Should send() throw if the XHR object's window's document is not fully active?

[bzbarsky]

We seem to have a fully active check in open() but not send(). That doesn't make much sense to me...

[annevk]

Does having the check help at all? Most other APIs don't have it...

[bzbarsky]

That's the other question, yes. @smaug---- might know why we do these checks in Gecko...

[smaug----]

The initial reason in Gecko was https://bugzilla.mozilla.org/show_bug.cgi?id=403168
but the setup we have for wrappers and such is quite different these days.

However, what should happen to the network connection if XHR is created using non-active document, like if the response is super slow, what kills it? I assume browsers normally cancel relevant network connections when unloading a page, but if the page has been unloaded already...

[annevk]

@smaug---- not sure, maybe the connection is not established until such a time that the document does become active...

[annevk]

Tests that other browsers not pass because of this:

• XMLHttpRequest/open-url-multi-window-2.htm
• XMLHttpRequest/open-url-multi-window-3.htm
• XMLHttpRequest/open-url-multi-window-4.htm
• XMLHttpRequest/open-url-multi-window-5.htm

Also XMLHttpRequest/open-url-multi-window-6.htm seems to not pass anywhere. I think based on this and the somewhat flawed security reason to add this initially, we should just remove this.

[annevk]

Hmm, it's not that simple, Chrome will throw on send() instead if the document is "detached" (whatever that means). Safari times out.

[annevk]

I think the main reason the check is in open() is because we need the document's base URL for open(). I wonder what happens if you pass a "detached" document to send() though.

[domenic]

Ran into this in jsdom as well, trying to implement the spec but failing opoen-url-multi-window-3.htm because we didn't include some kind of detached check.