🚀 go-pugleaf

rheiger@osnbe.Olivetti.ch (Richard H. E. Eiger) writes:
..
: Every node charges the previous one on the communication path for the service
: provided. This gets propagated back to the source of the call. This way we
: don't introduce a huge overhead for billing centers. We don't need complex
: and treaties among a big number of participants, but only rather simple
: agreements between the sites directly connected. Hence there will also only
: be very few international treaties between bordering sites.

I don't particulary like the topology implications of your idea, and in the
end, I don't see having bilateral (or agreements like CIX) between connectivity
service providers as a big problem. CIX is -the- example that this works
with a very, very small overhead (and after the agreement of ANS with CIX,
you will see practically all US regionals joining CIX).

: Ideally the sites can make enough money on the non connectivity services to
: reduce the pure traffic (routing) costs close to zero. In my opinon the
: connectivity is not potentially the most profitable service of the "Internet".

One of the main problems of the "Internet" today is that it is -near- to
impossible to earn money from "non-connectivity" services. This is less
a technology problem (WAIS, for example has provisions for this), it's
just practically impossible to find a service to provide that somebody
else is not doing for free .

: To accomodate for very small sites intermediate sites could have a few class
: c networks with different subnetting schemes available (unless we could use
: different subnet masks within the same network).

It's not very likely that we will see "toaster"net with the current IP
network layer, and it's pretty clear that ISO CLNS will not do the
job either. For more information on the next generation IP (PIP, Nimrod,
TUBA) etc. you can join the Big-Internet mailing list or at have a look
at the archives of the list.

Allocating class C addresses is not a problem from an address space point 
of view (only a very low percentage of the class C addresses have been 
allocated), it is a problem from the growth of the routing tables. The 
solutions to this are being implemented now (CIDR, C#), they just don't have 
enough "customer" acceptance (as everybody has noticed that has tried to 
squezze a class B address out of me in the last month) yet.


-- 
CHUUG/EUnet Switzerland				Simon Poole
Zweierstrasse 35	Tel: +41 1 291 45 80
CH-8004 Zuerich		Fax: +41 1 291 46 42	poole@eunet.ch