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

[David P. Reed]

At 05:41 PM 6/18/2002 +1000, grenville armitage wrote:
and packet losses in the ISPs own network causes the
>enduser's TCP to generate repeat packets which are then
>counted against the monthly cap.

Note that packet losses also cause the TCP to send more slowly, which 
reduces the cost to the user in either case:

         1. if losses are congestion losses, the situation gets better and 
fewer packets get lost,

         2. if losses are due to bad lines or equipment, the user reduces 
his own cost automatically.

Of course slow speed and high charges also causes the user to switch to a 
competitor - the case where one would have an incentive to "profiteer" by 
charging for lost packets is where switching costs are high (like in the 
state-franchised monopoly access markets, but then your recourse is to 
elect less corrupt politicians who won't support such profiteering).

Not that I think that charging per byte sent is a good idea from a business 
point of view.  The incentive in such a system is to operate with totally 
clogged pipes, which boosts latency.  Ae proper charging model relates to 
meeting user latency needs, not accepting bits to be delivered at the whim 
of the network.