Re: Three states for a binary bit (was Re: TCP (and SCTP) sucks on high speed networks)

["David P. Reed" <dpreed@reed.com>]

At 09:09 AM 12/6/00 +0000, Jon Crowcroft wrote:
>a pure end2end system that adapts to network conditions has a lag - at
>best its around 1/4 RTT (asuming BECN for sender adaption and FECN for
>receiver adaption)

Quantifying lags in terms of RTT is wrong.  RTT's are being employed in 
ambiquous ways (as they, sadly, often are) here.

There's the "speed of light" RTT which uses prioritized packets that bypass 
data packets, and the queue-driven RTT which uses packets that travel in 
FIFO queues.  The latter are affected by congestion.

One can do end-to-end adaptation just as fast as "inside the net" 
adaptation with the same time constants, because the same information can 
be available at the "speed of light".

But it is just silly to tie the applications' hands behind their backs and 
then say they can't implement overload management.