01:36
<Alice>

Just to confirm: AOM192 is a non-starter in terms of the proposal to have open and closed shadow roots behave differently, right? https://github.com/WICG/aom/issues/192#issuecomment-1132237093

I know there were similar concerns around getInnerHTML (DSD) and elementtiming:

Today in AOM though we discussed maybe having an API where you can pass in an array of shadow roots. If the API doesn't distinguish between closed and open shadow roots, maybe this has legs?

I wrote this up as an issue: https://github.com/WICG/aom/issues/195
07:13
<annevk>
nolanlawson: yeah maybe, although when I see Alice's example there it doesn't seem like the <input> element should have to know about the details of the component; so either it should delegate the actual pointing to the component or something else needs to change. It doesn't seem like a good solution for that case would be that the <input> element has to know about the internals.
07:18
<Alice>
I don't think it does need to know about the internals. The relationship could be set by the optionlist. But also, the example is just that - an example.
07:20
<Alice>
If your point is that there's literally no valid use case for wanting to refer into deeper shadow roots, that's a separate question.
08:22
<annevk>
Alice: that would be a rather strong assertion, but let's just say that I haven't seen a compelling example thus far
08:23
<annevk>
(otoh, if <input> was a custom element instead my draft PR against HTML would allow for such references)
08:24
<annevk>
(but it doesn't address serialization and if we need to address serialization that would be a lot harder for arbitrary shadow roots, so perhaps it shouldn't allow arbitrary shadow roots until we figure that out)
08:28
<annevk>
Domenic: did you see my HTML PRs for reflect? They're all somewhat trivial
08:29
<Domenic>
annevk: yeah, sorry, don't think I'm going to get to them today, but will try to prioritize Monday
08:54
<annevk>
Domenic: all good, just making sure they're not lost in the noise
15:06
<Ms2ger>
Does anyone else have issues with bikeshed update failing to download some files?
18:24
<nolanlawson>
Alice: that would be a rather strong assertion, but let's just say that I haven't seen a compelling example thus far

I tried to add some examples. I know Westbrook came up with the "recycler" example – he might have more. https://github.com/WICG/aom/issues/195#issuecomment-1444184540

In general, the strongest argument from my POV is developer ergonomics. Developers like shadow DOM's style encapsulation and slotting, and when given the tool, they'll use it to create very granular components that fit their organizational or technical needs. Folks do this with React/Vue/Svelte/et al as well, but in that world, they can still wire up elements using ARIA attributes, whereas with native web components, they cannot.