Issue #151 - Make it possible to reject a promise from getHighEntropyValues

[miketaylr]

This partially addresses #151 by allowing UAs to reject the promise created from getHighEntropyValues. Once we have consensus on the rest, we can remove the issue and clarify what it means to "decide".

PTAL @amtunlimited

cc @othermaciej

---

Preview | Diff

[miketaylr]

gentle ping @amtunlimited ❤️ 🌈 🐰

[amtunlimited]

Sorry about the delay. Been a crazy...time?

Quick question: If I'm reading this correctly, does this mean that if one hint is rejected, the client does not get any hints?

[miketaylr]

Quick question: If I'm reading this correctly, does this mean that if one hint is rejected, the client does not get any hints?

Correct. You don't get any hints from getHighEntropyValues if the UA decides you can't have one. Right now there's no platform-notion of "partially resolve a promise", and it seems better to know up front that you didn't get some of the data you wanted then have to double check every item you did get back for a non-null result.

Perhaps a non-normative Note would be useful to encourage UAs to give some more context as to why they just sent a NotAllowedError?

[amtunlimited]

Perhaps a non-normative Note would be useful to encourage UAs to give some more context as to why they just sent a NotAllowedError?

IIRC we explicitly didn't do this for the headers (e.g. setting some sort of "not set" header). If I can find the relevant discussion I will link it here.

Is there a notion of "optional" in JS to represent not-present vs intentionally blank?

[miketaylr]

Is there a notion of "optional" in JS to represent not-present vs intentionally blank?

Yeah, that's what null represents (vs undefined). I thought about returning null for values that the UA wants to refuse, but I think getting a rejected promise for the whole lot is conceptually simpler.

[amtunlimited]

LGTM