Presume the existence of a future world where many providers are
operating on a global basis and absent the current somewhat US-
centric interconnection web.
In said world, the cost of serving a given-size customer or ISP connection
which contains 90% local traffic and 10% far-distant traffic ("international")
is different than serving one which is 10% local and 90% distant traffic.
In this situation, one can either ignore the actual cost-per-destination and
charge based on an assumed traffic distribution, or one can recover based
on some extraction of the actual customer traffic profile. Note that a
based method does not have to be "per-byte"; it can be as simple as having
spliting current monthly utilization fees into local/international
The problem with the former case (assumed traffic model and price) is that
one risks too conservative a profile (with higher net price and loss of new
customers) or too aggressive a profile (with great growth potential but the
strong attraction of heavy distant-traffic customers and unrecovered costs).
This, btw, is today's model for the vast majority of ISP's; we all arrange for
some form of international connectivity and hope that these costs do not
dominate our overall infrastructure costs. Of course, this approach only
encourages distant-heavy usage applications (e.g. international IP voice/fax)
to migrate to profile-insensitive Internet services and is eventually self-