Wikipedia talk:Requests for comment/Tony1

Reply to Outside view from SandyGeorgia

• The point about developers not responding. UC Bill has already provided one patch which IMHO addresses part of the problem and might be creating a new one. However, he seems willing to continue development if the requirements can be clarified. dm (talk) 13:37, 9 September 2008 (UTC)

Huh?

> Once it's unlinked, it's gone.

Because once a date is unlinked, it can never be relinked? Am I missing something here? WhatamIdoing (talk) 05:10, 10 September 2008 (UTC)

Right now, you have data that's been linked and identified/formatted as a date by an editor, there's only a few forms it can take. Once you make it freeform text, editors will do whatever they want and it will be much more difficult to parse it out programatically. Theres a lot more edge cases. Was it linked in the first place, was it deliberately not linked, etc. dm (talk) 06:23, 10 September 2008 (UTC)

Can I have an example? Are we exclusively talking about dates in text, as opposed to dates in references? WhatamIdoing (talk) 20:02, 10 September 2008 (UTC)

Here is the regex to match a date in the format MYD, linked:

/(.*)\[\[{$month_regex}[ _](\d{1,2})]] *,? *\[\[(\d{1,4}([ _]BC|))]](.*)/iu

Now, if you simply remove the "\[\[" and "]]" you end up with this:

/(.*){$month_regex}[ _](\d{1,2}) *,? *(\d{1,4}([ _]BC|))(.*)/iu

The problem with that is that it will match this:

May 1995

The regex will treat that as if it were the date May 19, 95.

If you modify the regex so that it requires a space following the comma, that will prevent false matches like the one above, but it will cause the regex to NOT match dates like this:

May 12,2005

There are ways to fix the regex so it will match unlinked dates correctly, but it is MUCH harder to do if you want to catch all the slightly-misformatted dates. With linked dates, it's much easier.

Understand? --Sapphic (talk) 02:46, 11 September 2008 (UTC)

I have two disagreements with this line of argument:

1. Back when autoformatting was recommended, many editors set up regexes to link dates, including standardising dates that were in non-standard formats, so it can be done and we have the technology
2. There are a lot of linked dates out there that are in non-standard format e.g november 5th, 2002, 2007-12-09. Just because an editor has linked it doesn't mean they did it right.

Colonies Chris (talk) 11:12, 11 September 2008 (UTC)

Do I understand? No. Are you telling me that people have been putting that unintelligible ~70-character-long string of apparent nonsense into regular articles on purpose? And the pro-linking people are mad because that string of nonsense has been turned into plain old text, i.e., "May 1995", that any human can read, understand, and more importantly, edit easily? Surely I have misunderstood something here. WhatamIdoing (talk) 02:25, 12 September 2008 (UTC)

Yes, you've misunderstood. The unintellible string is a regular expression, a formal specification of a pattern against which text is to be matched. A bot can use a regular expression to detect strings that look like dates. Sapphic is asserting that it is very easy to code a bot to reliably detect linked dates and unlink them, but very hard to code a bot to reliable detect unlinked dates and link them. Hesperian 02:55, 12 September 2008 (UTC)

That makes sense. Thanks for the clarification. So the re-linking would have to be supervised by a human (using a script), i.e., exactly the process that Tony1 is using now to de-link the dates. WhatamIdoing (talk) 03:23, 12 September 2008 (UTC)

That is the crux of why we're asking to hold off further unlinking while we propose something which can take the existing linked dates and make them into something more palatable for everyone. If we wait till Tony et al are done, you've just made the job much harder. Unless Tony and everyone will be as eager to go do it over again :) dm (talk) 03:59, 12 September 2008 (UTC)

"Numerous complaints"?

My experience suggests otherwise. Over the last few weeks I've unlinked hundreds of articles on a wide range of subjects. In almost all cases there was no reaction of any kind, pro or anti. In three or four cases I was asked why, I gave my reasons, and they were accepted. In two cases, editors objected in immoderate terms and would not discuss it. That's not "numerous complaints", it's just the routine difficulties that arise as a change of policy starts to affect people who weren't aware of the change. Colonies Chris (talk) 12:33, 10 September 2008 (UTC)

I agree. I used to enthusiastically link dates for autoformatting, until the recent discussion convinced me that WP is better off without (the current brand of) date autoformatting. I have since been removing such links wherever I see them, and have yet to see anyone complain about it. -- Donald Albury 14:14, 11 September 2008 (UTC)

RfC instructions

Please see the RFC instructions; other than to endorse a view, please edit only your own section. If you want to start a threaded discussion, the talk page can be used. SandyGeorgia (Talk) 05:19, 13 September 2008 (UTC)

Outside view by Arthur Rubin (reply to Orderinchaos)

I see about a 5% grammatical error rate in the sample edit Orderinchaos mentions above. Not bad for a person, but unacceptable for a bot which will make millions of edits. — Arthur Rubin (talk) 17:02, 11 September 2008 (UTC)

It's not a bot, but a human-supervised script; it does not make "millions of edits", and does not run without scrutiny by its operator. Tony (talk) 04:58, 13 September 2008 (UTC)

Which bot would that be? One out of 'millions' is not a good statical basis. Ceoil ^sláinte 19:35, 13 September 2008 (UTC)

I'm curious whether we can get diffs of any grammar errors. I find no "sample edit" mentioned by Orderinchaos. WhatamIdoing (talk) 18:17, 15 September 2008 (UTC)

Orderinchaos wrote "edits like this give me confidence...." They don't give me confidence, due to the 5 grammatical errors added. — Arthur Rubin (talk) 00:55, 16 September 2008 (UTC)

Am I being unusually dense? I found no grammar errors. Are you aware that "16 September" is considered a legitimate date format for text? It is not in the American (middle endian) style, but it is widely used in the rest of the world. WhatamIdoing (talk) 01:45, 16 September 2008 (UTC)

Perhaps he's referring to missing commas after weekdays; depending on your user prefs, those errors were there anyway. For some users, they were camouflaged by user prefs, for others they weren't, which is Tony's whole point. SandyGeorgia (Talk) 01:49, 16 September 2008 (UTC)

I don't think so. He specified that the errors were added; there were no commas after the weekdays in the original text. WhatamIdoing (talk) 02:34, 16 September 2008 (UTC)

Right, but it's possible Arthur Rubin doesn't understand the errors were there anyway, because his user prefs may be set such that he doesn't see them; if that's the case, it would mean he hasn't yet understood the whole issue with autoformatting (that it camouflages errors for some users, while most still see them). SandyGeorgia (Talk) 02:40, 16 September 2008 (UTC)

I would be interested to know exactly what were the 5 errors the diff above introduced, because I only see one in the 104 dates adjusted, and that one is a known issue that may be addressed already in the script. Gimmetrow 02:55, 16 September 2008 (UTC)

Based on his comments in the "Conclusion" section, I think Arthur is blaming Tony for not correcting errors (missing commas) that were already present in the text (but possibly camouflaged by autoformatting). Of course, since Arthur declines to copy and paste a line or two out of the diff to provide us with an example, we're just guessing here. WhatamIdoing (talk) 05:11, 27 September 2008 (UTC)

Rv closing

Sapphic, I see that you reverted the closing of this RFC/U. Are you aware that there is no magic 30-day timer for RFC/U's? They're always closed by humans, whenever it seems appropriate. RFC/U is not a 30-day badge of shame. WhatamIdoing (talk) 18:02, 19 September 2008 (UTC)