Kai Ninomiya

> We'll try to solicit some feedback about the options before the meeting. Feedback we got was that we should just enqueue a microtask from inside `getCurrentTexture()`. This has deterministic...

As for video frames, we don't want the possibility of video frames expiring between rVFC callbacks and rAF callbacks, so probably don't want to change the behavior we have now....

Agreed let's defer on the topic of unobserved canvases for the moment. I think that is a reasonably separable discussion. --- > There is no browser-side rendering that ever happens...

Can you point to a real world sample, or at least provide a concrete example, of microtasks being used in this way? It's certainly possible, but I haven't seen it,...

This would be an "explicit present" which we probably last discussed several years ago at one of the F2F meetings (I think San Diego 2019?). We had decided our swap...

Refcounting is a less weird way of describing it. Makes it sound better! Brandon and I discussed, and support the refcounting-microtask option for V1. We'd like to later design an...

Thanks @tidoust for debugging and filing the Bikeshed issue! This reminds me, is there a way to get spec-prod to use a pinned version of Bikeshed? It would be convenient...

> I note disabling HTML validation is not going to work for publication to /TR because the publication rules checker will refuse to publish a document with invalid HTML content...

This is great, thank you! We'll definitely be considering this feature again after V1 is complete.

Notably it didn't seem to make sense to have a generic "internal" error as well as an "out-of-memory" error. It would make more sense to have "pipeline-failure" and "out-of-memory" or...