The zero-settlement scheme works ok as long as interconnected
parties are of about the same size.
That makes them suffer equally from routing bloat.
Incidentally, that was part of the reasoning after Sprint's
IXP peering policy.
In any case having zillions of small ISPs to be at top-level IXPs is
economical and technical insanity, and it seems that it can only be
fixed by rather steep settlements at IXPs to make it a very expensive
prestige shopping.
BTW, "settlements for routes" is a rather interesting thing.
Imagine, say, Sprint peering with Joe Blow Internet-And-Burgers.
Let's assume that route costs $100/yr, so Sprint pays $1M/yr to
JBIB and JBIB pays back $300/yr. Sounds like a good business
for JBIB 
The better idea is auctioning and trading the routing slots, though
it is unclear where the proceeds from the initial auction should
go. Probably to fund the Network Police which would chase the
bootleg route injectors.
I don't know what you make of it, but i think that sucks a lot more
than the zero-sum model we have now.
--vadim
Not speaking for Sprint.
PS Given the choice between fixing the technology and fixing the
society it is always more prudent to go after technology.
From list-admin@merit.edu Tue Jan 30 20:56:10 1996
Received: from merit.edu (merit.edu [35.1.1.42]) by titan.sprintlink.net (8.6.9/8.6.9) with ESMTP id UAA11534 for <avg@titan.sprintlink.net>; Tue, 30 Jan 1996 20:56:08 -0500
Received: from nico.aarnet.edu.au (nico.aarnet.edu.au [139.130.204.16]) by merit.edu (8.7.3/merit-2.0) with SMTP id TAA25560 for <nanog@merit.edu>; Tue, 30 Jan 1996 19:59:35 -0500 (EST)
Received: (from gih@localhost) by nico.aarnet.edu.au (8.6.10/8.6.10) id LAA25214; Wed, 31 Jan 1996 11:56:53 +1100
Message-Id: <199601310056.LAA25214@nico.aarnet.edu.au>