Re: [rtcweb] STUN for keep-alive

[sisalem <[e-mail]>]

  as far as I remember this was more thought for multicast sessions. for 
unicast the interval was 5 seconds (+-50%)

cheers

On 9/14/11 5:04 PM, Randell Jesup wrote:
> On 9/14/2011 7:45 AM, Jonathan Lennox wrote:
>> Christer Holmberg writes:
>>> Well, it depends on the amount of outgoing media traffic, but in 
>>> cases where you only receive media you would still need to send 
>>> keep-alives.
>> Remember, RTCWeb is (probably) mandating RTP/RTCP mux.  I'd need to 
>> do some math to be certain, but I'm pretty sure that in almost all 
>> normal circumstances, the RTCP regular reporting interval should be 
>> significantly smaller than the STUN keepalive interval.
>
> Agreed; I think it would require a bizarre session setup (RR=0) to 
> force the RTCP interval out far enough to cause problems.  (Don't 
> forget to include the +- 50% in the calculation.)
>
> RFC 3556:
>
>       o  If RR is zero, then the proportion of participants that are
>          senders can never be greater than RS/(RS+RR), and therefore
>          non-senders never get any RTCP bandwidth independent of the
>          number of senders.
>
>
> Obviously, RR values near 0 could also cause a problem.  A statement 
> about not using RR values that would result in RTCP intervals greater 
> than N should be enough I think.
>
> BTW: Have we mandated that RTP & RTCP be muxed always to all 
> destinations, or that it must be supported and offered?  If you're 
> talking to a legacy device that doesn't mux, what happens?
>