[546] in Commercialization & Privatization of the Internet
NREN Presentation at St. Louis IETF (rough notes)
daemon@ATHENA.MIT.EDU (Martin Lee Schoffstall)
Fri Apr 5 03:56:27 1991
To: com-priv@uu.psi.com
Cc: vcerf@nri.reston.va.us, nren-discuss@psi.com
Date: Wed, 03 Apr 91 12:32:52 -0500
From: "Martin Lee Schoffstall" <schoff@psi.com>
This is the written brief that I got from one of the PSI staff members
who attended. I'm hoping that Vint Cerf can supply a more complete
brief as he real recording devices.
Marty
----------------------------
Two presentations on the NREN were made at the IETF meeting
in St. Louis, Missouri that just ended. The first presentation was entitled
"Architecture and Goals for the Interim Interagency NREN", and was
jointly presented by William Johnston of the Lawrence Berkeley Laboratory,
and Peter Ford of the Los Alamos National Laboratory. The second
presentation was entitled "NREN Legislative Update", and was presented
by William Bostwick of the Federal Networking Council.
The first presentation started out with the presenter noting
that the networking component of the High Performance Computer Communications
(HPCC) legislation had a number of goals, among which were: to expand
the U.S. portion of the Internet; to support the education, science and
research communities; to support the components of the HPCC pertaining
to basic research and to human resources; to support the Grand Challenges.
It was further noted that the NREN is neither a beginning nor an
endpoint in the development of networking. Instead, a point was made
that the NREN will be an evolution of the current Internet that will
occur as a result of increased cooperation with industry. Lessons learnt
from internetworking research and the Internet will be applied in the NREN.
The presentation continued with a statement on the criteria by
which the success of the NREN will be determined. These criteria were to
be the extent to which the bandwidth and connectivity requirements of science
and education could be successfully addressed; and the improvements in
the quality of education in terms of increases in the number of qualified
[education] professionals and improved curriculums. Further, it was
noted that the extent of the NREN would be determined by what people
were willing to pay for. However a point was made that "what people
are willing to pay for" does not necessarily imply privatization. On the
other hand, it was noted that it does not preclude privatization either:
it was stated that network connectivity does not necessarily equate to
network staff. I got the impression that the presenter meant that portions
of the NREN may in "private", non-government supported networks.
The presenter then went on to enumerate the portions of the
NREN as consisting of a gigabit Research and Development network which
DARPA would be the lead agency for, an interim interagency NREN for which
the lead agency would be the NSF, and two other components. It is significant
to note here that none of the NREN portions enumerated addressed the
needs of the K-12 community.
After the overview of the NREN was given, as above, requirements
were enumerated for the architecture of the NREN implementation. These
requirements were as follows: to accelerate the availability of high
performance computing resources; to spur gains in the industrial
competitiveness of the United States in world markets; to support
the scientific and technical aspects of the Grand Challenges; to provide
availability to large data repositories; to provide ubiquity of access
to computing resources; to increase cooperation with the data communications
industry. It was noted that the intent of the NREN was to make network
infrastructure as ubiquitous as telephone infrastructure in the United
States, and to serve the needs of various Federal organizations with
specific service and connectivity requirements. It was further noted
that this second intent of service to Federal organizations needed
to be reconciled with the requirements for ubiquity of access.
The presenter then discussed the need for security in the NREN.
In elaborating on this, it was stated that the NREN security model would
have to acknowledge that the network was as important as other computing
resources, and hence appropriate security measures to protect it would
be required. It was further noted that recognition would have to
be made that different levels of security would be necessary. To this end,
research and development would be necessary to support the implementation
of various security models that address the disparate security levels
required.
It was also briefly noted that, to this end, the NREN would have
to provide support for virtual private internets, policy routing and
address privacy. The presenter was of the opinion that a large commercial
infrastructure could best address these needs.
At this point, a request was made that either of the presenters
speak to the following issues, which had not been addressed in the
presentation thus far: whether the NREN would be multiprotocol; whether
international connectivity was anticipated; whether connections to commercial
service providers was anticipated. In all three instances, the presenters
answered in the affirmative, noting the evolutionary nature of the
NREN with respect to the common service providers.
The presenters were then asked to elaborate on the way in which
the NREN model of need handled commercial components of the Internet
currently connected to the NSFNet backbone. One speaker declined to
comment on this matter, noting that this was a policy issue. The second
speaker felt that commercial connectivity was a good thing, and ought
to happen, but did not feel up to addressing the issue.
The matter of commercial connectivity to the NREN was further
pursued by a comment from the audience that commercial connectivity
had implications with respect to access control and privacy, and was
thus not just a policy issue. The speakers replied by stating that
access control and privacy were issues for all users, and hence no
distinction needed to be made.
The presentation continued with a listing of some of the technologies
that would be used on the way to a gigabit per second NREN. These were:
SMDS, B-ISDN, ATM technologies, Frame Relay technologies and SONET.
A question was asked at this point as to whether the presenter felt
that SMDS and Frame Relay were functionally similar technologies,
and if so, why it was felt that both needed to be used. The presenter
answered that what was meant was that both technologies would be studied
and evaluated with respect to their applicability to the NREN. It was then
futher noted by the presenter that T1 and T3 technologies would still be
around, and that it was anticipated that the NREN would initially use
these technologies. Further, the issue of multiple protocols was revisited
with a statement that the NREN would run both IP and CLNP, a multiplicity
of routing protocols, and a variety of software in support of NIC and
NOC functionality. It was reiterated very strongly at this point that
the NREN would be multiprotocol.
The presenter noted that appropriate use policies dictated the way
the Internet looked and functioned today, and that such policies would
influence the technologies used in the NREN.
A comment was then made by a member of the audience that recent
years have seen rapid advances in both hardware and "base" transport
technologies, without commensurate development in user services and
applications. It was pointed out, for example, that some still used
the same electronic mail program that they used 15 years ago. The presenter
responded to the effect that large scale networking was still in its
infancy, was still evolving, and that continuing work would yield
advances in this area.
The presentation then continued with the presenter listing the
things he felt were "right" about the current Internet: that it continued
to evolve, as opposed to being static; that interoperability was widespread;
that it was "free", or appeared to be "free", to end-users; that it
provided a stable environment for the development of new applications
of networking; that it had a large community willing to expend the
effort to ensure that it continued to work. The presenter then went
on to enumerate what he felt were the problems with the current Internet:
it had scalability problems, with problems in IP address space
and routing protocol scalability being specifically cited; it had
little or no accounting facilities; the absence of agreement on an
approach to address the issue of security, much less any security
facilities; the significant amount of network expertise that was
needed to successfully connect to the existing network; the lack
of resource guarantees.
At this point, two models were outlined for the implementation
of the NREN network from a technical standpoint. The first model, which
was labeled the state of current affairs, involved a network of level 1
bit pipes, with level 3 routers providing network service. In discussing
this model, it was stated that among the advantages of the model were
that there were cost benefits, in that it was possible to make bulk
bandwidth purchases, and that the model was based on currently available
technologies, and hence it was clear that a network could be
realized based on these technologies. It was pointed out however that
a drawback of this model was that bandwidth distribution was
static, leading to inefficiencies with respect to adequately addressing
the changing bandwidth requirements of network users.
A second model was then discussed, which the presenter felt was
better for the purposes of implementing a network. In this model, level
2 packet services based on such technologies as SMDS and Frame Relay
would be used in conjunction with level 3 routers to realize the network.
It was felt by the presenter that the common service provisions of level
2 service would be advantageous. It was also noted that there might
be cost benefits to level 2 services, although it was too early to
tell. Further advantages mentioned were that link-level resource
guarantees were provided, and that higher-level infrastructure development
would provide a more robust network. The presenter also mentioned that
widespread level 2 service availability was anticipated in the 1992
or 1993 fiscal years.
A comment was made from the audience that the second model suffered
from problems with geographic distribution: currently, different areas
have different degrees of level 2 service availability, and carriers
unsure of market demand are reluctant to invest in level 2
infrastructure. The presenter agreed that this was a problem that needed
to be resolved, commenting that a full mesh of service was desirable,
with many common carriers providing interconnectivity as required.
After outlining the two implementation models, the presenter moved
on to an outline of the actual implementation plan for the NREN. The
presenter stated that development would continue based on T1 and T3
technologies, as per the first model, with a migration to the
technologies of the second model occurring down the road. Factors that
would determine the exact schedule for migration included cost considerations,
the requirements of the HPCC, and the availability of level 2 services. It
was reiterated that the implementation plan required the addressing
of accounting problems, specifically the lack of accounting in the
current Internet, as well as extant problems with scalability, routing,
security, reliability, robustness, service guarantees and user services.
The management plan for the NREN was presented next. This portion
of the presentation started out with the enumeration of the purposes
of the NREN management structure. These were: to facilitate the transition
of the NSFNet to the interim interagency NREN, and ultimately to the
full-fledged gigabit NREN; to establish a balance between network
stability and the need to progress through the use of new technologies;
to provide a forum for the critical review of the NREN; to provide
community control over the NREN.
A plan was then described whereby the NREN would be managed by
a not-for-profit public corporation with the necessary expertise. A
diagram was put up, showing the managerial structure of the corporation,
named the Corporation for Public Networking (CPN) for expository purposes
only.
The management structure diagram shows that the CPN is managed
by a board of directors responsible to a user community. Responsible
in turn to the board of directors are the managers of a number of areas,
such as "network operations", "user services" and "NIC services". The
user community was depicted as being made up of Federal agencies such as
NASA, DARPA, DOE and the NSF, academic and educational representatives such
as CREN, EDUCOM and FARNET, the scientific community such as SIAM and ACM,
the telecommunications industry such as AT&T, U.S. Sprint and Bellcore
and the Internet community, as represented by the IETF.
A Technical Review committee and an Internetworking committee
were also depicted in the diagram. It was stated that the two
committees would provide consulting expertise to the CPN on technical
issues. However a point was made that the recommendations of the Technical
Review committee would be nonbinding on the CPN. Although it was not
stated explicitly to be the case, I assume that this is also true of the
Internetworking committee.
At this point, something akin to World War 3 started. A complaint
was raised that no end-users were included among the depicted user
community, and that furthermore no K-12 representation was present
among the educational representatives. In response, the presenters asserted
that end-users were represented by the various Federal agencies.
Heated discussion ensued, during which time a point was made that
unless there is direct end-user representation on the board of directors
of the CPN, that board would be more interested in running the corporation
than in addressing end-user needs.
At this point, a representative of the National Science Foundation
spoke at some length on the NREN. Points were made that the plan being
presented was only preliminary, and that one of the purposes of the
presentation was to solicit input from the community. The policy and
legislative processes relevant to the NREN were also discussed, as
were the roles of such organizations as the Office of Technology Assessment
and the U.S. Congress. The NSF representative also stated that
the connection of private networks was necessary, and that the NREN
could not be built solely with Federal funds, but that private
commercial investment was necessary. It was also pointed out that the
K-12 community would be represented by the Department of Education, and that
furthermore, the Congress would ensure that their needs were addressed through
legislation.
At this point, it was also noted by a member of the audience that
gigabit networking was necessary for the good of America.
A significant discussion also ensued when one of the presenters
asserted that the points being made about K-12 networking requiring
low-speed connections were absurd. It was further stated by the presenter
that K-12 institutions needed gigabit networking as much as any other
community. Upon being asked how it was proposed that gigabit connections
to every K-12 institution in the country would be paid for, the presenter
responded that it was necessary, and that people would have to be
convinced to pay for it.
A question was also raised with regard to who would be responsible
for determining appropriate use policies in the NREN. In response,
it was said that the board of directors would make such determinations,
in response to input from the user community.
Other questions raised included one about the implications
of a single corporation managing the NREN. Specifically, the question
was raised as to whether a single corporation implied a single
NREN network, and how the management plan addressed the issue of transitioning
from a centrally planned network to a more competitive environment with
multiple network providers. It was stated in response that the management
plan presented was a first step, in that there was no direct Federal
control of the NREN in the plan. Instead the NREN was controlled by a
separate corporation. The presenters also mentioned that there may be other
management models. For example, the CPN could be chartered by Congress, or
regulated by the Public Utilities Corporation.
A suggestion was also made during discussion that a user advocacy
group to which the board of directors of the CPN was responsible was
necessary.
The presentation was eventually terminated due to time constraints.
The second presentation, entitled "NREN Legislative Update"
was considerably less controversial. In essence, Bill Bostwick
of the Federal Networking Council presented an overview of recent legislative
activity pertaining to the NREN.
It was stated that Authorization Hearings were held on S272 by
the Senate Commerce Committee on March 5, 1991. Authorization
Hearings were also held on HR656 on March 7, 1991 in a joint meeting
of the Senate Commerce Committee and the House of Representatives.
Further hearings will be held by the Senate Energy Committee on April
11, 1991.
The presenter felt that compromises would have to occur in the
definitions of the various roles and responsibilities of the agencies
participating in the NREN: the DOE, NSF, DARPA and NASA.
A description of the HPCC program, now in its 2nd printing with
100 000 copies can be obtained from
PMES,
c/o NSF,
Computer and Information Sciences Engineering,
1800, G Street NW,
Washington D.C. 20550.
It was also suggested that a phone call to either of the following
may also suffice to obtain a copy of the HPCC program description:
NSF (202) 357-7936
DOE (301) 353-5800
The presenter noted that the HPCC was announced through the Office
of Science and Technology Policy, and that the total budget for the year for
the HPCC in the Administration budget was $638 million. The networking
component for the year totalled $92 million.
