08:23
<Tim van der Lippe>
Would be great if that could be addressed. Should I maybe file an issue on the repository instead if WebKit is open to improving the script?
14:29
<annevk>
I suspect that already exists
14:59
<annevk>
Hmm, the person in favor of cross-origin storage has gone full AI. Except that as far as I can tell AI hasn't solved the basic privacy issue. Good times.
15:00
<Noam Rosenthal>
Link?
15:01
<annevk>
https://github.com/whatwg/fetch/issues/1954 is the latest in a series of things I've seen come by.
15:07
<Noam Rosenthal>
There's a lot in https://github.com/WICG/cross-origin-storage#privacy-considerations, though I haven't dived into it
15:12
<annevk>

That starts with

In browsers that still support third-party cookies, user agents are expected to make this API available only in contexts where third-party cookies are enabled.

meaning it's a Chromium-only API as Chromium has a privacy hole.

15:13
<annevk>
But anyway, recent posts sought to address that, but it seemed primarily AI-driven without addressing the known issues with hash-based caching.
15:15
<Noam Rosenthal>
gotcha
15:41
<annevk>
Ah I just found most of the remaining context in my notifications: https://github.com/WICG/cross-origin-storage/pull/73
15:46
<Noam Rosenthal>
Github actions is down/moody: https://www.githubstatus.com/ So some html builds are failing
15:48
<annevk>
Also impacts WPT. Not too bad if you're in Europe I suppose.
15:53
<Noam Rosenthal>
yea I'm sure Thomas Steiner consults with our privacy experts and is open to feedback, despite the AI
15:58
<annevk>
Sure, but maybe do that before you ask other companies to take a look
16:15
<Noam Rosenthal>

For whoever is editing the HTML standard source with LLMs (most/all of us?) I'm playing around with a skill+script that first splits the file to sections by h2, than performs the edits, and then re-concats. I found that otherwise LLMs (at least gemini) keeps trying to write bespoke python scripts to edit the giant file and spends a lot of tokens/time on this just-in-time tooling

https://github.com/noamr/web-archeologist/tree/main/skills/html-spec-splitter

16:15
<Noam Rosenthal>
See fix (https://github.com/whatwg/html/pull/12846) Tries to follow what blink does which I think is sensible
16:26
<zcorpan>
@noamr:matrix.org: maybe we can reconsider splitting source
16:30
<Shannon Booth>
on a very loosely related question, is there a single page version of the MR diff viewing thing linked on MR? or is it split into the different sections only
16:34
<Noam Rosenthal>
I'm always supportive of that.
17:31
<annevk>
What's the MR diff viewing thing?
17:33
<annevk>
Optimizing for AI while that's still changing week-to-week seems premature. I'm always open to people splitting a logical chunk away from HTML into its own standard though, as we did with Fetch, Infra, XMLHttpRequest and various other aspects.
17:34
<Noam Rosenthal>
It's not just for AI, also for github diffs and for IDEs. I don't know what having a single file is optimized for
17:46
<Noam Rosenthal>
but anyway, until there's any consensus on that, that kind of tool can help with the current version of what AI does
17:50
<Noam Rosenthal>
Re modularization, I think contenders can be canvas (8%, was a separate spec before...), microdata (2%), messaging (1.4%), drag&drop (1.3%). But that would make a small and temporary difference in terms of file size
18:06
<Noam Rosenthal>
Green now
18:10
<Shannon Booth>
Ah, sorry, I mean whatpr
18:19
<Luke Warlow>

https://html.spec.whatwg.org/multipage/forms.html#category-listed - is it just me or does the definition for listed elements read as if the form IDL attribute can also be used to specify a form owner?

These elements also have a form content attribute, and a matching form IDL attribute, that allow authors to specify an explicit form owner.

18:21
<Noam Rosenthal>
Isn't that what that attribute does?
18:22
<Noam Rosenthal>
(it doesn't specify that it's an actual element vs. an IDREF)
18:22
<Luke Warlow>
Yes the attribute does do that. But the form IDL attribute doesn't? To me this reads as if both allow you to do that.
18:24
<Noam Rosenthal>
ah yea it reads a bit implicit in that sense as if the IDL attribute is also a setter
18:46
<annevk>
That does not exist. Tobie would remember why. Prolly some tools cannot deal with the large input size.
19:10
<zcorpan>
I didn't mean changing what we publish, to clarify
21:36
<smaug>
jarhar: FWIW, we've been considering more pip-window like API for "Persistent Widgets". That would solve various issues around session history handling for example. (but might introduce others)
21:39
<smaug>
(but perhaps a new element is effectively that in some sense. The old proposal was iframe, and that wasn't good)
21:53
<jarhar>
thanks for the feedback! yeah having a PiP-like window makes sense for some cases, but the primary use cases seem to be more focused on the embedder having full control over how its rendered like an iframe
22:48
<smaug>
well, embedder would have. It would be bound to something in the new page
22:49
<smaug>
new element isn't less magical
22:50
<smaug>
pip approach might be also simpler for things like focus handling during page load. Maybe. But a new element might be then simpler after page load