RTPReceiver encoding parameters are useless

[murillo128]

Correct me if I am wrong.

On chapter 8.3 RTP matching rules the rtp packets are routed to the correct RTPReceiver based on the following parameters:

• parameters.muxId
• parameters.encodings[i].ssrc
• parameters.encodings[i].fec.ssrc
• parameters.encodings[i].fec.ssrc
• parameters.codecs[j].payloadType
• parameters.encodings[i].rtx.payloadType

First weird case, we check the payloadType on the codecs, but the rtx payloadType on the encodings (???). Luckily this will be solved by #444

We only use encodings for looking for the ssrc, but we then don't do anything with it (is it relevant at all that a packet is received by a particular encoding? i don't think so, we know the ssrc and the codec already).

So my question is what are encodings really used for in receving side? We just need a sequence of <ssrc,fecssrc> at most.

Moreover, as discussed in #439, if we just allow one incoming stream for simulcast, we just need one ssrc and one fecssrc, don't we?

So if we do #440 we could simplify to:

dictionary RTCRtpReceivingParameters {
             DOMString                                 muxId = "";
             sequence<RTCRtpCodecParameters>           codecs;
             sequence<RTCRtpHeaderExtensionParameters> headerExtensions;
             unsigned long                             ssrc;
             unsigned long                             fecSsrc;
             RTCRtcpParameters                         rtcp;
};

As usual, am I missing anything?

[ibc]

We only use encodings for looking for the ssrc, but we then don't do anything with it (is it relevant at all that a packet is received by a particular encoding? i don't think so, we know the ssrc and the codec already).

The "RTP matching rules" section is a bit special in the sense that it is not intended for API users, but for ORTC implementors (browser vendors). It helps anyhow but it does not provide complete guidelines for the implementor.

if we just allow one incoming stream for simulcast, we just need one ssrc and one fecssrc, don't we?

Agreed. Also note that the SSRC for received RTX payloads should also be unique (not sure why, but in the real world it is unique). So I would add a rtxSsrc to that dictionary.

[murillo128]

@ibc matching rules actually defines what input parameters are required in order to work. We should not request the API user to fill more parameters that the ones strictly required.

[aboba]

A table has been added, indicating which attributes within RTCRtpEncodingParameters relate to the RtpSender, and which relate to the RtpReceiver. Is this sufficient?

[robin-raymond]

@aboba I believe it is what was needed. Anything further requires a proposed redesign which is beyond the scope of this bug and the original statement about encodings being useless in the title of the bug is inaccurate.

I filed a separate bug to clarify when encodings are not specified to describe exactly the behaviour expected. As far as I'm concerned, this bug is closed.