Re: WebIDL: how to address the various audiences and constraints?

On Sep 30, 2009, at 3:40 AM, Robin Berjon wrote:

> On Sep 30, 2009, at 12:13 , Maciej Stachowiak wrote:
>> On Sep 30, 2009, at 2:43 AM, Robin Berjon wrote:
>> On Sep 29, 2009, at 21:20 , Cameron McCormack wrote:
>>>> As for having a simplified version first including only what’s  
>>>> needed
>>>> for those specs that need Web IDL done quickly, maybe.  HTML5 is  
>>>> by far
>>>> the biggest user of the esoteric ECMAScript features.  I guess I  
>>>> would
>>>> like to know, for the authors of dependent specs, how quickly  
>>>> they need
>>>> Web IDL done.
>>>
>>> WebApps has a document in LC that depends on it (Widgets 1.0: The  
>>> widget interface), and it's a really trivial document to test — we  
>>> don't expect it to be long before we can transition, but it is  
>>> blocking on its dependency on WebIDL. I'd say it's at most one  
>>> month before its progress is hampered by process alone.
>>>
>>> It's too early to tell but DAP has some low-hanging fruits that I  
>>> would expect it to be possible to make quick progress on (famous  
>>> last words — I know). Here we're looking at a 3-6 months window.
>>
>> As I understand W3C Process, a spec can enter CR with a dependency  
>> that is still a Working Draft. What is not allowed is for a  
>> document to *exit* CR and transition to PR or REC while depending  
>> on a Working Draft - all dependencies at that point must be CR  
>> maturity or higher. In the timelines above are you referring to  
>> completing Last Call and entering CR, or are you talking about a  
>> timeline to complete CR and enter PR? I believe only the latter  
>> would be blocked by a delay in Web IDL.
>
> To the best of my knowledge you won't find anything in Process  
> stating what maturity levels you can reference; it's a PubRules  
> thing. PubRules says:
>
>  - "In general, documents do not advance to Recommendation with  
> normative references to W3C specifications that are not yet  
> Recommendations."
>  - To transition to PR you should check that you're only referencing  
> PR+ specifications.
>  - To transition to CR you should check that you're only referencing  
> PR+ specifications.

It's a requirement to reference only PR+ specifications to enter CR?  
That doesn't sound right. It would make it very hard to ever get to CR  
with a spec that has a significant dependency chain of new specs, and  
would make mutual references completely impossible. I can find  
historical counter-examples:

DOM 3 LS entered CR on Nov 7, 2003, and it referenced DOM 3 Core which  
at the time was a WD and entered CR on the same day. Selectors Level 3  
entered CR in November 2001 (it later went back to WD) and it cited  
multiple Working Drafts normatively.

>  - To transition to LC I couldn't find anything.
>
> It's quite possible that other rules are documented elsewhere —  
> these are the ones I found. I couldn't find the rule that is passed  
> from mouth to ear about the ability to reference no more than two  
> maturity steps below (which you seem to be quoting).

I couldn't find it either.

> So based on the above it looks like you can't touch CR if you depend  
> on something that isn't in PR. Note however that this isn't listed  
> as a strict rule — it's phrased as sort of the recommended behaviour.
>
> I'm fine with the two-steps rule but it'd be nice to have a pointer  
> or agreement from the Team on that.

We should probably seek input from the Team on the actual rules on this.

Regards,
Maciej

Received on Wednesday, 30 September 2009 11:45:41 UTC