Peter Ford <peterf@microsoft.com> wrote:

Metcalfe certainly is not the first, or the last, to call for "reservations" and last time I looked a lot of the work in the IETF is centered on this area. This does bring up the interesting point of will/should the Internet business model change to support reservations.

Peter, i nave no problems with resource reservation or charging based on resource reservation as a business model. It is perfectly valid. What i have problem with is that nobody ever managed to explain how a data network of size of Internet providing resource reservation can be built. The problem is that in "fat state" schemes (i.e. keeping per-connection state in gateways) consistently yield linear growth of state in gateways at exchange points as function of network size. Which, given exponential growth of Internet is not going to fly. Remember that Moore's Law has smaller exponent built in. I wouldn't even dwell on differences in session arrival patterns between POTS and Internet which make all Erlang-based telco technology to be totally useless to handle Internet kind of traffic. I feel sorry for folks who work on RSVP et al, they're very much like those brilliant minds who spent their lives improving perpetuum mobiles. Some of the devices were truly amazing feats of engineering. No amount of technical cleverness can deal with the fundamental scaling problem. It's like the second law of thermodynamics.

Let's assume some kind of "promise" beyond the current best effort service is offered. Do you really believe this will not be based on usage?

I know that application targeted by resource-reservation can happily work without. There are numerous proofs by existance. My question is -- what exactly resource reservation is going to solve? Is there a problem to start with? --vadim