Normative: Clarify text about host added properties #301
Conversation
spec.html
Outdated
| This specification does not recommend any specific addition. | ||
| In the Web embedding, HTML and WebIDL will specify which | ||
| interfaces are included. The Web Platform and Web-like | ||
| environments may decide to include `setTimeout`, `EventTarget`, |
There was a problem hiding this comment.
AFAIK setTimeout is IO. And as I explained in #284 I strongly against the host being able to add things into the realm without being able to have a clean version. This also hurts the API to be cross-platform safe.
There was a problem hiding this comment.
This is just an editorial note and we can remove the example.
Anyway, please keep the discussion to the respective issue.
There was a problem hiding this comment.
This PR is a clean way to describe the changes due to the feedback in #284 (comment) . We will need to follow up with details in the HTML side at some point, but I don't see why that should block Stage 3 (though it will block shipping in web browsers).
| platform capabilities with no authority to cause side effects such | ||
| as I/O or mutation of values that are shared across different | ||
| realms within the same host environment. |
There was a problem hiding this comment.
with no authority to cause side effects such as I/O or mutation of values that are shared across different realms
This is not enough for the SES requirement.
Imagine the host provide a counter { [[Get]]: () => #counter, [[Set]]: (val) => #counter = val } property on the Realm's global object, this provides a shared global state within the Realm (not shared across different realms). It provides a communication channel in the different compartments within the same realm.
There was a problem hiding this comment.
I guess I read the text to cover that case--I see the code sample as a value shared across realms, at a high level. I guess we just need to iterate on the wording so that it's maximally clear.
There was a problem hiding this comment.
If #counter is a host-defined pre-realm internal state, it will be a mutable state within the Realm and passes the current spec text but violate the SES requirement.
|
Something known that would block shipping in web browsers should of course block stage 3 - since the point of stage 3 is to ship it in web browsers. |
|
My guess is that a stage 3 would facilitate implementation without holding on what properties are added through HTML. This clears path for experimenting in this integration, while providing a fairly stable contract of a spec candidate from the ECMAScript side. |
I don't think this is how we've been doing things. We have frequently continued to iterate on HTML integration PRs after Stage 3 in recent proposals. test262 is also usually required for browsers to ship, but it's not a Stage 3 requirement. I think we should leave it up to the HTML folks whether they feel like things are complete enough on their end to go to Stage 3. |
Ref #284