Re: [webidl] Followup for bug https://www.w3.org/Bugs/Public/show_bug.cgi?id=23133 to to deal with inherited interfaces that are [NoInterfaceObject]. (#38)

> The only thing [NoInterfaceObject] does is remove the global property; the interface still exists.

Sorry, but no.  The spec says, at <http://heycam.github.io/webidl/#NoInterfaceObject>:

    If the [NoInterfaceObject] extended attribute appears on an interface, it indicates that an interface
    object will not exist for the interface in the ECMAScript binding.

And then in http://heycam.github.io/webidl/#interface-prototype-object it says:

    If the [NoInterfaceObject] extended attribute was not specified on the interface, then the interface
    prototype object MUST also have a property named “constructor” 

You can see this in Gecko:

    (new XMLHttpRequest).__proto__.__proto__.hasOwnProperty("constructor")

tests false.  Blink has a different behavior, but that's a bug in Blink per current spec.

> this breaks the correpsondence 

Not any more than if you had an object with no "constructor" property on a prototype chain in general...

We could change things around so that constructor objects are in fact required to exist but simply not be exposed on the global with [NoInterfaceObject].  That's not how things work right now, however.

Of course this whole business of having [NoInterfaceObject] stuff on proto chains is pretty broken to start with, imo.  The only case I where this happens (at least in Gecko's IDL) is XMLHttpRequestEventTarget, which isn't even implemented anywhere except Blink and Gecko as far as I can tell.  So maybe we can just drop the [NoInterfaceObject] from there and then disallow having [NoInterfaceObject] ancestor interfaces for something that has an interface object?  @annevk?

---
Reply to this email directly or view it on GitHub:
https://github.com/heycam/webidl/pull/38#issuecomment-69807414