08:42
<annevk>
TabAtkins: would it be easier if I don't update DOM while you convert it? I can probably wait another week or so
10:17
<annevk>
Does anyone happen to know for which HTML elements notification of end tag popping during parsing is important?
10:17
<annevk>
<script> is obviously one of them, are there any others?
10:19
<zcorpan>
</object>
10:20
<zcorpan>
</style>
10:20
<zcorpan>
</video> </audio>
10:23
<zcorpan>
i think that's it, but impls might have more. i think blink did something for </input>
10:24
<annevk>
there's no such thing in the spec
10:24
<zcorpan>
for </input> no
10:25
<annevk>
it's weird that the contract for the others isn't quite as explicit as it could be either
10:27
<zcorpan>
media elements tried to avoid it, but <track> screwed it over
10:27
<Ms2ger>
How about picture?
10:27
<Ms2ger>
link?
10:27
<zcorpan>
nope
10:28
<zcorpan>
not per spec at least
10:30
<zcorpan>
(picture doesn't do anything at all, it's the img that looks at its parent and previous siblings when it is inserted)
12:13
<gsnedders>
SimonSapin: should I be using tinycss2 or is there something better?
12:24
<SimonSapin>
gsnedders: I haven’t made a tinycss3, if that’s your question :)
12:26
<gsnedders>
SimonSapin: "is this what I should be using to parse CSS in Python?"
12:26
<SimonSapin>
yes
12:28
<SimonSapin>
gsnedders: let me know if you find something that could be improved or added
12:36
<gsnedders>
SimonSapin: Will do. Though I scarcely need more than parse_color. Some nice way to serialise stuff would be nice, though.
12:36
<gsnedders>
But rather out of scope, I suspect.
12:37
<SimonSapin>
gsnedders: https://pythonhosted.org/tinycss2/#tinycss2.ast.Node.serialize
12:38
<gsnedders>
SimonSapin: I was more thinking about serialising color3.RGBA
12:38
<SimonSapin>
though the RGBA type is not a Node
12:38
<gsnedders>
Exactly.
12:38
<gsnedders>
Also one can bikeshed format endlessly :)
12:39
<gsnedders>
A dict of (str) keyword -> RGBA would be nice, too.
12:39
<SimonSapin>
I don’t think it needs to be more complicated than a string interpolation one-liner
12:40
<gsnedders>
Oh, it's not that complex. The only fun thing is multiplying it so you have either a 0–255 int or percentage
12:40
<zcorpan>
serializing the alpha isn't obvious (see www-style)
12:40
<SimonSapin>
that dict is tinycss2.color3._COLOR_KEYWORDS
12:41
<SimonSapin>
https://github.com/SimonSapin/tinycss2/blob/a64d0a64057fa3b73146b8bc7759a882d53550ed/tinycss2/color3.py#L180
12:41
<gsnedders>
Ah, cheers. I hadn't bothered digging into that yet.
12:41
<gsnedders>
zcorpan: is there a tl;dr of that?
12:42
<SimonSapin>
it’s undocumented, but unlikely to change
12:42
<gsnedders>
This is a script I'm likely not gonna touch after today, so not really an issue.
12:55
<zcorpan>
gsnedders: how do you roundtrip rgba(1,1,1,0.5) when storing alpha in 8 bits?
12:56
<gsnedders>
zcorpan: Ah, but tinycss2 stores it as a double. Then you just go for the shortest unique representation.
12:57
<zcorpan>
gsnedders: ah ok
15:22
<ondras>
when using registerElement with the "extends" feature, one has to use is= to instantiate? <x-stuff> cannot be used then?
15:23
<caitp>
that doesn't sound right
15:24
<ondras>
I vaguely recall asking about this some time ago, receiving response approximately compatible with what I wrote
15:24
<ondras>
but I am not sure
15:24
<ondras>
(and my poor english might be also the cause)
15:25
<caitp>
is= was supposed to be, if I understand, a way to work around the dumb aspects of the parser, and a way to ensure hard-to-polyfill features work well (although I'm not positive about that last one)
15:46
<annevk>
Domenic: did you get any further writing up how ARIA bindings would work?
15:46
<annevk>
Domenic: I'm wondering if we can reuse that to explain title="", contenteditable="", etc.
16:11
<Domenic>
annevk: no further than https://github.com/domenic/html-as-custom-elements/blob/master/docs/accessibility.md. I can turn the strawmen there into something more concrete maybe, if there is implementer interest.
16:11
<Domenic>
Why does title="" need explaining?
16:11
<annevk>
Domenic: e.g. you might want to overwrite the default behavior
16:12
<Domenic>
Mehhh just use a different attribute then
16:12
<Domenic>
If I were to say "explain title" I would mean "add `window.showOSTooltip("string", { x, y })"
16:13
<Domenic>
then I could explain title as some sort of global mutation observer + mouseover hooks
16:13
<Domenic>
+ I think it impacts accessible name calculation
16:14
<Domenic>
but that is already explained at the bottom of the above link
16:14
<annevk>
Domenic: see https://lists.w3.org/Archives/Public/public-webapps/2014AprJun/thread.html#msg324
16:14
<annevk>
Domenic: if you want to allow <my-element someattr> we need to account for the fact that some day "someattr" might be a global attribute
16:14
<Domenic>
Sure. How is that related?
16:15
<annevk>
Domenic: it's the same scenario, except "title" is already a global attribute
16:15
<Domenic>
You can already use <div someattr>; how are we accounting for that?
16:16
<annevk>
We tell people to please use data-* or they might have a problem going forward
16:16
<Domenic>
I still don't understand the problem. Maybe jump to a strawman solution and tell me what new capabilities I will have when it is implemented?
16:18
<annevk>
The problem is clashes and I don't really have a good solution
16:19
<Domenic>
Well, what new capabilities will I have if we solve the clashes solution?
16:21
<annevk>
You can use any attribute name (including builtins) without future repercussions
16:22
<Domenic>
I guess I pretty much agree with Hixie's arguments in that thread that this isn't a big deal. But if you wanted to solve it maybe the custom element definition should blacklist or whitelist attributes.
16:22
<annevk>
Domenic: see. e.g. https://lists.w3.org/Archives/Public/public-webapps/2014AprJun/0440.html from that thread
16:23
<Domenic>
E.g. registerElement(..., { ..., attrs: ['title', 'somefutureattr'] }) => 'title' and 'somefutureattr' won't get special treatment from the browser
16:23
<annevk>
Yeah maybe
16:23
<Domenic>
https://lists.w3.org/Archives/Public/public-webapps/2014AprJun/0440.html doesn't make much sense... you can use setters .osc2 for the attribute data-osc2
16:50
<annevk>
true
17:54
<annevk>
So annoying that GitHub doesn't search through commit messages
17:54
<annevk>
Does anyone know what resulted in IDL adding back the [[HasInstance]] override?
17:54
<Ms2ger>
Maybe bz
17:55
<hasather_>
annevk: search locally?
18:15
<TabAtkins>
annevk: That would be a little easier, yeah, but not strictly necessary. But I plan to finish up this weekend, so if you're okay with waiting a little bit, that'd be cool.
18:16
<annevk>
TabAtkins: sure thing
22:41
<Domenic>
"Don't expose GC behavior" strikes again! https://github.com/yutakahirano/fetch-with-streams/issues/15#issuecomment-72119046