[05:11:01] This is what one of the sections looks like when the whole thing is defaulting because there are no labels (let alone lexemes). Any suggestions on improvements? The pencils are links to Wikidata. I could try to roll them out on even the currently unlinked QIDs, but wanted to seek feedback first. : https://tools-static.wmflabs.org/bridgebot/39031738/file_81087.jpg [05:16:57] 0716 [06:05:23] well a fallback could be a solution, but it's a very bad solution (especially for minorized languages) (re @u99of9: This is what one of the sections looks like when the whole thing is defaulting because there are no labels (let alone lexemes). ...) [06:07:00] I guess that for cases like these there is no good solution... [06:33:35] Another section of that same language looks like this. I got the impression before that English-default is not desirable (and IMHO the red crosses make it worse). It would be good to have guidance. : https://tools-static.wmflabs.org/bridgebot/4e109b21/file_81092.jpg [06:36:17] Are the pencils and crosses are from a user script? [06:36:20] Yeah I think English is a bad default here ; at the very least it should default to Romani first and even that would be suboptimal... [06:36:20] Ping @amire80 what do you think? (re @u99of9: Another section of that same language looks like this. I got the impression before that English-default is not desirable (and IM...) [06:38:31] No, we have put them in. I added the pencils recently here: Z36490. The red crosses are from Z36095. (re @Jan_ainali: Are the pencils and crosses are from a user script?) [06:39:50] Sorry, my question is not about whether there is a better fallback for this specific language. I tried to choose a language which has no fallback options, so must go to the full-on backstop function. (re @NicolasVIGNERON: Yeah I think English is a bad default here ; at the very least it should default to Romani first and even that would be suboptim...) [06:42:03] I would think that fallbacks should only be applied if their output would be acceptable in Wikipedia. That would lead to a very limited usage of fallbacks, and it is not what I used to do myself, I was more generous before. [06:42:31] It would be better to show an error, I think. This helps to collect the errors, and do e.g. statistics of what errors show up, because of what missing part, etc. [06:43:08] And articles with errors could be treatd explicitly when being integrared, either to be stopped or be specially viewed. [06:43:46] Sure, my point is fallback is bad but some fallback are worse (to be crude) (re @u99of9: Sorry, my question is not about whether there is a better fallback for this specific language. I tried to choose a language whic...) [06:48:10] IMO integration is a separate question. Yes, unavailability of language-configuration in that language (whether errors or defaults) should be a very strong factor against integration in that language. (re @vrandecic: And articles with errors could be treated explicitly when being integrated, either to be stopped from integration or to be put o...) [06:49:39] If you are right (which I'm not sure and will argue), the errors are not good enough yet to give regular users information about how to fix them. (re @vrandecic: It would be better to show an error, I think. This helps to collect the errors, and do e.g. statistics of what errors show up, b...) [06:52:35] but we can't tell if a result is a default, it just looks like a succesful evaluation? (re @u99of9: IMO integration is a separate question. Yes, unavailability of language-configuration in that language (whether errors or defaul...) [06:53:50] what's missing? can't we say in an error message "there's no Croatian function for Z1234"? (re @u99of9: If you are right (which I'm not sure and will argue), the errors are not good enough yet to give regular users information about...) [06:53:55] if by "we" you mean the computer, then sure. But if you mean the community that is choosing to integrate that article, then I think they are better placed to see if the text reads well in their language. (re @vrandecic: but we can't tell if a result is a default, it just looks like a succesful evaluation?) [06:54:25] yeah, I meant the computer. [06:54:42] good point about the community [06:54:57] Then I think you have the wrong definition of "we". Humans first and everything... (re @vrandecic: yeah, I meant the computer.) [06:57:57] @vrandecic here's my other request on integration (re @u99of9: My request: Once you have the tech ready to host AW articles on local language wikis, please don't enable it until we have had a...) [06:59:24] It's a generic "we" I think. Just like the generic "you" it becomes confusing. (re @u99of9: Then I think you have the wrong definition of "we". Humans first and everything...) [06:59:25] I have read that and the team has noted it. We will have an honest, open-ended conversation when it gets to it. (re @u99of9: @vrandecic here's my other request on integration) [06:59:30] To be fair, we also need to remember that the articles will change (expand), which will sometimes make for text which no longer all works. Should that be like version control where the community has to decide whether to accept the new version yet? (re @u99of9: if by "we" you mean the computer, then sure. But if you mean the community that is choosing to integrate that [06:59:30] article, then I th...) [07:00:34] that is *not* the plan — the hope was that once integrated, not every single version needs to be approved. that would overwhelm the communites. (re @u99of9: To be fair, we also need to remember that the articles will change (expand), which will sometimes make for text which no longer ...) [07:03:10] Okay. So with your erroring plan the article would become unavailable or have error messages through it (perhaps a lot will build up since it's harder to clear an error than to click approve on a revision)? (re @vrandecic: that is *not* the plan — the hope was that once integrated, not every single version needs to be approved. that would overwhelm ...) [07:03:27] to be fair, I was thinking of "we" as in "the sociotechnical project that is Wikimedia, including its communities and the Websites with their technical capabilities". In this case, if the computer can tell whether an article output has issues, it seems beneficial as it could effectively help the human contributors to deal with those (re @u99of9: Then I think you [07:03:27] have the wrong de [07:03:27] finition of "we". Humans first and everything...) [07:04:36] I don't have the answer for that yet — something I would like to have input. But with the erroring it becomes possible to do such things. (re @u99of9: Okay. So with your erroring plan the article would become unavailable or have error messages through it (perhaps a lot will buil...) [07:05:55] Yes, computers are good helpers. If it helps, I think something like David S's tool could tell us whether an HTML fragment is relying on a default function. (re @vrandecic: to be fair, I was thinking of "we" as in "the sociotechnical project that is Wikimedia, including its communities and the Websit...) [07:07:36] Is “the plan” documented somewhere? (re @vrandecic: that is *not* the plan — the hope was that once integrated, not every single version needs to be approved. that would overwhelm ...) [07:11:47] I'm not the expert here. Maybe because I haven't worked on them, the errors I see are often unhelpful. But, for example: "Wikifunctions returned a failed response: No matching lexeme for item in language" should have a link to the item. "Wikifunctions returned a failed response: Reached rate limit in orchestrator" should point to the function that failed. (re@vrandecic [07:11:47] : what's m [07:11:48] issing? can't we say in an error message "there's no Croatian function for Z1234"?) [07:12:59] ha! that's a good question! I would love to say yes and say here, but I first need to look. It's not meant to be a secret. I'll try to find it, or produce it! (re @Al: Is “the plan” documented somewhere?) [07:14:34] agreed! I thought we can do better. I noted this as something to look at. (re @u99of9: I'm not the expert here. Maybe because I haven't worked on them, the errors I see are often unhelpful. But, for example: "Wikifu...) [07:16:54] Currently an error cannot incorporate a call-to-action; it must be plain text. (re @vrandecic: agreed! I thought we can do better. I noted this as something to look at.) [07:16:58] I would be much more encouraged to try to improve the first of these displays than the second: : https://tools-static.wmflabs.org/bridgebot/35a5d788/file_81095.jpg [07:16:59] https://tools-static.wmflabs.org/bridgebot/ae4636a1/file_81096.jpg [07:17:23] Thank u, 6835 (re @join_captcha_bot: Hello ‎@eviandriani5, please write a message with the sequence of numbers highlighted in this video to verify that you are human...) [07:18:03] (The first is https://abstract.wikipedia.org/wiki/Q408 in Swedish. The second is https://abstract.wikipedia.org/view/en/Q335618 in French.) (re @u99of9: I would be much more encouraged to try to improve the first of these displays than the second:) [07:20:44] the text can be a CTA? or do you mean a link? What would be linked for "Lexeme for item missing"? (re @Al: Currently an error cannot incorporate a call-to-action; it must be plain text.) [07:21:39] And if the second stated the relevant QIDs or ZIDs? Or is it about the red wall? (re @u99of9: I would be much more encouraged to try to improve the first of these displays than the second:) [07:27:03] Both. The rational incentive is from seeing meaningful structure and knowing that even a single edit will improve it. The emotional disengagement is from the red wall. Wiki vs Brittanica if you like. (re @vrandecic: And if the second stated the relevant QIDs or ZIDs? Or is it about the red wall?) [07:31:32] Neither picture includes a proper CTA. IMHO the default/fallbacks should include a linked ✏️, and the errors should include links. But that is not the difference between the pictures. (re @vrandecic: And if the second stated the relevant QIDs or ZIDs? Or is it about the red wall?) [07:34:01] understood [07:39:54] Here are my earlier musings (and Al's response) on missing data. (re @u99of9: There is a spectrum of ways we can deal with errors and missing data. 1. Throwing exceptions will rarely help an AW user. 2. Re...) [07:40:24] I mean a message containing a link. If there’s nowhere to link to, for action, there should be a link to information about how to perform the action. I think it will often be necessary to add a link from a sense to an item when only the item is known, but that feels like a Wikidata workflow. (re @vrandecic: the text can be a CTA? or do you mean a link? What would be [07:40:24] linked for [07:40:24] "Lexeme for item missing"?) [07:52:14] Good to hear. I am often surprised by tech changes on Thursdays, even though I follow many phabs. (re @vrandecic: I have read that and the team has noted it. We will have an honest, open-ended conversation when it gets to it.) [07:59:21] Taking this in a different direction... Perhaps the timing of the newsletter announcing changes could be more closely synchronised with when those changes started applying? How far in advance do you know which changes will ship each week? (re @u99of9: Good to hear. I am often surprised by tech changes on Thursdays, even though I follow many phabs.) [08:01:42] I think my fifth point is particularly important. There are other solution options in this space, no doubt, but it should always be possible for a Wikipedia community to block part of an article, and I have taken “part” to apply (by exception) to any call. The “blocked unless checked” policy implied by “version control” would apply at article/section level, I think, w [08:01:42] [08:01:43] ith lower level blocks applied by exception (if necessary) to the revised content before the revised version is released to that community by unblocking the article/section. Something along those lines… 🤔 (re @u99of9: Here are my earlier musings (and Al's response) on missing data.) [08:39:36] That's entirely my fault. Folks on the team want that too. I'm just always late with assembling and writing the newsletter. Sorry! (re @u99of9: Taking this in a different direction... Perhaps the timing of the newsletter announcing changes could be more closely synchronis...) [08:39:50] I'm writing one right now [09:04:07] I'm sure we all keep you very busy! (re @vrandecic: That's entirely my fault. Folks on the team want that too. I'm just always late with assembling and writing the newsletter. Sorr...) [09:04:41] Release it next Wednesday with next week's updates? (re @vrandecic: I'm writing one right now) [09:11:37] but what about this week's updates? [09:46:51] I thought it was somewhat insane at the time. I still don't understand how it could work. But I now do see why you want it. Without something like that, for the foreseeable future the most fully developed AW articles will only be publishable in English! (re @Al: I think my fifth point is particularly important. There are other solution options in this space, no doubt, [09:46:51] but it should always...) [10:33:54] It could be just a function wrapper: “if language not in […] f(x) else g(y)”. As I said, it is for exceptions where, for example, the configured function’s output is unacceptable only for specific items, perhaps because the wrong lexeme gets selected and (conceivably) yields offensive or defamatory output in a particular language. I can see the same approach working for c [10:33:54] [10:33:54] ontrolled rollout of new content in an established article, especially where this relates to living persons. (re @u99of9: I thought it was somewhat insane at the time. I still don't understand how it could work. But I do now see why you want it. With...) [13:53:00] The pencils have started showing up in Swedish. What do you think as a fixer/reader? Thumbs up or down? (re @Jan_ainali: Are the pencils and crosses are from a user script?) [13:57:00] As a fixer, it was really smooth! [13:57:00] As a reader I am less sure. But on the other hand, I still hope we would have a way for the client wiki to say that we don't want to show abstract articles that are not "done". (re @u99of9: The pencils have started showing up in Swedish. What do you think as a fixer/reader? Thumbs up or down?) [14:07:57] I like the way the same fragment highlights across the abstract and rendered columns. I wish that if I would click on one of those, it would auto scroll to get the linked fragment on the other side into view. [14:10:32] Sounds like a good suggestion for T429214 (re @Jan_ainali: I like the way the same fragment highlights across the abstract and rendered columns. I wish that if I would click on one of tho...) [17:56:54] Does anyone feel like looking at a test implentation I wrote and tell we where I'm wrong? https://www.wikifunctions.org/view/en/Z36502 [18:06:50] nvm, I solved it :) [22:07:19] It looks like it is currently working? (re @wmtelegram_bot: Does anyone feel like looking at a test implentation I wrote and tell we where I'm wrong? https://www.wikifunctions.org/v...)