[238] in Commercialization & Privatization of the Internet
a recent article on NREN
daemon@ATHENA.MIT.EDU (Martin Lee Schoffstall)
Thu Feb 28 14:24:07 1991
To: com-priv@psi.com
Date: Thu, 28 Feb 91 15:28:39 -0500
From: "Martin Lee Schoffstall" <schoff@psi.com>
The Economist
February 16-22, 1991
Page 24
Data Communications Network: Common Electronic Policy
America's electronic networks are abuzz with debate
over...the future of America's electronic networks. Since the
1970s, the government has been building subsidised data-
communications networks so that researchers could more
easily send electronic mail and share computer resources
and and the results of experiments. The effort has been a
spectacular success. Many scientists and engineers now
chatter away all day to each other by computer; traffic is
growing by 30-40% a month. One reason it is growing so
fast is that it is often provided free, making scientists as
subsidy-dependent as sugar growers or train drivers.
In late January, Senator Albert Gore began his second
attempt to pass a bill to promote high-performance
computing and communications. The bill proposes to spend
$1 billion on computer research. A big chunk of that would
support a $400m plan to build a "National Research and
Education Network"-a high-speed computer network, many
times faster than the fastest parts of today's networks.
Meanwhile, the National Science Foundation (NSF), a big
channel for federal networking subsidies, is rethinking its
own support for networks.
The biggest network, called the INTERNET, is itself made up
of over 2,000 smaller networks. Some of them are
networks built to connect a state or region; others grew
nationwide to serve the needs of specific groups, like
computer scientists and research universities. Each group
brings its own special concerns to the debate, from access to
supercomputers to computers for kindergarten children.
The arguments grow hottest where these networks
converge-in this case, the so-called NSFNET backbone.
Take a deep breath and ingest a bit of alphabet soup. The
NSFNET is the fastest part of the INTERNET. It uses that
speed to provide both access to supercomputers and high-
volume, long distance traffic such as that from New York to
California. NSF sub-contracts the running of NSFNET to
Merit, a company based in Michigan. Merit, in turn has
looked to IBM and MCI for equipment and expertise.
Though NSFNET costs about $10m a year to run, a $3m-a-
year subsidy from the NSF plus further money from IBM
and MCI enable its services to be offered free to most of the
academics and researchers who use them. In the autumn
of 1992, however NSF's contract with Merit expires.
As a first step towards weaning networks from subsidies,
the NSF is thinking about subsidising users directly. In
theory NSFNET's services could then be provided on
competitive terms, with both sibsidised and unsubsidised
users free to choose from their own provider of high-speed
services. Eventually, subsidies on today's services would
wither away, and the government could devote its
resources to encouraging the development of tomorrow's
services.
Fine in theory, but in practice it is difficult to decide who,
exactly, are the users who should be so subsidised.
NSFNET's biggest customers today are 30 or so regional
networks set up as "retailers" of networking services to
their local schools, colleges and businesses. It is these
regional networks to which the NSF now tentatively favours
switching the subsidies next year. In the short term, this is
a simple way to ease researchers on to the market. Over
the longer term, however, subsidising the regional
networks could kill the very competition it is meant to
promote.
The regionals already compete with many of the services
offered by the two commercial vendors of INTERNET
services--UUNET and Performance Systems International
(PSI), both based in Virginia. Many reckon that
competition would be better served by giving the NSF's
money to universities and researchers instead--though the
NSF worries that such a scheme could be wastefully hard to
administer.
But fair competition among regionals is only half of the
equation. There is also the problem of replacing services
now provided by NSFNET with new arrangements that are
at once competitive, responsive to users and--with
government money as a carrot--capable of undertaking
high-risk investments in as-yet-untired networking
technology.
Merit, MCI and IBM have pooled their expertise into a
consortium called Advanced Network Services, which will
bid to take over from NSFNET after 1992. That raises the
worry that a privatized version of NSFNET might inherit an
over-privileged competitive position. Not that anybody
agrees how to define privilege in this new industry.
