[1102] in Commercialization & Privatization of the Internet
Re: Talking Pointsink of this:
daemon@ATHENA.MIT.EDU (Richard Mandelbaum)
Wed Jul 31 16:13:24 1991
To: HABEGGER_J@bronze.colorado.edu
Cc: com-priv@psi.com
Date: Wed, 31 Jul 91 16:10:38 -0400
From: Richard Mandelbaum <rma@tsar.cc.rochester.edu>
Jay, I found your document to be very interesting even though I disagree
with many of its points. I don't have time, at the moment to answer it
explicitly and will try to do so later in August. For the moment the
following excerpt from an article of mine which will appear in the
future might be of interest as it expresses a somewhat different
point of view.
THE FUTURE OF THE "BACKBONE"
The NSFNet backbone is now the primary route for inter-regional
Internet traffic. It started life as a 6 node 56 kb backbone in 1986,
was upgraded to a 13 node T1 backbone in 1988/89 and will become a 16
node T3 backbone by the end of 1991. The current cooperative agreement
with ANS, Merit, IBM and MCI expires at the end of 1992. Since 1988,
the NSF has committed $28 million towards the backbone, with the State
of Michigan, IBM and MCI contributing at least an equivalent amount in
a mixture of either money, labor or equipment. In the past year
controversy has emerged concerning what should happen in the post 1992
period. The key questions are as follows:
What role should NSF play in 'providing' a backbone following the
termination of the current backbone contract?
Should an essentially similar solicitation be released for an
additional 3-5 year period?
Should the mid-levels be funded and encouraged to buy backbone services
in the marketplace?
What other possibilities exist to ensure the continued availability of
high performance backbone services to the research and education
communities?
The majority NTTF position is that,
"NSF should provide funds for a state of the art backbone for at least
three to five years. State of the art is defined as a cycle of
bandwidth upgrades to OC-12 (622 megabits) plus enhancements to basic
network services such that the NREN is capable of supporting, at a
minimum, end to end imaging and video teleconferencing.
... mission agencies such as NASA and Energy have always favored
central funding and
if an NSF administered NREN is not a state of the art network for
research (in addition to its other roles), then the mission agencies,
which have a high priority need for leading edge network connections
to campus research faculty, will likely embark on their own network to
provide broadband connections to these individuals." (1)
(1 Letter from Mike Roberts. Feb. 6 , 1991)
Bill Schrader, president of PSI, in his presentation at the recent OTA
workshop called for a modification of the "NREN bill"
"to require ALL operational funds to be distributed DIRECTLY to
America's post- secondary, secondary and elementary institutions and
non-profit research centers to be used to purchase networking
services.
to minimize the amount of funds reserved for network research, which
will now be partially provided by commercial firms in their own
interest." (2)
(2 Bill Schrader, Why the NREN is in trouble. A presentation to the
Office of Technology Assessment Workshop: The Privatized NREN., Feb 14, 1991 )
There have also been strong statements, especially from the mid-level
community, about the major drawbacks in funding only end-user
institutions. In particular, Alison Brown, director of OARNet has
written,
"I don't necessarily disagree that providing subsidies directly to the
educational institutions is a desirable end-state, but I'm pretty
convinced that trying to do it now would un-do much of the good NSFNET
has done, although it might do some of the commercial providers of
Internet services some good in the short run.
One of the prerequisites for giving subsidies to the end user is that
they can obtain a single connection from any vendor and reach anyone
who purchases Internet services from any vendor, just like I can buy
long distance from MCI and still reach people who buy long-distance
service from AT&T. ... The Internet currently has a high level of
technical interoperability but at present has virtually no basis for
using each others facilities ... and there is CERTAINLY no set of
economic agreements in place which are comparable to those worked out
by the phone companies. I believe ANS and NSF are aware of the need
for such agreements, but they need time to work them out, and until
PSI and ALTERnet and other commercial providers yet to come are also
included, there is still no basis for assuming I will reach sites on
ALTERnet if I am a PSI customer. ...
Right now the mid-level networks are confronting this very dilemma,
and there is no benefit in forcing the campuses to confront it until
it can be solved by the mid-levels, which have both more clout and
more ability to solve it. ...
There isn't much evidence that the schools in the U.S. are unhappy
with the service they are getting through their regional networks and
NSFNET. ... Why not leave the current structures in place for
supporting the academic marketplace and concentrate on how we move into
providing additional connectivity to commercial services and firms? I
think we're not seeing the forest for the trees on the issue of
subsidies to schools vs. state or regional providers." (3)
(3 Letter from Alison Brown to members@farnet.org. March 1, 1991)
The critical issues for the regional networks, firmly rooted as they
are in the research and education community, center on:
a) the committment to eliminate usage restrictions on the Internet so
that true economies of scale, obtainable only with heavy commercial
involvement in the use of the Internet, can be achieved.
b) the desire to support the "Have-nots", the K-12 sector, the small
schools, and education in rural areas, perhaps through distance
learning.
c) the need to support network research and advances in the deployment
of high- speed networks (0C-12, or even higher); and
d) the willingness to "level the playing field" so that the regional's
are not a source of unfair competition to the for-profits.
a. Elimination of Usage Restrictions
It seems clear that having a backbone provider who is fully-funded by
the federal government insures continued federal involvement in usage
questions - a responsibility that the NSF wants to terminate! A
continuation of the status-quo is thus something which the regionals
should oppose.
b. Supporting the Have-Nots
On the other hand, direct distribution of funds to end-user
institutions, though nice on paper is not really practical. In the
past year NYSERNet started projects which have, or will, extend
internet connectivity to 3 school districts in New York City, in the
South Bronx, in Bedford- Stuyvesant /Fort Greene in Brooklyn, in Howard
Beach in Queens. NYSERNet started a distance-learning project in the
high schools of Livingston-Steuben-Wyoming Counties in upstate New York
. It is connecting the science high schools in New York City, Bronx
H.S. of Science, Brooklyn Tech,Stuyvesant and others as well as
initiating an experimental counseling program aimed at New York State
High Schools and modelled on Cornell's Uncle Ezra program. NYSERNet
is about to connect up the New York State Library system and has
already connected the New York City Public Library. It has extended
connectivity to the Buffalo Museum of Science and the New York Museum
of Natural History.
Each of these projects involved subsidies from the Federal government,
the State, private foundations or industry. The projects would not
have been attempted if the schools had to wait for the government to
do them. They would certainly not have been iniated if subsidies went
directly to end-user institutions. No individual in School District 9
in the South Bronx has ever applied to NSF, DOD or DOE for a grant.
It is not a hot bed of cutting edge research. It does however have
students and teachers who can benefit from being connected to the rest
of the Internet Community. Neither Columbia nor the City University
of New York is truly interested in School District 9. PSI, which is
interested in them as a customer, is not yet able to subsidize their
entry into the Internet. By default the regionals are left with the
job of extending and broadening the Internet to these new classes of
users.
Once the difficulties associated with connecting the "have-nots" have
been solved, the problem of supporting them still exists. To this end
NYSERNet is trying to set up an end-user services division. Who should
fund it? One can wait for the end-user's the principals or
superintendents with vision to convince their school board or local
education bureaucracy of the importance of networking. More
realistically, NYSERNet will attempt to secure monies on their behalf.
c. Support for Research
The Marketmaker concept offers a way out of the dilemma of whom to
fund. The Marketmaker is not in competition with either the telephone
companies or the new Internet providers such as ANS or PSI. The
Marketmaker is a customer of these providers. Encouraging the
regionals to move in the Marketmaker direction and get out of the
operations end of networking will make it more acceptable for all
parties to consider the regionals as the best loci for funding.
This does not solve the problem of the high-end, gigabit backbone. No
commercial provider will supply this service at a cost the academic
community can afford. On the other hand, having the federal government
run the high-end backbone again raises the issue of acceptable usage.
A solution to the high-end backbone problem should contain the
following four elements:
1) structuring the backbone contract so that it's manager is not in
the 'retail' TCP/IP business or at the very least cannot subsidize
divisions which are in that business.
2) restrict use of federal monies to covering only for the marginal
costs of extending a T1, T2, or T3 production network to the higher
bandwidths necessary for new applications or research.
3) have all regionals and commercial providers pay for access to the
backbone; and
4) fund the regionals, via ear marked federal grant money, to enable
them to connect the have-nots, as well as support the researchers.
This is a moving window model. "Have-nots" eventually become "haves",
with their place being taken by new legions of "have-nots".
Experimental uses become mundane with new uses appearing. Ultimately,
new telephone company services such as B-ISDN or SMDS might provide
gigabit-on-demand services. This will provide a long-term solution for
the high-end backbone, but is not likely to occur within the next 3 -5
years. In the interim, a model of backbone funding and regional
funding, combined with a move by the regionals, out of the operations
business and into the Marketmaker mode, might be a solution.
____________________
This is something I wrote for another reason, but I would like to he
ar
what Com-Priv people think about some of these ideas.
Talking Points
Internet Privatization and Commercialization
by Jay Habegger
July 30, 1991
#1 Definitions. PRIVATIZATION: to transfer a good that has
previously been provided by the public sector to private
sector provision. COMMERCIALIZATION: this term often used
interchangeably with privatization by many members of the
Internet/NREN community, but here it refers to the policy of
allowing and promoting the use of the Internet to carry
commercially oriented traffic.
#2 There is not a theoretical justification for public
provision of Internet type networks. Access to these
networks is not a "public good" in the economic sense, i.e.
it does not have the properties of nonexcludibility or
nonrivalous consumption, and the good is therefore eligible
for provision by the private sector. It is important to
recognize that Internet type networks are provided by public
entities today only as a result of historical accident. The
technology was pioneered by DARPA and implemented by NSF to
solve the need for remote access to supercomputer centers.
If AT&T had acted upon an offer from DARPA to operate the
ARPANET in the early 1970's [1] the situation today might
be very different, perhaps more closely resembling the
present X.25 marketplace.
#3 Even if Internet type networks are eligible for private
sector provision there may be public policy considerations
that argue for public sector involvement. The distinction
between what is theoretically required and what is a
political decision to further public policy objectives
should be kept clear. For example, the desire for some
level of universal access to the Internet is a public policy
objective that may require public sector involvement.
Because there may be public policy objectives that require
some public sector involvement, does not imply, however,
that government needs to provide the good. For example, in
case of poverty or hardship the U.S. government provides
food-stamps, but the U.S. government does not operate
supermarkets or set the price of food.
#4 There are many constituencies involved in the Internet/NREN
debate that each possess a public policy agenda. Many of
these constituencies are likely to resist any attempt at
privatization that does not satisfy their public policy
objectives. For example, institutions presently enjoy free
use of the backbone are likely to resist any privatization
scheme that reduces or eliminates this benefit. Many
constituencies are likely to favor commercialization over
privatization in order to maintain their subsidies.
#5 The issue of privatizing the Internet can be reduced to
three questions: 1) how to privatize the NSFNET backbone; 2)
how to privatize the mid-level networks; and 3) how to
handle the issue of IP address assignment and name
registration.
#6 The NSFNET backbone is essentially already privatized in
that ANS - a private firm - provides the bandwidth for the
NSFNET for a fee out of the total bandwidth managed by ANS.
This is analogous to how public entities purchase telephone
service; they lease a portion of the bandwidth managed by an
IXC, such as U.S. Sprint or AT&T.
#7 Problems arise from the fact that NSF acts as a purchasing
agent for almost all of the potential backbone customers.
NSF essentially purchases bandwidth from ANS and then offers
it free to institutions and mid-level networks which must
only finance the cost of a link to a NSFNET backbone POP.
Instead of the ANS monopoly that is often discussed, there
exists an NSF monopsony. This situation is analogous to the
case if the U.S. government contracted with one IXC and then
offered free long distance telephone service to all U.S.
citizens provided that they use the carrier that the
government had contracted with. Obviously, there would be
little incentive for an individual to use some other IXC and
pay for long distance service.
#8 Once the problem is recognized to be an NSF monopsony rather
than an ANS monopoly it becomes clear that instead of
eliminating NSF acceptable use policies they should be
strengthened. Without acceptable use rules, NSF would be in
a position to offer free use of the backbone to all
potential customers eliminating any hopes of creating a
competitive backbone market. In the presence of strong
acceptable use rules, some potential customers, e.g. for-
profit entities, will be forced to pay for backbone service
and choose a backbone provider. Commercializing the NSFNET
backbone in the sense of the definition in #1 above should
not be considered.
#9 The obvious solution to the present monopsony situation is
to stop the NSF from acting as an agent for almost all of
the potential customers. One method of achieving this is to
allow the present NSFNET to cease operation in 1992 and
provide each institution which NSF wishes to have access to
the backbone with an "Access Token". NSF may create
different classes to tokens based on the size and need of
the institution. A small college, for example, may on need
a 56 kbps gateway to the backbone while anything less than
1.5 Mbps may be insufficient for a large university. The
token could be presented to the service provider of choice
by the institution in return for privilege of connecting to
the service providers network at the closest POP. The
service provider could then present the token to NSF for
remuneration of a standard access fee.
#10 The Access Token scheme eliminates the need for the need to
distinguish between commercial and non-commercial traffic.
Instead of providing bandwidth, NSF would be subsidizing
access to the network for approved sites. Any traffic
originating or leaving an approved site would by definition
pertain to the mission of that site.
#10 Another solution that may be considered at first is for NSF
to distribute the backbone contract over all backbone
providers. Under this plan each backbone provider would be
responsible for providing access to an equal portion of the
institutions and mid-levels that have their access sponsored
by NSF. This plan suffers from one serious problem. The
contract would have to renegotiated each time a backbone
provider entered or exited the market, or a cartel would be
created which would likely prevent any other firms from
entering the backbone market.
#11 In contrast to the backbone market, there is a possibility
for some publicly supported mid-level networks to become
regional monopolies. Although these mid-level networks are
not granted any special market privileges by the state
legislatures or state PUCs, a monopoly can arise because
these networks are often subsidized by both the NSF and
state taxpayers. In some cases the management of a mid-
level network that enjoys a subsidy may see it as an
opportunity to price service below cost in order to pursue a
policy agenda, e.g. providing universal access to
educational institutions within the state.
#13 The simplest method of forestalling the possibility of a
regional monopoly is for NSF to end subsidies to mid-level
networks. Note that if NSF grants an Access Token to a mid-
level is it is still providing a subsidy to the mid-level,
which leads to the conclusion that NSF should only provide
Access Tokens to institutions and not mid-level networks.
Thus, Access Tokens would flow primarily from institutions
to mid-levels and backbone providers that offered a plethora
of POPS. This creates no problem because a mid-level can
present the Access Tokens to NSF and the mid-level can in
turn pay a backbone provider for a connection.
#14 The use of an Access Token system is likely to create a
situation that is similar to the provision of other
utilities. The institution must merely trade its Access
Token for access to the POP of a local service provider.
Whether the local service provider is a mid-level that has a
connection to the backbone, or actually a backbone provider
makes no difference to the institution which is only
concerned about the most functional access to the resources
available on the Internet that can be obtained with the
Access Token.
#15 IP address assignment and name registration do not need to
be performed by a public entity. The best organization for
performing these functions might be a non-profit consortium
with membership open to all backbone providers. This
organization may be constituted along the lines of an
organization such as SPARC International which assures
compliance with the SPARC microprocessor architecture.
ENDNOTES
[1] Lawrence G. Roberts, "The Evolution of Packet Switching,"
Proceedings of the IEEE, Vol. 66, No. 11, November 1978. pp.
1307-1313.
