The Virtual System Model :

1 Motivation In recent years, many researchers have explored mechanisms that allow systems to span large networks. Few, however, have looked at the problems that will confront the users of these highly distributed systems. There will be a huge amount of information available within such a system, and that information will be scattered about. Even if users can access it all, there will be so much that is irrelevant that it will be diicult to identify the pieces that are of interest. This problem has two aspects. First, users must be able to keep track of objects about which they already know and in which they have indicated an interest. Second, users must be able to identify objects of potential interest about which they do not already know. The mechanisms that solve the two sub-problems should work together; users are often willing to spend time organizing information for their own use, and it should be possible for the meta-information, so speciied, to be shared with others. In recent years, several approaches have been devised to address these problems, but these approaches break down in environments that scale both geographically and administratively. Among the approaches that have been proposed are user-centered namespaces, databases, attribute-based naming, resource discovery agents, and a uniform global namespace. I will describe these approaches and the problems with each. I will then describe my own approach, and a preliminary implementation testing the concepts of that approach. I will end by succinctly stating my \thesis" and describing what I propose to cover in the dissertation.