Re: Provider credibility - does it matter? was Re: Inter-provider relations

[Jeremy Porter <jerry () fc net>]

In message <Pine.BSI.3.91.961025162433.20949C-100000 () avon-gw uk1 vbc net>, Jim 
Dixon writes:
> A large part of our Atlantic traffic (most, I believe) is US users 
> accessing our customers' (eg the BBC, you may have heard of them) or 
> our customers' customers' Web sites (eg www.bullfrog.co.uk).  As I 
> said, US networks pay pennies to carry traffic to us, and we pay at 
> least ten times as much to carry the packets across the Atlantic.  We 
> carry our packets, and we carry packets for US networks.  We pay at 
> least 90% of the carrying costs.
>
> Europe heavily subsidizes the US Internet.  It's not just VBCnet: the 
> European Internet community pays something like 90% of the costs of 
> traffic between Europe and North America.  The same applies to the
> rest of the world.
>
> If you paid attention to the discussion at hand, you might notice 
> that I was criticizing the Economist's model of the Internet.  One
> of the things wrong with that model was their failure to notice the
> huge subsidy that the rest of the world pays to the US Internet.
> There are a lot of other things wrong with the model.  Poor models
> make for poor reasoning, just like talking without listening makes
> for a poor dialog.

The Economist's model is incomplete, however I really doubt
it could be argued that Europe subsidizes the US Internet.
A large reason the cost of international leased lines are so
high is due to protectionist tarrifs in international telecom.
I understand it is cheaper to buy a leased line from the
us to Europe than the other way around.  (This at least has
been discussed on nanog before).  At any rate the most
the costs differences will do is is subsidize US access to Europe.
It doesn't effect the cost of US-US internet.  
Until someone invents a metric that measures Host Value,
there will be no easy way to estiablish the relative value of two
networks to each other to set up rational economic models between
providers.

Its one reason I advocate partial transit arragments and "pay for peering".
I do not advocate traffic based settlements, as the accounting
overhead is unacceptable, and the net results are usually a wash.
(Seeing as for any given random customer base, your net flows will
be equally distributed, assuming you market to different classes of
customers equally, and since bandwidth is usally only bought
symetrically it makes sense to market this way...)

I'd be happy if I could get techincal BGP peering setup with a large
national ISP, and pay based on a ratio of routable IP addresses
or some such. There are some details to work out on the costs etc,
but if Sprint has 10,000 prefixes out of 40,000, therefore cost of
partial transit should be about 25% of full transit.  Obivously
this isn't the best metric for many number of reasons, but it is a 
starting point for deteriming the relative value of two networks
to each other(i.e. how many customers).
There could probably be major changes to the voice network
with deregulation on local loops, etc if more rational settlements
systems were in place, but Telephone companies like to
suck money out of both ends of the pipe (To quote a former(?) FCC Chairman).

---
Jeremy Porter, Freeside Communications, Inc.      jerry () fc net
PO BOX 80315 Austin, Tx 78708  |  1-800-968-8750  |  512-458-9810
http://www.fc.net
- - - - - - - - - - - - - - - - -