[e2e] TCP, retransmissions, and ISPs with byte caps

[David P. Reed]

At 05:34 PM 6/18/2002 +0200, Philippe Strauss wrote:
>then came the LANcity. my colleagues tried it and shipped it back due
>to a terribly long time for the modem to reconnect after
>a plant outage or noise burst (several hours for ~ 500 modems)
>modulation and access to the return path was close to the docsis1.0
>standard I think.

As a customer, I have never observed the several hour case, and there are 
several hundred LANCity modems on our branch of the tree.   Minutes, 
yes.  Hours, no.   But who knows what your plant looks like.


>terayon and it's teracomm 1000, a well designed system on
>the transission side: spread spectrum CDMA, good performance even
>on noisy plant.
>it's allocation scheme on the upstream path, most of the time,
>impose a 20ms min RTT for the basic UBR traffic contracts.
>with a non UBR you get better RTT, same values as you
>see with the LANcity.

Bad design.   Must have been designed for ATM bigots.  20 ms is enough time 
for a signal to travel 4,000 miles.   You can do arbitration orders of 
magnitude faster than that on a radio channel in freespace, which is much 
harder to work over than a coax plant, which is like a cleanroom in comparison.

This is what you get when your engineers only think about bitrate and not 
about latency.  They probably drive cars that go from 0-60 in 5 seconds, 
after waiting 7 seconds for the turbocharger lag.