Re: [MMUSIC] RTSP & NAT: Symmetric RTP

[Magnus Westerlund <magnus.westerlund@era.ericsson.se>]

Hi David,

David Yon wrote:

> Using SSRC was mentioned a long time ago, although at the time 
> dismissed as too protocol-specific.  But the context of the discussion 
> has changed, and if that works then I have no problem with it in 
> principle.  That said, there was at least one other concern voiced 
> about whether 32 bits is sufficient.  It also begs the question of 
> what format to use for an endpoint-ID, and whether this format is to 
> be fixed or variable.  One approach would be to make the endpoint-ID 
> format specific on a per-protocol basis, with a limited number of 
> choices (it seems like string and integer would cover most needs).

My feeling is that 32 bits are enough for this. Because a brute force 
attack is not possible without the receiving end-point notice it. If 
good random numbers are used then it seem difficult enough to guess 
within a few message the correct one out of the 4.2 billion ones possible.

> I'm not sure I followed you exactly, but I read your first paragraph 
> to mean that using a windowed protocol approach for endpoint ID is 
> even more of a problem in the context of RTCP.  Is that a fair 
> interpretation?
Yes, the problem as I see it to safely demultiplex between ordinary 
traffic and the endpoint ID message.

Best Regards 
Magnus Westerlund 
Multimedia Technologies, Ericsson Research ERA/TVA/A
----------------------------------------------------------------------
Ericsson AB                | Phone +46 8 4048287
Torshamsgatan 23           | Fax   +46 8 7575550
S-164 80 Stockholm, Sweden | mailto: magnus.westerlund@era.ericsson.se



_______________________________________________
mmusic mailing list
mmusic@ietf.org
https://www1.ietf.org/mailman/listinfo/mmusic