7/31/92

Comments of EDUCOM Networking and Telecommunications Task Force on
National Science Foundation Draft Solicitation for NSFNET

I. CONTEXT
The draft solicitation does not adequately provide a context for the
next iteration of NSFNET and its relationship to other federal
programs such as the IINREN, Gigabit testbed program, NREN itself,
commercialization, and the rapidly developing call for a national
information infrastructure.  Since the funds proposed to be
allocated to the new cooperative agreement(s) are so limited, it is
necessary to clarify these relationships and to indicate NSF's
priorities and goals for NSFNET. This is especially important in the
current atmosphere, which has increasingly and mistakenly assumed
that the NSFNET effort is responsible for accomplishing broad data
network connectivity for the nation's business and public sectors.

II. STABILITY AND CONTINUITY
In contrast to the situation which prevailed in 1987, NSFNET now
supports a large and diverse community of research and education
network users who are very dependent on the services provided by
NSFNET.  It is essential that the new award(s) provide for a smooth
transition to the new architecture and new provider(s), including
appropriate test regimes.  The requirement for stability and
continuity of service should be explicitly established by NSF as the
governing priority in making schedules and establishing budgets for
the transition period.

III. ROUTING AUTHORITY
Routing has become a preeminent issue in the evolution of the
Internet.  The routing authority described in the solicitation is
too narrowly defined and ill situated organizationally to fulfill
the critical role required over the next several years as the
Internet adopts new routing and transmission technology, new
addressing protocols, and continues to grow at double digit rates.
There is no intrinsic reason, and a doubtful economy of scale, in
combining the RA with the NAP provider.  NSF should place the RA
into another agreement, or possibly combine it administratively with
the new NIC award.

IV. NETWORK ACCESS
Although the Network Access Point (NAP) architecture described in
the solicitation and the companion Aiken, Braun, Ford IINREN
Implementation Plan provides some short term relief to several
current routing and network interconnection difficulties, the scheme
leaves open the important issue of how many NAPs will be needed,
both initially and over the course of the agreement.

There is a high probability that the high speed backbone will be
based on broadband public services of one or more interexchange
carriers, and that the bandwidth offered to NSFNET will be shared
with other transport services of the carrier(s).  If this is the
case, transport services will be widely available, perhaps in every
LATA.

The ABF study, which describes the desirable migration of the
Internet toward a broadband mesh network, also cautions that routing
technology may limit such progress in the short term.

These issues, i.e., short term routing restrictions versus the
prospect of widely available transport services, have important
topological and economic impacts on the end to end services of
NSFNET.  NTTF believes that NSF should undertake a review, with
community participation, of the appropriate number of access points
to the high speed backbone, and accompanying increase in routing
complexity, and establish criteria for evaluating proposals for
establishing access points.  The review should also address whether
NSF can safely transfer the responsibility for funding and operation
of the access points to the midlevel and private networks, thus
eliminating the need for direct NSF involvement in this aspect of
the evolution of the network.

V. VERY HIGH SPEED BACKBONE NETWORK
Several carrier representatives have suggested that the proposed
operational and financial separation of backbone and NAP services
will increase the total cost of NSFNET to the government and network
users.  In order to provide a basis for evaluating this issue, the
solicitation should be revised to permit bidders on the cooperative
agreement to propose both bundled and unbundled backbone and NAP
services.

Although the solicitation speaks of end to end services, the
description of the very high speed backbone network (vBNS) covers
only a limited number of connections to the proposed NAP's and the
NSF supercomputer centers, leaving a substantial gap between the
NAP's and many end users.  If NSFNET is intended to support NSF
research projects in a wide range of institutional settings,
including those which are part of HPCC Grand Challenge applications,
then access to the high performance backbone must be available on an
end to end basis.  It is not reasonable to assume that existing
midlevel and campus networks will be able to step up to the
challenge of continuing to offer leading edge network services
without a continuation of NSF support for their role in the delivery
of end to end services.

EDUCOM NTTF recognizes that there is an inherent tension between
needs for end to end high performance services, and needs for
additional connectivity at low and medium speeds within the research
and education community served by NSF.  The solicitation needs to
provide additional clarity about how progress can be made on both of
these objectives within the limits of available funds.