RE: Early and late rogue RTP packets

["Dan Wing" <dwing@xxxxxxxxx>]

I haven't seen anything written up on such a thing, and I expect the
different definitions are because of PSTN gateways and other devices (SBCs?)
which re-use UDP ports too quickly; the UDP ports need to remain unused for
a while for all the same reasons TCP ports need to remain unused for a while
(TIME_WAIT).  The existing values for TCP's TIME_WAIT are probably too long
for today's Internet, but re-using a UDP port immediately between two calls
is just asking for trouble.

Related to this topic, I expect you have read
draft-stucker-sipping-early-media-coping-03.txt?  This document shows some
interesting behaviors all of which get even more interesting when combined
with an endpoint that wants to establish an SRTP call.

-d
 

> -----Original Message-----
> From: owner-ietf-rtpsec@xxxxxxxxxxxx 
> [mailto:owner-ietf-rtpsec@xxxxxxxxxxxx] On Behalf Of Nhut Nguyen
> Sent: Monday, February 05, 2007 11:32 AM
> To: ietf-rtpsec@xxxxxxx
> Subject: Early and late rogue RTP packets
> 
> Hi everyone,
> 
> Congrats on the approval of the upcoming BoF session! Looking 
> forwards to great discussions!
> 
>  
> 
> BTW, anybody knows of a generally agreed upon definition of 
> late and early rogue RTP packets? There are some references 
> out there but the definition seems conflicting based on 
> whether the reference timing is based on detection or on the session:
> 
>  
> 
> 1)     Early rogue RTP packets: detected BEFORE analyzing 
> packet, e.g. packets received after a session has been 
> closed. Late rogue: detected AFTER analyzing packets: e,g. 
> malformed RTP packets.
> 
>  
> 
> 2)     Early rogue: packets received BEFORE the session ends 
> and late rogue: received AFTER the session ends.
> 
>  
> 
> As you can see depending on which definition to be used, RTP 
> packets received after a BYE message can be called EARLY 
> (definition 1) or LATE (definition 2)! :-(. (one option is to 
> ignore early or late, just rogue RTP packets :-))
> 
>  
> 
> What is the consensus? Inquiring mind wants to know!
> 
>  
> 
> Cheers,
> 
>  
> 
> Nhut   
> 
>