Re: Reducing normative dependencies

I think there are two concerns here.

- We wish to be able to add new stats as we need them, and we want to do
so without going to the rather onerous process step of opening up the
webrtc-pc document - we foresee that we want to add stats while the
webrtc-pc document is in states where we don't want to touch it. This
was the original motivation for having the stats document be a separate
document.

- We want to be able to point implementors to a single document and say
"implement that and its dependencies", and have the result be a fully
working system - and a fully working system needs to implement at least
the stats API and a core set of stats.

The last point actually indicates that we would WANT to have a normative
dependency from PC to stats - I remember Cullen arguing this point in
our Washington interim meeting.

(it also points out that we want to have some declaration in the stats
document that "the following are the required stats for level X of
implementation", so that the requirement for a conformant system doesn't
change every time we add another stats value. Additional problem.)

But there may be other ways to skin the cat.






On 12/05/2014 01:32 AM, Cullen Jennings (fluffy) wrote:
>> On Dec 4, 2014, at 1:36 PM, Martin Thomson <martin.thomson@gmail.com> wrote:
>>
>> On 4 December 2014 at 08:25, Cullen Jennings (fluffy) <fluffy@cisco.com> wrote:
>>> I don't want to be adding new normative dependencies to the PC spec at this point if we can avoid that. Previously we had agreed to keep the structure for the stats in PC spec so that other specs could define new stats but not have PC spec normatively depend on the all the specs that defined stats. I think this is the right decision and we should stick with it. Let me know if there is a good reason not to.
>>
>> If this is in relation to https://github.com/w3c/webrtc-pc/pull/37,
>> then I think that I agree.  If we move stats out, the dependency graph
>> needs to be: stats -> pc, and not pc <-> stats.  The former de-risks
>> pc, which is a good thing.
>>
>>
> No,exactly the opposite. PR 37 makes PC normatively depend on stats which it did not do before. Right now stats depends on PC. That is fine with me. However PR 37 makes PC normatively depend on stats which has many problems including one has to implement all of stats to be compliant with on PC. 
>
>
>
>
>


-- 
Surveillance is pervasive. Go Dark.

Received on Friday, 5 December 2014 14:34:01 UTC