Re: [ppsp] re-charter issue discussion(cont'):transport mechanism

[Wesley Eddy <[e-mail]>]

On 4/10/2012 5:50 AM, Arno Bakker wrote:
> On 30/03/2012 10:53, zhangyunfei wrote:
>> Hi all,
>> To continue the re-charter discussion in PPSP WG meeting, I would like
>> to discuss more on the transport mechanism issue.
>> I happen to know in RTCWeb, the transport protocol between two peer
>> browsers seems to been decided: For media part, srtp/rtp over UDP(I am
>> not quite sure of if it is accurate). For data transmit, Datagrams over
>> SCTP over DTLS over UDP.
>> My question is: If this related to PPSP transport mechanism? Or totally
>> two different things?Any different requirements? Just to provide such
>> info to incur discussion.
> 
> Hi all
> 
> to keep the re-charter discussion brief could we perhaps split the
> discussion into (1) where media transport aspects should be discussed
> and (2) what the transport actually is.
> 
> For (1), I'd like to keep it inside the peer protocol spec, as mentioned
> before.
> 
> For (2): The RTCWEB work is definitely relevant, so thanks for keeping
> us up-to-date, Yunfei. There are of course differences, notably PPSP is
> also about time-shifted streaming, i.e. video-on-demand. This means
> real-time is not always relevant and we also need to think about
> post-view seeding. The latter makes LEDBAT support important.
> Furthermore, sender authentication and privacy may be less important for
> us than in the RTCWEB teleconferencing scenarios.
> Also PPSP is more about one-to-many than RTCWEB, although they may be
> looking at that too.
> 
> I think we should definitely advocate our work in RTCWEB, but in Canada
> they weren't ready for P2P scenarios yet, and I haven't followed their
> progress since.
> 


In my view, PPSP should not be distracted by RTCWEB or
go to any effort to mimic its decisions.  The methods
of operating are very different between the RTCWEB
browser-to-browser media flows between the involved
users and the PPSP method of using a tracker to find
peers in a swarm, and then doing chunk-based retrieval
of the media from the swarm rather than direct from
a source like RTCWEB does.

I am in agreement with Arno's point 1 that media
transport within the peer protocol is highly desirable.

-- 
Wes Eddy
MTI Systems