06:58
<annevk>
Yeah that seems busted somehow. Might be a Bikeshed issue.
07:33
<annevk>
freddy: did you review https://github.com/whatwg/html/pull/12583? I'm curious where we ended up as you also had a Trusted Types proposal around sanitization.
07:43
<zcorpan>
Merged, thanks!
07:57
<freddy>
Adding to my queue, but likely not reviewing this week
08:01
<annevk>
Noam Rosenthal: AI has found a lot of blockers for these 1 TT + 4 HTML stacked PRs. Given that we likely need a couple more rounds here based on what I'm seeing I'd rather merge my PRs.
08:02
<Noam Rosenthal>
Can you send these blockers over?
08:03
<Noam Rosenthal>
That's OK of course, I'll manage with the rebase
08:04
<annevk>
Yeah will do. I guess I'll create a gist again as that seemed to work okay, right?
08:04
<Noam Rosenthal>
yea sure, whatever format. comment/gist/fax all works
08:05
<Noam Rosenthal>
interesting how different people get different results on LLMs for these reviews. Both zcorpan and myself did a few rounds. I wonder if it's a matter of prompt, personal context, or which LLM engine
08:07
<Noam Rosenthal>
annevk: also I'd like to see if anything in the API shape has an issue - I'd like to at least move this to stage 3, or if this is only logical bugs/editorial issues
08:07
<annevk>
https://gist.github.com/annevk/973d21ae390332babf94d6abddaf9a08 Note that the initial HTML PR is covered by both files. Probably the second file has a better review for it.
08:07
<annevk>
I was using Opus 5 at max effort FWIW.
08:09
<annevk>
I think we have agreement on most of these features, except for maybe the Trusted Types integration. At least I know freddy wanted to look at that.
08:10
<zcorpan>
Might be worthwhile to set up a loop that reviews with different models
08:17
<annevk>
It seems my own PRs have also been invalidated, but hopefully they are not too hard to rebase on top of out-of-order streaming.
08:40
<annevk>
Noam Rosenthal: I guess I should wait with further refactoring. I really want to turn the insertion location into a tuple. Though if you can think of changes that would make your PRs easier to review, perhaps it would be good to make those as smaller standalone changes first.
08:45
<annevk>
See https://github.com/whatwg/html/issues/1706 for context. Before vacation I was just trying to address a number of old HTML parser issues which is how I stumbled upon this mess.
08:47
<Noam Rosenthal>
Yea there's a ton of mess in the parser. I am more familiar with it now after integrating it with all the new stuff (sanitizer/streamHTML)
08:48
<Noam Rosenthal>
Rebasing was actually quite easy
08:57
<Noam Rosenthal>
Though the bottom 2 PRs in the stack don't rely on anything special in the parser. It's just new calls to the fragment parser. Where it touches the parser internals is with streaming
09:55
<Noam Rosenthal>
freddy: we discussed this a few weeks ago and at WHATNOT
10:21
<Michael Reeves>
Dominic Farolino: just following up on this, no worries if you're busy!
10:25
<Noam Rosenthal>
I can address the "make insertion into a tuple" thing before, and also making the "fragment case" more normative. I think doing those would make the reviews of the upper two PRs easier (though they shouldn't affect the bottom two at all)
10:36
<annevk>
Thanks! And interesting, I would have expected it to impact the sanitizer integration.
10:38
<Noam Rosenthal>
yea it does affect that, that's the 3rd PR in the stack
10:39
<Noam Rosenthal>
The first two are mostly independent of parser stuff, but they touch the fragment parser initialization so I preferred to put them first. they're also much simpler
10:39
<annevk>
Oh I guess this shows my confusion with stacks.
10:40
<Noam Rosenthal>
yea it goes bottom up... I guess it's more like a queue
10:41
<Noam Rosenthal>
point is, the first two are quite independent, and this stack is built in a way where each PR can be submitted sequentially and we don't need to review everything at once
10:41
<annevk>
Noam Rosenthal Luke Warlow: so for innerHTML, an XML document, and Trusted Types, the story is that TrustedHTML is accepted and a string is not, right? Would we throw for this new type so we can make it work later?
10:42
<annevk>
Okay, I guess we should focus on Trusted Types for now then and hopefully freddy can review it soon. Or am I incorrect in assuming that has to land first as well?
10:43
<Luke Warlow>
Noam Rosenthal Luke Warlow: so for innerHTML, an XML document, and Trusted Types, the story is that TrustedHTML is accepted and a string is not, right? Would we throw for this new type so we can make it work later?
Yes TrustedHTML is accepted and a string is rejected, I think the idea is that this new type is rejected too (because sanitizer just doesn't work for XML)
10:43
<Noam Rosenthal>
it has to land first as well in this stack, but I could change the order if absolutely necessary (rather not)
10:49
<annevk>
Alright cool. Let's aim for that then.
10:50
<annevk>
Okay, yeah that makes sense to me. I doubt we'll ever put it in the effort, but who knows.
11:55
<zcorpan>
Noam Rosenthal: are the extra commits in https://github.com/web-platform-tests/wpt/pull/61748 expected?
12:27
<Noam Rosenthal>
Maybe it's not stacked correctly on top of the other ones. But there are a few new commits for streamHTML together with template for
13:46
<Noam Rosenthal>
I will check on Monday. Perhaps let's focus on the bottom of the stack first (trusted types, positional methods) and then I will re-align the rest once that lands? My initial goal with stacking is that reviewers will be able to see the whole picture but we can land them one at a time
14:17
<danae404>
Would someone be able to send me a link to the WHATNOT meeting today? Also, if there's time in the agenda, could someone add https://github.com/whatwg/html/pull/12735? (I think I don't have permission to add agenda+ myself)
14:41
<annevk>
Hey danae404, I added you to the calendar. I'll also fix your permissions in a second.
14:42
<danae404>
Thank you!
14:44
<annevk>
danae404: also, it seems it was already discussed two weeks ago and people are generally supportive based on the comments on the PR.
15:02
<danae404>
annevk: I wasn't sure if it needed further discussion regarding whether or not to restrict it to added tracks
15:45
<annevk>
zcorpan Noam Rosenthal: could you review https://github.com/whatwg/html/pull/12703?
16:11
<Noam Rosenthal>
Afk until Monday, will look then if no one beats me to it
16:57
<annevk>
Dominic Farolino: https://github.com/web-platform-tests/wpt/commit/ed37f83e79426194b075cfc29a5454894ae6f36f has the crashing test. https://github.com/whatwg/html/commit/57811ef0dca48568665d65eafd827764e9e454a2 is the HTML parser fix.
16:58
<annevk>
It's done, ta!