Re: [MMUSIC] BUNDLE TEXT: De-mux procedures (June 19th)

[Paul Kyzivat <[e-mail]>]

On 6/19/13 9:37 AM, Bernard Aboba wrote:
> Harald said:
>
>  > Absolutely, I'm saying that BUNDLE should not specify a procedure for
>  > de-muxing. Instead, it should point to the specs that already specify
> how the
>  > de-muxing is done.
>
> [BA] RTP/RTCP multiplexing is covered in RFC 5761, and de-multiplexing
> of RTP and DTLS is covered in RFC 5764, so there is no need for BUNDLE
> to create new text on those aspects.

I'm replying here, rather than to Harald because both of you say more or 
less the same thing and its more concise here.

IIUC, RFC 5764 says how to demux DTLS from DTLS/SRTP, but not how to 
demux DTLS from other RTP. I think it turns out that the same mechanism 
works for both, but that is by accident rather than specification.

I don't know if/when someone would want to use vanilla RTP with DTLS, 
but somebody will probably find a reason. (Not RTCWEB.) It will take 
more than a reference to 5764 to cover it.

Also, 5764 only covers a single m-line using DTLS payload packets. That 
also needs to be addressed explicitly.

> However, there is still the open
> question of whether payload type can be required to be unique on all
> bundled m lines.  If so, then PT demultiplexing is viable; if not, then
> a combination of PT and SSRC de-muliplexing will be needed.

I agree.

	Thanks,
	Paul