The workshop was organized around six sessions. The goal of the sessions was to stimulate discussion, and so each was started by brief presentations followed by considerable discussion. These discussions have been summarized by session.
The workshop was attended by 64 researchers, resulting in lively discussions. Here were some of the highlights. See the above summaries for further detail on each of the sessions.
The session on High Level Communications was aimed at the lowest level of middleware, providing enhanced communication services. The discussion included two significant issues. One was the need for caching supported by multicast communications. It wasn't clear, though, where such caches should be located throughout the system. Another major issue discussed was the mechanims for providing applications with more information about the availability of services and the associated parameters (e.g. bandwidth, delay). The issue was whether enhanced services at the network level should be attained by the underlying subnetworks providing information about the availability of such services, or if such support should be generated only by making measurements on an end to end basis of what service was available. This was a topic of considerable discussion.
The session on Security-Related Issues asked the question "Are they services, APIs, protocols, or what?" While this is a question that permeates much of middleware, it was of particular interest in the area of security. Another question explored in this session was the trade-off between providing common security services for multiple applications vs. securing the important applications on an application by application basis.
The topics discussed in the session on Distributed Objects and Procedures touched on the systems issues in using distributed objects. The discussion made clear there are multiple technical communities involved and surfaced the question of the role of standards.
The WWW Revisited session raised the question of whether the network service level should provide a richer model of interaction vs. achieving that increased level of service through application level adaptation. Another question made clear through the discussion in this session is that multiple systems are being proposed as the general solution. In other words, several groups are developing frameworks for general interactions over the network that require adoption of those frameworks. Is this a reasonable approach for achieving the needed common infrastructure?
The session on Services for Remote Collaboration focussed mainly on synchronous collaboration. It noted that current techniques mainly involve synthesis of services, but there are some unique requirements for collaboration.
The Naming and Addressing session discussed the needs of middleware and related services for coherent naming and addressing strategies. Significant discussion occurred on the question of the various usages for names and the question of whether that necessarily implies the need for different naming structures.
In summary, the workshop dealt with a number of areas of middleware. To our knowledge, this was the first workshop focussed on middleware, and no doubt will not be the last. It was by no means complete in its coverage. By intent, it raised more questions than it answered. None the less, it provided a useful and interesting forum for discussion of the many issues involved in developing the technology of middleware needed for evolution of the global information infrastructure.
I want to thank the program committee, session chairs, panelists and the attendees for making the workshop a success.
Barry Leiner
Workshop Chair