I'm not familiar with ReSpec. @marcoscaceres would be great for this if he has time.

👍 Thanks! I was mostly wondering whether it seemed reasonable to you to assign the [[supportedAlgorithms]] internal slot to the SubtleCrypto interface in WebIDL? And/or is there someone else I should ask about WebIDL stuff?

Yeah that's fine. Looking at the patch it is somewhat weird that when accessing the slot it's not made clear what instance it's being accessed on. Usually you'd see something like "this's [[slot]]" or "instance's [[slot]]".

Yeah, that's true. As originally envisioned, [[supportedAlgorithms]] was a free-standing internal object, so there was only one instance. By making it a property of SubtleCrypto, my intention was to keep that the same, but come to think of it, that doesn't actually really work, since there's one per Window (and implementations would technically have to run the steps to register an algorithm once per window, which seems unreasonable), so maybe this patch is not correct. Do you know of a reasonable way to declare a "global internal object"?

(Or perhaps this concept should be reworked entirely. I believe the original intention was to make the spec extensible by other specs, which could register additional algorithms on this object, but it's not actually necessary to have an object that can be extended at runtime, say. So perhaps there's a better way to express this spec-level / implementation-level "registry". But ok, if so that should probably be a separate PR.)

My general approach to registries is to "just say no" and require a PR with the appropriate sign-offs when an extension is needed.

It does seem good to work out what the scope of this is first, before tying it to a specific object.

OK, yeah. I tend to agree, I can get rid of it in a separate PR.

For now, I've made it a free-standing object again and renamed it from [[supportedAlgorithms]] to supportedAlgorithms, to make ReSpec happy, and also make it look less like it should be an internal slot of some other object. Hopefully this seems reasonable.

Happy to help, but give me day or two... better next week - but this is a good start.

This all seems good.

As a followup, we might want to add a .pr_preview.json file too.

👍 Thanks, will do!