005-05-15	15 May 2005	15 May 2005	15 May 2005
[[2005-05-15]] [B]	May 15, 2005	15 May 2005	2005 May 15	2005-05-15	2005-05-15	2005-05-15	2005-05-15
[[2005]]-[[05-15]] [B]	May 15, 2005	15 May 2005	2005 May 15	2005-05-15	2005-05-15	2005-05-15	2005-05-15

What I am talking about is the bottom style. That little bastard, even with DynamicDates turned on, produces crap for I.P. users (the 99.97% of our readership referred to as “others” in the far-right column). When DynamicDates is turned off, it seems this style turns into even crappier looking stuff like April 152009 (need to confirm the facts on that one). What I was saying is that IF we want to turn off DynamicDates at some later date, the dates that haven’t been delinked by Lightbot will no longer format. Commas won’t appear. A number of things may happen. We need to get to the exact bottom of the facts (see below). But what I am saying is that after we’ve removed linking from most articlespace, and only our intrinsically chronological articles like 1985 have oodles of linked (and autoformatted) dates, IF we want to turn off DynamicDates (which I still think is unwise), we can first have Lightbot go through and convert dates formatted in the bottom style to dates in one of the two styles immediately above it. Greg L (talk) 20:42, 15 April 2009 (UTC)

---

ALL: I’m wearing my fingers out using trick aliasiing techniques to over-ride the cursed autoformatting to ensure that everyone sees the same thing when we write examples—even if someone has their date preferences turned on to one setting or another. Maybe all parties participating here can agree to turn off their date preferences so we are certain to all be seeing the same thing?

Short of that, here is “date‑link blue”: <font color="#002BB8"> April 152009 (example date that doesn’t link) </font color>. Maybe that will help. Greg L (talk) 20:12, 15 April 2009 (UTC)

---

Question: Can we setup an example subpage where DynamicDates doesn’t work? If so, I will be more than pleased to contribute some example dates. Greg L (talk) 20:22, 15 April 2009 (UTC)

P.S. Or, alternatively, if there is someone who really, fully knows exactly how DynamicDates works, add another column to the above table illustrating “DynamicDates disabled.” Greg L (talk) 20:54, 15 April 2009 (UTC)

I don't think I really, fully know exactly how DynamicDates works, but AFAICT without it links to dates would work exactly like any other "ordinary" link, and I think I really, fully know exactly how "ordinary links" work. BTW, you can disable DynamicDates on a particular link by adding a colon after the opening brackets ([[:).

What you type	DynamicDates disabled
[[May 15]]	May 15
[[15 May]]	15 May
[[May 15]], [[2005]]	May 15, 2005
[[May 15]] [[2005]]	May 15 2005
[[May 15]][[2005]]	May 152005
[[15 May]] [[2005]]	15 May 2005
[[2005-05-15]]	2005-05-15

--A. di M. (formerly Army1987) — Deeds, not words. 21:19, 15 April 2009 (UTC)

• That seems to make sense now. I can wrap my mind around all of that. Thank you, A. di M., for adding the column to the top table and for adding the bottom mini-table. [FWIW, It wasn't me who added the column to the top table. --A. di M. (formerly Army1987) — Deeds, not words. 22:47, 15 April 2009 (UTC)]
  [I see that now. It was added by an I.P. editor who traces back to the Office of the President at the University of California. Probably a logged-out regular here.] Greg L (talk) 00:20, 16 April 2009 (UTC)

  So… my point is that we wouldn’t want to disable DynamicDates right out of the gate. Even if Lightbot goes through and delinks 99% of the now‑linked dates in regular articespace, there will still be all those intrinsically chronological articles (1985, and April 15) that Lightbot will leave alone. These articles contain a wide variety of date syntax and three (now) four of the syntaxes don’t take well to having DynamicDates simply turned off. The type shown at the bottom would become a broken link (since it wouldn’t be parsed into its two component links) and two three other date syntaxes would look improper to our I.P. users. That’s no good at all. I simply can not see how it would be wise to have so many dates on Wikipedia look like crap (by simply turning off DynamicDates) and having humans chase all these ugly-ass dates down and manually repair them. What the hell(!), that’s what bots are for.

  Before DynamicDates could be turned off, an updated version of Lightbot would have to go through Wikipedia (after it’s delinked non‑MOSNUM-compliant dates), find these three four types of date syntax, and put them into a form that will look proper with DynamicDates turned off.

  As I mentioned above, whatever we do, any changes we make to Wikipedia should have the most benign impact for our I.P. users during the transition. That should be our main objective here. Dancing and getting drunk on the grave of date autoformatting can wait. As far as I can see, it is clearly best for our readership to leave DynamicDates turned on while Lightbot is let lose to make Wikipedia MOSNUM-compliant (assuming Lightmouse doesn’t tell us he’ sick to death of dealing with us. Poor bastard, he’s doing a bunch of the work here and he’s just a volunteer, like everyone else).

  I see the bottom line as this: we can table, for the moment, discussion of turning DynamicDates off, invite Lightmouse back into the fold (after being scolded “naughty naughty–you”) and see if he is willing to do a bunch of heavy lifting to make Wikipedia MOSNUM-compliant without being attacked left and right by editors who haven’t known of any of these proceedings but suddenly notice their precious linked dates being delinked. They’ll want to have Lightmouse turned into Soylent Green. There needs to be a decision by the ArbCom instructing that he is supposed to be responsive to legitimate concerns of editors if his bot activity is not in compliance with MOSNUM and/or has unintended behavior. But he also needs to know that he has the community’s backing if some editor flies in and claims that MOSNUM is all screwed up and he wants his linked dates back in his favorite article. No ANI's and no ArbComs. It’s settled. And Lightmouse can go back to providing the service to Wikipedia that he has long done (quite well too). Greg L (talk) 22:25, 15 April 2009 (UTC)

What happened to no polling?

I'm sure we used to have several policies/guidelines/etc. saying we don't decide things by poll. (WP:PNSD, for example.) Stifle (talk) 10:57, 15 April 2009 (UTC)

Well, given how divisive this dispute has been and numerous RfC's we've already had, we were left with little option than to have a poll. With the greatest respect Stifle, this has been advertised for months and the poll concluded on Sunday - why wait till now to voice your concern? Ryan Postlethwaite^{See the mess I've created or let's have banter} 11:02, 15 April 2009 (UTC)

Perhaps it is rather worth considering that that this was more than just a poll. The instruction at the top of each section was "Please indicate your vote under ONE option, accompanied by a concise explanation for your choice. Your explanation is important in determining the community consensus." By asking for an explanation of contributors' choice, it can be seen as half-way house to an RfC, without the distraction of having threaded responses. I actually think this style has some benefits over a full-blown RfC, particularly where the focus is on determining consensus on specific, narrowly-defined issues. --RexxS (talk) 15:27, 15 April 2009 (UTC)

I was just interested in knowing. I like voting, myself; it's much cleaner. Stifle (talk) 08:32, 17 April 2009 (UTC)

Autoformatting - take II

I wouldn't mind a little help here to get things worked through. What is clear is that dates are no longer going to be routinely linked (at this stage, I'm not wishing to look at how dates will be mass unlinked, but I'll look at this later as it's an important point). With dates not being routinely linked, autoformatting will not work correctly. A major problem is that the majority of dates that aren't linked won't be autoformatted, but the few that are will be autoformatted - that's something that I suspect most would believe is unacceptable. However, the autoformatting of dates in article histories and logs is still required. What I want to know is the options we have. Would disabling DynamicDates (setting $wgUseDynamicDates = false in LocalSettings.php) lead to autoformatting being stopped in articles, whilst leaving the option to change date preferences in article histories and logs intact? Ryan Postlethwaite^{See the mess I've created or let's have banter} 21:48, 15 April 2009 (UTC)

• I think your concerns are addressed in my 22:25, 15 April 2009 (UTC) post, above. As to getting de-linked dates properly formated in fixed text, that is something Lightbot automatically does. Greg L (talk) 22:39, 15 April 2009 (UTC)

• Date autoformatting is not considered desirable by a majority of Wikipedians, but a significant minority of opinions expressed are interested in its potential. Linked dates are despised by an overwhelming majority of Wikipedians. I'm coming around to Greg's suggestion that delinking first followed by switching off Dynamic Dates would be the path of least disruption to giving the community what it wants within as short a timeframs as is possible/reasonable.

  BTW, I'm just assuming the discussion here is based on an informed view of the functionality of Dynamic Dates. If a more authoritative input is required, then let's take the time to get it before switching it off. Ohconfucius (talk) 03:25, 17 April 2009 (UTC)