On Thu Feb 18 17:08:18 PST 2021 Ryosuke Niwa via webkit-dev <
webkit-dev at
lists.webkit.org> wrote:
> The latest proposal does solve much of the problems we've identified in the
> past, and it's looking to be a promising direction. Do you have any example
> website or app that you can share with which resulted in the observed 15%
> improvement in FCP? That would be a very intriguing observation and would
> substantiate the support for this feature.
Glad to hear you think it's promising! The example is from the AMP team,
and you can see more detail at the Origin Trial results summary here:
https://docs.google.com/document/d/1QmBKxQLE81PsMzyPCvYhzqRAn4hxz61Jzk0uXJAhaVc/edit> However, as I commented on
https://github.com/whatwg/dom/pull/892 and have
> previously stated during the last F2F and other avenues, the currently
> proposed semantics of getInnerHTML is deeply problematic. We want
> consistent semantics across different kinds of Web API, and what's being
> proposed is very much different from what we had discussed what we would do
> for selection.
I just replied to the comment you made yesterday on DOM issue #892. I looked
back at the F2F minutes (
https://www.w3.org/2020/03/23-components-minutes.html)
to refresh my memory, and it turns out the current API was actually suggested
(by annevk) at that meeting. I don't see any comments from you about the API
shape there, but perhaps I missed it?
In terms of API shape, can you clarify what you feel is deeply problematic
about it? I assume you're talking about the need to pass in closed shadow
roots, to enable them to be serialized. I see that your comment there suggests
the requirement to pass in *all* shadow roots, even open ones. But the linked
comment on selection actually refers to just closed shadow roots also. I'm
unclear why you would want/need to pass in open shadow roots?
> Also, have people figured out how scoped custom element registries can
> integrate with this feature in the future? Given that's the other most
> frequently requested feature, it would be very regrettable if we later
> found out some inconsistencies or awkwardness in their integrations.
Yes, I've talked several times to justinfagnani about Scoped Custom Element
Registries, and he sees no problem integrating with DSD. We'll likely
just need to add an attribute that opts the shadow root out of using the
global registry, but that seems straightforward. Do you have any particular
concerns?
Thanks,
Mason