Re: [rtcweb] DTLS, DTLS-SRTP, and 5-tuples

[Paul Kyzivat <[e-mail]>]

On 3/9/15 9:17 AM, Simon Perreault wrote:
> Le 2015-03-09 08:54, Christer Holmberg a écrit :
>> DTLS itself knows nothing about ICE.
>>
>> So, IF DTLS assumes (implicitly or explicitly) that a single 5-tuple is used, the appropriate WG at least need to be consulted about whether usage of multiple 5-tuples will cause any issues - technical or security.

ISTM that this subject is closely related to multipath transports - TCP, 
RTP, UDP, SCTP. ICE provides a way to negotiate the multiple paths.

Looking around, work on multipath transports is spread around. For TCP 
and SCTP it was done in transport. There is a draft for multipath RTP in 
avtcore. I see a little email discussion of multipath UDP in transport, 
but (based on my one minute spent googling) not any real action.

To date ICE hasn't considered this to be a general multipath transport 
problem - just a way to negotiate a single path transport. But the 
discussion here is showing flaws in that approach. But 
draft-wing-mmusic-ice-mobility was closer to really dealing with multipath.

This may need a tactical solution to keep rtcweb going in a timely way 
and a more strategic approach to get this all properly cleaned up.

	Thanks,
	Paul