Re: Proposed Charter Changes

On Wed, Apr 29, 2015 at 9:55 AM, tim panton <thp@westhawk.co.uk> wrote:

>
> On 29 Apr 2015, at 18:17, Justin Uberti <juberti@google.com> wrote:
>
>
>
> On Wed, Apr 29, 2015 at 6:49 AM, Eric Rescorla <ekr@rtfm.com> wrote:
>
>>
>>
>> On Tue, Apr 28, 2015 at 11:50 PM, tim panton <thp@westhawk.co.uk> wrote:
>>
>>>
>>> > On 28 Apr 2015, at 16:28, Cullen Jennings <fluffy@cisco.com> wrote:
>>> >
>>> >
>>> > I put a diff at
>>> >
>>> > https://github.com/fluffy/webrtc-charter/compare/gh-pages...fluffy:ekr
>>> >
>>> > I like the text you put in this because I think it reflects the
>>> relativity of what the WG intends to do.
>>>
>>>
>>> I have a couple of serious problems with this text.
>>> Firstly I don’t like the characterzation of low-level and high-level
>>> APIs.
>>> I substituted 'object-orientated' and 'declarative SDP’ - (I considered
>>> ‘opaque SDP blob’)
>>>
>>> The reality of the current API is that everyone has to mess with the SDP
>>> to get what they
>>> want, and frankly there is nothing lower level than regexps on SDP.
>>>
>>
>> I have no strong opinion on this phrasing.
>>
>
> I don't care for the new terminology.
>
>
> Honestly I’m not wedded to the words either, but ‘low' and ‘high' make a
> set of
> assumptions about implementations that we probably shouldn’t be making in
> at this stage.
>
>
> PeerConnection can be implemented atop the ORTC objects. Ergo, ORTC is a
> lower-level API.
>
>
> Um, I’d heard that ORTC could also be implemented on top of PeerConnection.
> (although why is a whole other question).
>
>
> Clumsiness in an API doesn't affect its position in the stack.
>
>
> True, but is this a stack? Or are these are side-by-side APIs that drive
> the really low level stuff
> like crypto/audio/video/networking? We shouldn’t be presupposing that it
> _is_ a stack
> in a charter unless we are absolutely certain that it _has_to_be_ . I’m
> (as you can tell)
> unconvinced.
>
>
I am pretty sure that it is. PeerConnection is the session negotiation
layer atop the underlying RTCRtp* and Transport objects. The lower layer
pieces have no negotiation facility.

This makes the distinction between PeerConnection and the lower-layer
objects easy to reason about.