[8251] in Commercialization & Privatization of the Internet
Re: ANS, the CIX, the rain, the park and other things (fwd)
daemon@ATHENA.MIT.EDU (Hans-Werner Braun)
Fri Nov 12 11:11:51 1993
From: hwb@upeksa.sdsc.edu (Hans-Werner Braun)
To: com-priv@psi.com
Date: Fri, 12 Nov 93 8:10:58 PST
>Beg your pardon Ittai but according to material given by MERIT to NSF
>and released to me by NSF under FOIA almost a year ago, there is no
>difference between the NSFnet "services" and the ANSnet backbone. Your
>response to Karl would lead readers to believe that there is someway
>to distinguish between the two forms of traffic. My research indicates
>that this is not correct.
This is all (not only the attached note) getting relatively boring, as
it had been discussed so many times already. For one thing I guess it
was not *that* important to NSF back in 1990 how the NSFNET awardee
exactly implemented its service provisionings and partnership
relationships, as long as there was a likelihood of services being
provided as required, with a potential for further improvements in the
service qualities (bandwidth, ubiquity, prospect of driving things so
enough stuff is in place (not only with the awardee and its
partnership) to have a sustainable environment after th five years,
etc.).
Regarding the NSFNET/ANSnet differentiation, you can really make a
point any way you see fit, and it becomes a somewhat irrelevant
argument. The same switches and routing tables are being used, and
there is at least an ability to vary policy at least at the periphery
by filtering inbound or outbound routes. So, it is a little (though
perhaps not much) more than blind trust that NSFNET sites are
containable within the confines of NSF policy. At this time, NSF's
paradigm is that they pay for a service on a larger infrastructure.
Besides that, I guess the closest parallel coming to my mind on short
notice would be if Gordon would ask ATT (random example) for the
equivalent of an 800 service (where he bulk (actually, centrally) pays
for calls), but with the modification of not being only going to a
single (800 number) point (many-to-one mapping), but, say, many 800
numbers (without being 800, but normal, numbers) are able to
communicate, while centrally paid (many-to-many mapping). GordonNet
(the many-to-many 800-style network) may get a good deal from ATT as
Gordon makes a case to them that he is not selling services at large,
but for a good cause and a restricted environment (e.g., everyone
promises only to use it for the making and distribution of Gordon's
Cook Book). In such a case, ATT would not have to change anything on
its network physically. All it would have to do is do the billing to
Gordon's Hideout, rather than the individual "users." So, from
Gordon's point of view, he would have a "network service," called
GordonNet, which is physically mapped into the larger ATT network. He
could even expand the scope, perhaps paying additional money to his
service provider, to create a newletter (say, the Internetional
Enquirer) after being done with his book. So, he is paying for a
service for provisioning of a network to him. Is it his? Is it
dedicated? Separate fiber in the ground? Will he always be able to make
the calls he wants without getting busy signals? WiIl he have to defend
himself on rsrch-pub@nsf.gov every other month? Only the IE will
know.
Hans-Werner
