User talk:Mikeblas/Archives/2018/August

Ancient History Citations

Hello Mikeblas, I am Sunriseshore

I saw you that you fixed a citation error I had made on the page but I am not exactly sure what this means or where the error was in the first place. Even after reading the article on citations I still have little understanding of the technical workings on Wikipedia. Please try to explain,

Thank you Sunriseshore (talk) 21:59, 29 July 2018 (UTC)

Hi! We write references with ref tags, like this: <ref name="something">Citation here</ref>. If an article has more than one definition for a reference with the same name, then it's an error. We can't write this: <ref name="something">New York Times</ref>, then this: <ref name="something">People Magazine</ref>, for example, because the names of the reference definitions are the same (they're both "something") but the content of the references are different (one is "New York Times" and the other is "People Magazine"). I think the Help:Cite errors/Cite error references duplicate key article gives a lot more details.

I hope that helps out; if not, let me know what other questions you might have. -- Mikeblas (talk) 22:17, 29 July 2018 (UTC)

Thank you for your quick answer But I see plenty of text with multiple references. How do we keep those from being errors? Ill read your link to try to better understand. And when can I remove the tag you left by the reference? Or how do I solve that problem Sunriseshore (talk) 23:31, 29 July 2018 (UTC)

You shouldn't see more than one definition of a given reference name. It's possible to have multiple definitions of the same reference name, but only if the content of the reference is exactly the same--including white space, capitalizatoin, everything. But that's a bad idea, since it's only a matter of time before someone decides to edit one or the other of the definitions... then, they're different and it's broken. If you have a specific example you'd like to me to look at, I'd be happy to do so. -- Mikeblas (talk) 01:06, 30 July 2018 (UTC)

Hmm, Thank for your answer. I think I will come back later with specific questions. — Preceding unsigned comment added by Sunriseshore (talk • contribs) 2018-08-07T16:22:19 (UTC)

Happy to help! -- Mikeblas (talk) 00:09, 8 August 2018 (UTC)

Why?

Any particular reason you needed to ping me twice in this edit of yours?—CYBERPOWER (Chat) 18:20, 15 August 2018 (UTC)

It looks like a bot you control has made edits that made an article worse instead of better, so I thought you might like to know. -- Mikeblas (talk) 20:13, 15 August 2018 (UTC)

The bot behaved normally. It did not break any citation templates. Duplicate named refs not set up correctly is not a bot issue, but rather an issue with the person who set them up.—CYBERPOWER (Chat) 21:51, 15 August 2018 (UTC)

The refs weren't duplicate before the bot edited them. Before the bot edited the article, the article showed no error messages for references. After the article was edited by the bot, the article did show error messages for references in the article. I don't think there's much room to interpret the edit made by the bot in this case as constructive. -- Mikeblas (talk) 22:28, 15 August 2018 (UTC)

I disagree. The whole point of assigning a name to a reference is they can be pointed to in other places in the article. It's absolutely pointless to duplicate the reference, including name else where. All someone has to do is mistakenly edit one of the references, even with just a space, to trigger that error, and not even realize it. The reference was improperly setup to begin with and thus I don't declare this a bot fault. It's not the bot's job to ensure references carrying identical names, have identical content. Though it would be a good bot job for a different to correct these cases.—CYBERPOWER (Chat) 22:32, 15 August 2018 (UTC)

Then, what is the point of this bot? Can you help me understand why you consider this change to be for the benefit of the article? -- Mikeblas (talk) 22:39, 15 August 2018 (UTC)

The point of InternetArchiveBot is to add archive URLs to no longer functioning links and references, not repair improper usages of ref name invocations. Each reference is treated separately.—CYBERPOWER (Chat) 22:44, 15 August 2018 (UTC)

And it pursuses that goal even if doing so adds errors to the article? Before the bot changed the article, there were no errors. There were three references with the same anchor name. That's not great, but since the content of the references was identical, there's no error. After the bot changed the article, errors appeared. For some reason, the bot decided to add deadurl fields and links to archived content to two of the three references. For the third, it added a dead link template and did not add the archive links. Since the three citations were identical and carried the same name, why did the bot not treat them identically? If it had, the bot would have achieved its goal of adding arhchive URLs and not created a citation error. But in this case, it's the bot itself that mistakenly edited one of the identically-named references (with far more than just a space) to trigger the duplicate name error. The bot apparently didn't realize it, and the author of the bot is defending that behaviour as correct. Why would that behaviour be considered correct? And what improvement does it make for the article? -- Mikeblas (talk) 23:09, 15 August 2018 (UTC)

That is actually a good question, re: 2/3 references being changed. I hadn't thought of that. That is something that would need to be investigated.—CYBERPOWER (Chat) 23:19, 15 August 2018 (UTC)

I will investigate why 3 identical references were not handled the same way, as that is a valid question. As for the rest however, the rest can be explained with GIGO. If references are not going to be set up correctly, then it's reasonable to expect that some automated process may break it. Either you have three of the same references without names, or just print it once, name it, and then invoke the reference in other places on the article. That's how it's intended to be used according to the documentation.—CYBERPOWER (Chat) 23:25, 15 August 2018 (UTC)

Thank you for considering my question. This question is, after all, why I pinged you in the edit I made to repair the damage done to the article by the bot. (Is there some other process for reviewing the behaviour of a bot that I should have followed?) I'm looking forward to understanding why the bot behaved as it did, and I hope my feedback can help prevent it from regressing articles in the future. While the situation where mutliple references have the same name (and precisely the same content) is far from ideal, lots of articles in the corpus are in this state. It seems like bots written to manipulate articles should do no harm, even if they decide to visit an article that isn't in prefect condition to begin with. After all, if every article was in perfect condition, bots wouldn't be necessary at all. -- Mikeblas (talk) 23:29, 15 August 2018 (UTC)

Unfortunately, writing a bot to think like a human is quite a daunting task. Writing a bot to think like millions of humans, well... Needless to say humans create so many different variables bots need to deal with, so not all of them can be considered when they're developed. IABot handles quite a fair amount of them, but evidently not all of them.

I'd love to explain what's causing it, but it doesn't actually seem to occur in the 2.0 rewrite of IABot, as can be seen here. It looks like the edits were made correctly.—CYBERPOWER (Chat) 23:43, 15 August 2018 (UTC)

This isn't an AI problem; we don't need the bot to "think", or even simulate thought. It just needs to be coded defensively. We know that reference tags have names. And we also know the names might be duplicated -- either with or without the content inside being precisely the same. As such, it's an evident requirement that the bot needs to react acceptably to input in that state. Note that the input isn't even ill-formed; it's acceptable to Wikpiedia's parser, and while certainly not preferred by editors, it works. "Reacting acceptably" can include include detecting the multiple-definition condition (which isn't hard), and then not editing the article.

That the bot edited the article even in this state, across the obvious requirement of handling a common case, indicates that the bot authors somehow think the changes the bot makes are more important than keeping error-free articles actually error free. Why would that be? I can't understand making that decision conciously, but your insistent responses seem to indicate otherwise.

I'm not sure I understand your comment that "it looks like the edits were made correctly". I think the diffs above from the article's history show that they weren't. -- Mikeblas (talk) 00:16, 16 August 2018 (UTC)

You misunderstand. I meant, referring to the last diff I posted, I ran IABot 2 on my sandbox’s with a copy of the article that pre-dated the bot’s edit and it didn’t wreck the references as it did on the article. It seems the problem may have already been fixed in the newest version of IABot.—CYBERPOWER (Around) 01:55, 16 August 2018 (UTC)

That's good news! -- Mikeblas (talk) 02:40, 16 August 2018 (UTC)