Re: [MMUSIC] a=rtcp-mux-exclusive vs a=rtcp:0 compression

[Roman Shpount <[e-mail]>]

On Thu, Apr 7, 2016 at 10:56 AM, Magnus Westerlund <
[e-mail]> wrote:

> Den 2016-04-05 kl. 18:36, skrev Roman Shpount:
>
>> I have modified the proposed SDP slightly. I have removed IN IP4 0.0.0.0
>> from a=rtcp attribute. I cannot find a place that states that a=rtcp is
>> required when rtcp-mux is used, so I removed a=rtcp from the
>> rtcp-mux-exclusive examples.
>>
>
> So, I would agree that a=rtcp is not mandatory based on RFC 5761, it do
> results in that the fall back for RTCP is m= port+1. But, as this is
> intended for the case of not having any fallback.
>

 I assumed since we are doing a new attribute that specifies that there is
no fallback, then a=rtcp attribute can be omitted.

However JSEP do contain text that appears to require that a=rtcp are
> included. This, may be changed to support the mux-exclusive usage. However,
> I haven't checked if this requirement is due to trickle ICE.
>

JSEP includes requirement for a=rtcp due to
https://tools.ietf.org/html/rfc5245#section-4.3 :

If the agent is utilizing RTCP, it MUST encode the RTCP candidate using the
a=rtcp attribute as defined in RFC 3605  [RFC3605].


This was already fixed in
https://tools.ietf.org/html/draft-ietf-mmusic-ice-sip-sdp-08#section-4.2 :

If the agent is utilizing RTCP and if RTCP candidate is present and is not
equal to the same address and the next higher port number of the RTP
candidate, the agent MUST encode the RTCP candidate using the a=rtcp
attribute as defined in RFC 3605 [RFC3605].

_____________
Roman Shpount