Bugzilla

https://github.com/whatwg/html/issues/5122 sketches out a plan that would eventually allow for origin isolation under certain circumstances.

I think we should make these changes to see if rolling them out as part of "resab" could be successful. There's little risk here as currently sites are not setting COOP+COEP and they would notice something would not work the moment they set them.

These are the changes:

* Rather than allowing SharedArrayBuffer and WebAssembly.Module same-site, we'd only allow messaging them same-origin.
* In a COOP+COEP (i.e., cross-origin isolation) environment, the document.domain setter returns early (as its first step).

Since this is a speculative change that might have to be rolled back I don't think we want to change web-platform-tests at this point, so we'll have to mark some of them as expected to fail. Google agrees with trying this out as well.

(This does not block enabling things on Nightly, it's strictly additive.)

https://github.com/whatwg/html/issues/5122 sketches out a plan that would eventually allow for origin isolation under certain circumstances.

I think we should make these changes to see if rolling them out as part of "resab" could be successful. There's little risk here as currently sites are not setting COOP+COEP and they would notice something would not work the moment they set them.

These are the changes:

* Rather than allowing SharedArrayBuffer same-site, we'd only allow messaging it same-origin.
* In a COOP+COEP (i.e., cross-origin isolation) environment, the document.domain setter returns early (as its first step).

Since this is a speculative change that might have to be rolled back I don't think we want to change web-platform-tests at this point, so we'll have to mark some of them as expected to fail. Google agrees with trying this out as well.

(This does not block enabling things on Nightly, it's strictly additive.)