[2225] in Commercialization & Privatization of the Internet
Re: Good and Bad Mid-levels - a Reprise
daemon@ATHENA.MIT.EDU (breeden@farnet.org)
Fri Jan 31 16:33:33 1992
Date: Fri, 31 Jan 1992 16:31:34 -0500
To: cook@tmn.com (Gordon Cook)
From: breeden@farnet.org
Cc: com-priv@psi.com, members@farnet.org
Gordon, gentle readers of com-priv,
On the subject of "good" and "bad" midlevels or regionals, and what the
government's policy toward them should be, let me hazard some remarks as
FARNET's executive director.
First, some definitions. FARNET is a non-profit association whose purpose
is the enhancement of research and education through the use of networking
technology. We have thirty-one members, with AT&T the newest. The
demographics look like this:
State networks 14
Multi-state "regionals" 7
National networks 3
Supercomputer nets 3
Universities 1
Provinical networks 1
Metropolitan networks 1
Interexchange carriers 1
Not all of these are networks. Of those that are, some were started with
direct NSF support under the "backbone plus midlevels" model of about 1987.
Others were started with seed funding from other agencies (NEARnet) or
state government. One is a Canadian provincial network (Onet); one (CREN)
predates the NSFNET by at least five years . None of them, to my
knowledge, is fully supported by NSF. All include some mixture of user
fees, state funds, and private-sector foundation funding. Many are
connected directly to the NSFNET backbone, but a number of them are not.
This is why it's not easy to generalize about "the midlevels" or "the
regionals". A reasonable working definition might be the set of networks
that have received direct support from NSF since 1987, either in the form
of cash awards or connections to the NSFNET backbone, under public
solicitations issued by NSF. The four primary solicitations (that I can
remember) are the backbone solicitation, the "connections" program (for
college and university connections to midlevels), the original midlevel
program (for networks which would provide backbone access for a state, part
of a state, or multiple states), and the second round of midlevel awards
(which were awards of new backbone nodes to existing organizations, such as
NEARnet).
NSF did not make its choices in a vacuum. It applied policy and service
criteria to the award of these grants and connections. I don't have any of
the solicitations in front of me as I write, but I am sure others do, and
they are a matter of public record. I was involved in responses to two of
them and have read and critiqued proposals in all four categories.
Important criteria that I recall are technical competence (including
specific requirements for 24 hour operations, uninterruptable power, etc.),
demonstrable need for network access, willingness and ability to become
self-supporting, broad commitment from potential users of the network,
leveraging of local/state/regional funding, and credible plans to provide
user services.
In my opinion the achievements of these programs have been considerable. I
speak from the vantage point of someone who worked on CSNET beginning in
1983, when "the network" meant "the ARPANET". NSF set up CSNET with the
explicit injunction that it must become self-supporting over a five year
period. It did. At its peak, in 1987, it had 200 member organizations,
in 13 countries. Today -- less than five years later -- the Internet has
an order of magnitude more reach than this.
I believe that NSF made two important decisions which contributed
decisively to this growth. One was its insistence on TCP/IP as the
protocol suite of the NSFNET backbone. Today, TCP/IP defines
interoperability. The other was its decision to push this technology out
into the regions, states, universities and colleges through the three-tier
architecture of the NSFNET (backbone-regional-campus), with NSF support at
all three levels.
Did they do a perfect job? Nope. Are we in danger today of being swamped
by our own success? Maybe. Are we working on managing this growth and
making the transition to the Internet (NREN, NPN) of tomorrow? Yes.
The HPCC program calls on Federal agencies to cooperate in the provision of
networking services to their constituencies. NASA and DOE, in particular,
have mission-critical requirements which they would like to meet through
the interim NREN, which is being managed by NSF. FARNET is working closely
with NSF, the mission agencies and the Federal Networking Council to define
precise service and operating requirements and to improve the ability of
all its members (and the rest of the Internet community) to meet them.
Our next meeting, in February, will be a two-day, hands-on workshop on
"hardening the network" in which a broadly representative group of people
work together to analyze some of the things that make the network "soft" in
the first place and look at how to fix them. NSF is supporting this
meeting, and the results will be widely disseminated.
FARNET's recommendations to NSF, issued in November, about post-1992
backbone services explicitly called on NSF to require provider
accountability for performance as part of its awards process. "The
[performance] criteria must be objective and measurable and should be
designed to ensure an acceptable level of service end-to-end throughout the
interim interagency NREN."
It's an oversimplification to think that all of the difficulties we are
encountering in operating large internets are the result of "bad" or weak
organizations. The tools, systems and procedures that are available to us
are not adequate to the task. You can fix some of those problems with more
people, but that wastes human energy, and the labor pool in this industry
is already too small. We need to make the technology better.
Whatever the shortcomings of particular organizations, there is a
widespread commitment among FARNET members to look critically at
performance issues and to work with the NREN community at large to improve
the reliability, stability and usefulness of the network. I do not think
that any of them expect to survive as wards of the Federal government.
By the way, we've thought about replacing "hardening" (which has some
unfortunate connotations) with other metaphors; "tempering" seems to be
the leading contender.
Yours for the well-tempered network,
Laura Breeden
FARNET
><<MESSAGE from>> Gordon Cook 31-JAN-92 12:58
> cook@tmn
> I have had a private suggestion that I used the wrong name for Fedreal
> help to the mid-levels. That it is not a case of the feds bailing out bad
> mid-levels at the expense of good ones, that it is the feds engaged in
> "hardening the regionals."
>
> Correct? Incorrect? Good policy or bad?
