[01:48:48] Did this answer your question @amire80 ? So, yes, you can call a function from a client wiki. The function call will be stored in the wikitext of the client wiki, not the result of the function call. That will be generated and cached. The functionality for calling a function from a client wiki is not implemented yet, and will not be until after launch of Wikifunctions. [03:36:20] But is that how abstract articles will be generated? I thought that there would be one abstract version, common for all languages, stored on Wikidata. And the client wiki somehow would indicate to want to use it or not. (re @wmtelegram_bot: Did this answer your question @amire80 ? So, yes, you can call a function from a client wiki. The function call will...) [03:38:59] Yes, that's correct! The functions will be in Wikifunctions, the abstract content (modulo community decision) on Wikidata. In that case we wouldn't need to create a function in the local Wikipedia, but it could be automatically connected. [03:45:50] But I realize that for "type generation" that seems odd to store on each item. So a client missing an article about Q357142 could call something like {{abstract|type=Q13382608}} (and another client might use {{abstract|type=Q947873}}). Or would we have bots running around adding a model to items, perhaps even involuntarily edit warring? [03:56:21] No, no. The abstract content could be just a call to the type-based text. So the abstract content for Q357142 could be just a call to Q13382608(Q357142) or something like that. [04:03:47] Ok. But then I imagine that when someone made a decent type article for skiers, they are likely to run a bot to add that to all items with that occupation. And when someone does the same for a television presenter they will do the same. So we need to come up with strategies to not have these two bots fight each other since this item is quite prominent in both occupations. [04:05:14] I'd hope that if an article has already content no bot would just overwrite that. That would be like Rambot overwriting the article for New York City. [04:06:29] Not the article, the abstract content. And potentially, and hopefully more often than not, with better abstract content. (re @vrandecic: I'd hope that if an article has already content no bot would just overwrite that. That would be like Rambot overwriting the arti...) [04:06:59] The abstract content is basically an abstract article. [04:12:04] Yes. So it might be a tiny abstract article about skiers, giving three sentences. And the content for television presenters might be a much more detailed model, generating more text, possibly tables with awards and infobox content. So what would we want to do then? (Less controversial would perhaps be to replace the skier function with an Olympic skier function.) [04:47:29] One could take the parts that are used in the two functions for skiers and TV presenters and create a bespoke article for Mr Svan, lovingly and manually combining the abstract content into a unique article. [04:48:06] One could refactor parts of the functions and then reuse them in both. [04:53:06] Agreed, that would of course be a nice solution. Still, I think we're going to need some written guidance on when it's fine to, at scale, replace abstract content. (Manually, this should always be fine if done in good faith.) [04:54:12] Yes, agreed. We'll need to figure that out, what will be the most useful for the readers. [07:19:39] Why after the launch? (re @wmtelegram_bot: Did this answer your question @amire80 ? So, yes, you can call a function from a client wiki. The function call will...) [08:15:58] Because before launch there are no functions in Wikifunctions to call yet? I'm not sure I understand the question [08:27:42] But this means that for some time after launch, functions won't be very useful, and this time may be quite long. (re @vrandecic: Because before launch there are no functions in Wikifunctions to call yet? I'm not sure I understand the question) [08:28:38] If the API works on day 1, then it may still be very useful (re @amire80: But this means that for some time after launch, functions won't be very useful, and this time may be quite long.) [08:28:50] I hope functions will still be useful, but you wouldn't be able to call them from the other Wikimedia projects [08:29:43] That's no different than Wikidata, which launched with Sitelinks that no one could use yet [08:31:04] It's basically just the question "do we implement that feature before or after launch?" It'll take equally long either way, but by doing it after launch we can serve other use cases sooner [10:22:36] If I recall correctly, the functionality to use sitelinks on the client was more or less ready on day one, and the clients were simply not immediately _enabled_. Hebrew and Hungarian Wikipedias had them enabled very quickly after launched, and other languages followed not so long after that. So the wait was very short. (re @vrandecic: That's no different than Wikidata, which laun [10:22:37] ched with Sitelinks that no one could use yet) [10:22:45] And the usefulness was almost immediate. [10:23:16] Will the capability to call functions from a client also be developed so quickly? [14:52:11] I hope so. If I remember it right, it's the first post-launch feature to be worked on. [18:38:18] Very interesting project. [19:31:59] I've been working for a while on Swiss political articles (particularly today the federal council by-election that took place this morning), it's especially visible on these articles how wasteful it is to have three/four different wikis dealing independently with the exact same stuff, so I'm really hyped for the launch of Abstract Wikipedia [23:23:29] Yay! Thank you both for the kind words