[dccp] RE: feedback on draft-ietf-dccp-tfrc-media-00.txt

["Phelan, Tom" <tphelan at sonusnet.com>]

Hi Sally,

Those look like good points to me.  Thanks.  I'll update.

Tom P.

> -----Original Message-----
> From: floyd at icir.org [mailto:floyd at icir.org]
> Sent: Tuesday, August 02, 2005 6:30 PM
> To: Phelan, Tom
> Cc: dccp at ietf.org
> Subject: feedback on draft-ietf-dccp-tfrc-media-00.txt
> 
> 
> Tom -
> 
> This is a very helpful and thoughtful document!  A few nits below,
> about the possibility of a higher initial window; mentioning ECN
> in a few more places; and the response to an idle period.
> 
> Many thanks!
> 
> - Sally
> http://www.icir.org/floyd/
> 
> -----------------------------------------------------------
> Slow-Start:
> One possibility mentioned in the appendix of CCID 3 is to allow
> an initial window of up to eight small packets, up to 4K bytes,
> instead of the current initial window of four small packets.
> (It is mentioned as possible future research, however, not
> standardized...)
> 
> 
> Capacity Probing and Lost Packets:  
> I think it is worth says here "lost or ECN-marked packets",
> just to make it clear that with ECN at the congested
> routers, there don't have to be any packet losses.
> 
> I would recommend the same in Section 4.1, where it says
> "This means that packet loss will likely be routine with this
> strategy."  It is worth explicitly acknowledging that an 
> ECN-capable router would change this.
> 
> 
> Idleness Penalty:
> "If the transmitter goes idle for a few RTTs, as it would if, for
> instance, silence suppression were being used, the transmit rate
> returns to two packets per RTT,"
> 
> I would clarify to say that the allowed transmit rate is halved
> for every 4 RTTs of idle time, and that the allowed transmit rate
> is only halved when it is at least four packets per RTT (so that
> it is never halved to less than two packets per RTT).
> 
> This could be corrected in Section 4.3.2 as well.
>