[js-api] Review object caching

[Ms2ger]

See the discussion in WebAssembly/threads#118

[Ms2ger]

The suggestion there seems to be that the spec requires that postMessage can produce an existing object in its deserialization. However, the HTML spec doesn't support that:

Set value to a new instance of the interface identified by interfaceName, created in targetRealm.

and the prose about Module agrees with that. (@binji added a test for this in web-platform-tests/wpt#12488.)

In the existing spec, it seems like we only need to remove the following sentence from a note in js-api:

This mapping is used to ensure that, for a given agent, there exists at most one JavaScript object for a particular WebAssembly address.

because it's not true.

The threads proposal will also need to be updated to take that into account.

[binji]

OK, this sounds good to me. Thanks for looking into it.

[littledan]

The analysis looks right to me; apologies for the confusion. @Ms2ger would you be interested in creating spec PRs and tests for this change?

[binji]

@Ms2ger ping

[Ms2ger]

Wrote some extra tests at web-platform-tests/wpt#18590.