After 4 years of operation the NERComP network is now a self-supporting success. Some of the reasons for its success are that (i) the network started small and built up utilization; (ii) the members, through monthly trustee meetings, practiced "participatory management" from the outset; (iii) unlike some networks, NERComP appealed to individual academic and research users who were terminal-oriented and who controlled their own budgets; (iv) the compactness of the New England region made it an ideal laboratory for testing networking concepts; and (v) a dedicated staff was willing to work hard in the face of considerable uncertainty. While the major problems were "political, organizational and economic" (1) we have found that they can be solved if the network meets real needs.

We have also found that it is difficult to proceed beyond a certain point without investing responsibility and authority in the networking organization. Conversely, there is a need to distribute some responsibilities such as marketing and user services back to the member institutions. By adopting a modest starting point and achieving limited goals the necessary trust and working relationships between institutions can be built. In our case the necessary planning has been facilitated by recognizing three distinct network functions: governance, user services, and technical operations. Separating out the three essential networing tasks and dealing with each individually through advisory committees, each with its own staff coordinator, has overcome a distracting tendency to address all issues at once. It has also provided an element of feedback between the end user and the supplier not usually present in networking activity.

The success of NERComP demonstrates that a distributive-type network can work. Our experiences in New England—which, because of its numerous colleges and universities free from domination by any single institution, is a microcosm for academic computing in the United States—indicate that such networks are best structured in a hierarchical form. This suggests that national networking should be based in part on the more than 30 existing state and regional networks (15).

With the groundwork now laid, we expect to see links among existing regional networks to complement development efforts now occurring at the national level. With Greenberger and others, we believe that one or more networking organizations devoted to the management issues discussed in this article will be required to facilitate resource sharing on a national scale. Because of their experience with these problems and their ability to provide service in many areas of the country through existing facilities, regional networks have a major role to play.