Notes from our WebIDL + ES5 phone chat today...

I took a couple of quick notes during our chat. I know I missed a few folks who were present down in CA, so please forward on if necessary.

Allen and I both agreed that it was a worthwhile discussion. We also thought it would be useful to do this again next Monday (10/5-to avoid relative dates :-) before Cameron heads back to AU.

Thanks,
-Travis


...preamble stuff...
(the notes joined the meeting in progress)



AWB: Exemptions for host objects present from original ES3

... Large set of secondary semantics (even syntax) needed to support the DOM

... Creates a divide between what ES5 programs can do and what the DOM provides



TL: Is WebIDL complete in terms of its "special object semantics"?

CM: Yes, as driven  by requirements from HTML5 and assuming that HTML5 is largely fleshed out by now.



CM: WebIDL is a mix of de-facto browser behavior and some "experimental" ideas (mixin prototype)

DS: Do you indicate which things are strange/deprecated and shouldn't be used for new specs?

CM: It's been suggested that this be done.

AWB: Based on list discussions it sounds like there are some things that we could drop because they are not interoperable?



DS: Is it is possible to have multiple ES bindings for different implementations, or is a single binding that should be held as the standard the better approach?

CM: Is there buy-in to have such a thing? If not, it doesn't make much sense



AWB: At least in ES committee, we haven't taken a union of what's out there, but an intersection of that behavior. For conflicts where there is no interoperability there is an opportunity to agree on a 3rd solution.

... often easier to get agreement on the new solution (venders can implement the new one and still support the old behavior if necessary)



DS: Two high-level decisions we could make...

... Should decide-- does WebIDL specify everything ever used in an implementation, or more of an idealized slice of what "should be" implemented?



MM: ES5 has a strict mode. Some existing browser "tricks" could be opted out of ES5 strict.

... Could consider moving "oddities" or funny behaviors to a non-normative section (appendix) such that an implementer can be considered "conforming" without having to implement the oddities.

... High-level bit about "can you do it in JavaScript?". "caller" cannot be done natively in the language.



TL: topic: multiple inheritance

... discussion of current spec idea...



CM: users would like the ability to override a mixin in once place (e.g., EventTarget)

TL: IE8 spread the mixins out over the affected objects (duplicating their definitions)

TL: IE8 has no equivalent "EventTarget" object in JavaScript.



<left the meeting>