Re: [rtcweb] consent freshness vs. circuit breakers

On Sep 9, 2013, at 8:45 AM, Bernard Aboba <bernard_aboba@hotmail.com> wrote:

> 
> Christer said: 
> 
> "Now, as we have decided to not use SDES, I guess that can be removed."
> 
> [BA] Actually, I believe that a sentence or two describing why the needs of consent freshness are not met by mechanisms such as SRTCP is worth keeping, and possibly should even be expanded upon.  One reason is that we also have the "circuit breakers" mechanism which *does* rely (in part) on SRTCP.   
> 
> Christer said: 
> 
> "But, based on that, I'd just like to verify whether there is still a need for the draft :)"
> 
> [BA] I do think that the draft is still necessary, even without SDES/SRTP.   The usage of ICE for consent seems to me to be less susceptible to a variety of issues than a dependency on SRTCP might be. 
> 
> However, I am concerned about the interaction between circuit breakers and consent freshness.  At IETF 87, we heard that the circuit breakers mechanism could fire even when the overall loss rate was quite low, due to burst losses.  That shouldn't be true of consent, since loss of consent requires up to 30 consecutive losses over a 15 second period.  However, it would appear to me that the current circuit breakers algorithms will cause media sending to be curtailed prior to loss of consent in many cases.  

Yes, consent needs to avoid the same pitfalls that were explained for circuit breakers.  

(For others that might be interested, the AVT presentation describing the research on circuit breakers is at http://www.ietf.org/proceedings/87/slides/slides-87-avtcore-2.pdf.)

-d

> 
> _______________________________________________
> rtcweb mailing list
> rtcweb@ietf.org
> https://www.ietf.org/mailman/listinfo/rtcweb