[rtcweb] Clarification on simulcast and RID and RepairedRtpStreamId

[Iñaki Baz Castillo <[e-mail]>]

Hi,

We are almost in 2019 and, personally, I don't know yet which the
proper/standard way for simulcast + RTX is.

Clearly, RID must be used to send simulcast so, instead of signaling
a=ssrc lines in the offer, the offerer signals a=rid values for all
the simulcast streams. The remote learns the associated SSRC upon
receipt of the first RTP packet by matching its RTP RID extension
value. This is clearly specified in
https://tools.ietf.org/html/draft-ietf-avtext-rid-09 and Firefox does
properly implement it.

Now, the issue is when adding RTX to simulcast streams. Theoretically
(if I'm not wrong) the sender should signal both the a=rid of media
streams and the a=rrid of their associated RTX streams, and then send
RTP RID in media RTP packets and RTP RepairedRtpStreamId in RTX RTP
packets.

NOTE: Not sure about a=rrid, I can't find it in any draft. So:

Q1: Is there any spec defining the SDP syntax to indicate
RepairedRtpStreamId values in the SDP?

Q2: Is THIS (RID + RepairedRtpStreamId) the proper way to go for
simulcast + RTX?


Also, when it comes to receive media with RTX it's clear that WebRTC
does not define how to receive simulcast streams for WebRTC clients
(browsers), so browsers must receive a single stream plus an optional
RTX stream.

In this case, as far as a=mid plus RTP MID extension is used (or
a=ssrc) signaled, there is zero need for RID. However, how is RTX
supposed to be signaled in this case? Chrome expects ""a=ssrc-group:
FID MEDIA_SSRC RTX_SSRC" in the remote SDP to be able to receive an
additional RTX stream.

Q3: Is this the way to go for receiving a single stream with RTX?


-- 
Iñaki Baz Castillo
<[e-mail]>