BRIAN KAHIN: I want to pick up on Nina Cornell's concerns about interconnection and the issue of the government's role. It is worth bringing up because it is very often overlooked. It is not even mentioned in the administration's information infrastructure document. The NSF [National Science Foundation] is involved in the program to restructure the NSFNET, which is in effect unbundling the NSFNET as it exists today so that NSF is no longer providing a single package in its main contract but has broken out the different roles of routing arbiter, a very high speed backbone that will be very restricted in use, and the newly conceptualized NAP [network access point].

In principle the NAP is an exchange or switch where anybody can interconnect, and it is a way to get onto the very high speed backbone if you have the right to do that. It is not clear how many NAPs there are going to be. NSF is going to fund them, and so we have the government funding something that is not subject to an acceptable use policy but that is general-purpose infrastructure investment.

To what extent will this architecture catalyze the larger infrastructure? How will the commercial Internet crystallize around this concept? I wonder if we are seeing an inversion of the paradigm from the network as a carrier to the switch as a carrier and what that means for understanding the cost-price differential, because it seems to me it is a lot easier to get at that when the carrier is as simple as a switch. That is not going to solve the "last-mile" problem, but it might be progress.

ROBERT CRANDALL: I still have some difficulty with imagining what the network or networks of the future are going to look like. I do not see any reason why there is one network, given what we know about what is happening to transmission switching costs. I can imagine lots of different networks into which I could connect with a piece of terminal equipment that has a set of switches that can connect me to lots of different networks and lots of different nodes.

One of the problems of this discussion is that it is based on what we know is the current network, the current public switched network developed under regulation. One of the things we know about other industries is that once we deregulate, the technology changes and market organizations change in ways we could not possibly predict. In fact, it may be that regulation is condemning us to one particular model that is incredibly inefficient.

KAHIN: Also, the network that we have been focused on is a homogeneous network, whereas now, especially in the Internet, we are entering an environment of heterogeneous networks, where the networks all differ in capability and functionality and speed.

CRANDALL: That is right, but there could be a bunch of different backbones, too, not only a bunch of different LANs [local area networks].

ALFRED AHO: This panel spoke of some of the experiences learned from regulation in Japan. In the United Kingdom, there is ongoing competition between the telephone companies and