Does removing an iframe make its current document stop being "active"?

[bzbarsky]

The relevant spec definition is:

At any time, one Document in each browsing context is designated the active document.

and https://html.spec.whatwg.org/multipage/browsers.html#history-traversal step 4 substep 3 explicitly sets a document as the active document of the browsing context.

When a frame is removed, we land in https://html.spec.whatwg.org/multipage/browsers.html#a-browsing-context-is-discarded which calls https://html.spec.whatwg.org/multipage/browsers.html#discard-a-document but none of that seems to unset the "active document" pointer from the browsing context.

I suspect, but it's worth testing what UAs actually do, that somewhere in here we should explicitly make the browsing context not have an active document or something...

[domenic]

Do you have suggestions for how to test this? i.e. an easy "is active" check?

[bzbarsky]

I don't have one offhand. XHR uses "fully active", not "active", in the places where it checks this stuff. The visibility API has conditions other than activity. Promise behavior in non-active situations is not really defined by any specs and probably not interoperable... We would have to look at occurrences of "active" in various specs to try to find one that uses this in a simple way. :(

[domenic]

@bzbarsky, can't we just test this by testing places that use "fully active"? If there are no nested iframes, then the two should be equivalent.

[bzbarsky]

That's the theory, in specs, but it may not match the practice in UAs.

That is, a UA could easily have "active" testing true, but "fully active" testing false, in the removed iframe case. And in particular, just because we determine that some APIs that tests "fully active" got false doesn't tell us much about APIs that should test "active".

[domenic]

OK... but I think it should be OK to assume the spec is correct, at least for the purposes of solving this bug and writing some tests. I don't think we should block on possible implementations where you can be fully active but not active.

[bzbarsky]

I think the problem is that we don't have good evidence that the spec matches reality, or would if it were implemented in web browsers. So I'm worried about us ending up accidentally specifying something that's not actually web-compatible. :(

[domenic]

Well, that's why we'd write tests, and only specify something that is compatible with tests that pass in at least a few browsers.