Re: [rtcweb] SRTP/AVPF/TCP in draft-ietf-rtcweb-transports-01

[Harald Alvestrand <[e-mail]>]

On 10/07/2013 10:21 AM, Magnus Westerlund wrote:
> On 2013-10-03 19:09, Parthasarathi R wrote:
>> Hi Harald/Magnus,
>>
>>  
>>
>> Thanks for the clarification.  Please clarify whether TCP/RTP/SAVPF will
>> be registered in IANA as separate MMUSIC specification or any another
>> simple mechanism exists to add IANA registry.
> I think we might need a MMUSIC or AVTCORE SDP field proto cleanup
> document. It is not only TCP/RTP/SAVPF that is missing, also AVPF and
> SAVP is missing.
>
> But, before diving into this. Are there really consensus that we will
> use IP/TCP/RFC4571 framing/RTP(SAVPF) rather than using TURN over TCP to
> get a stack saying IP/TCP/TURN/RTP(SAVPF)?

I have seen this thread suggesting two stacks:

- TURN over TCP to a TURN server, UDP to final destination
- RFC 4571 framing (I think) over TCP, all the way to the final destination

The latter would be done using ICE TCP candidates according to RFC
63something. I haven't yet seen one suggesting using the TURN framing
all the way to the final destination; I'm not sure I've seen a way of
signalling that either.

I think I've seen rough consensus that we should (SHOULD) support the
former. Ihaven't seen anything stronger than a MAY support the latter,
so for Transport, the only concern is that if we reference it, we
reference it correctly.


>
> Their SDP identification are quite different. The later uses m= lines
> indicating RTP/SAVPF with the TURN part in the ICE candidates.
>
>>  
>>
>> I’m interested in seeing the usage of DTLS key exchange for SRTP within
>> TCP/RTP/SAVPF profile. Even though DTLS is designed for connectionless
>> protocol, I assume that it works with TCP (connection oriented) protocol
>> as well. The optimized version of TCP/RTP/SAVPF shall use TLS as a SRTP
>> key exchange mechanism. Please correct me in case I’m missing something.
> In that case you definitely need an MMUSIC document. This as we need to
> figure out if using TLS to represent DTLS was a misstake as I think you
> need to separate DTLS and TLS when sent over TCP for the different
> purposes in that case.
>
> Cheers
>
> Magnus Westerlund
>
> ----------------------------------------------------------------------
> Multimedia Technologies, Ericsson Research EAB/TVM
> ----------------------------------------------------------------------
> Ericsson AB                | Phone  [phone]
> Färögatan 6                | Mobile [phone]
> SE-164 80 Stockholm, Sweden| mailto: [e-mail]
> ----------------------------------------------------------------------
>


-- 
Surveillance is pervasive. Go Dark.