Shouldn't the Window interface be marked as [ImplicitThis]?

[cdumez]

https://html.spec.whatwg.org/multipage/browsers.html#the-window-object

[PrimaryGlobal]
/sealed/ interface Window :​ EventTarget {

Why isn't the Window interface marked as [ImplicitThis]?

The latest Web IDL specifies [ImplicitThis] and indicates that it is used by the HTML specification for the Window interface:
http://heycam.github.io/webidl/#ImplicitThis

[cdumez]

@bzbarsky

[bzbarsky]

I'm pretty sure there are existing issues on this (I certainly recall discussion in them), but basically the whole implicitthis thing is silly as defined right now. It would need to be defined on Window, and also on EventTarget, and various interfaces that Window implements (via "implements" statements) and all the interfaces that are on the ancestor chain of the global in workers and anything implemented by any of those.

I think the last time this was discussed the conclusion was basically that the [ImplicitThis] behavior should be the only behavior in IDL. Then for operations/attributes for which it makes sense (i.e. the ones that expect to get the global as this) it would do the right thing, and for other ones you would get an exception, just like you get an exception for undefined if none of this [ImplicitThis] stuff is happenign (though UAs should feel free to make the exception as informative as they like, of course).

@domenic, does that correctly reflect the current state of thinking on this stuff? I can't remember where the discussion about this was....

[domenic]

Hmm, found https://www.w3.org/Bugs/Public/show_bug.cgi?id=18547 but that's pretty old... Ahah, it was at https://lists.w3.org/Archives/Public/public-script-coord/2015JanMar/0109.html, which has some good discussion

It seems in https://lists.w3.org/Archives/Public/public-script-coord/2015JanMar/0122.html we settled on the plan you mentioned, that [ImplicitThis] should be the only behavior in IDL.

Happy to keep this bug open as a tracking bug on HTML, but yeah, the work to do here is in IDL. I will update https://www.w3.org/Bugs/Public/show_bug.cgi?id=18547 with the plan of record so we can find it later.

[annevk]

What is /* sealed */ referring to? When can we remove that? (I did a bunch of digging, but nothing turned up.)

[annevk]

Ah, I forgot to dig in my inbox: "it means i wish that people couldn't subclass that class". We should probably turn that into a note of sorts.

[bzbarsky]

I don't think people can meaningfully subclass Window anyway, since it has no constructor, right?

[annevk]

I think this would mostly about someone creating a new environment for a worker or window/document and using subclasses there of what we have today. Not about developer-controlled subclasses necessarily. Only the latter might some day work for Document.

[bzbarsky]

Ah, I see. So this is about specs not subclassing window, ok.

Developer-controlled subclasses of Document totally work in Gecko, for what it's worth.