Workshop on Middleware

Session Summary

Services for Remote Collaboration

Coordinator: Werner Vogels (Cornell)

Speakers: Eve Schooler (ISI), Hans-Peter Dommel (UCSC), Michael Weber (Univ. Ulm) and Farnam Jahanian (Univ. Mich.)

Scribe: Lewis Girod and Bienvenido Velez-Rivera


One of the most promising properties of technology based on ubiquitous networking is the ability to make people collaborate in many new ways. Collaboration can be made more effective by removing boundaries in time and distance and by introducing novel techniques that would not be available in traditional settings.

The task of the panel was to investigate what Middleware should offer to build collaboration systems in such a way that they might scale in number of participants, in geographical location and in concurrent management of shared resources.

The panelists were recruited from the workshop participants and requested to prepare brief statements on a number of topics relating to the issues of scale. The presentations and following discussions focused on the problems that are encountered when building large scale systems for synchronous style of distributed collaboration. In their presentations the panelists showed the most challenging problems are in the area of scaling of real-time access to shared resource in geographically divers locations. A large number of these problems can be seen as generalization of problems that are being solved in the many areas of research in distributed computing


There were two major controversies between the other workshop participants and the panelists.

First, there was the issue that a number of participants was afraid that an approach that was too technology based would cause collaboration systems to be built bottom up, which was believed to be the wrong model. The active involvement of humans in the collaboration process warrants an approach that is human centered and where the system technology should be used to implement styles of collaboration that would match the current traditional techniques. There was a strong belief that by having the design of collaboration systems be dictated by technological advances, the resulting systems would not be useful or acceptable for human collaboration.

The second controversy dealt with the different styles of collaboration. A number of workshop participants were "upset" by the lack of attention the panel had given to the asynchronous/non-real-time styles of collaboration systems. The success of a number of research systems, the trials at the National Performance Review and the commercial success of Lotus Notes (TM) made the participants believe that solving a number of real-time distributed collaboration problems would not be relevant for the advances in collaborative systems. The most pressing problems to be solved now would be how to make the current asynchronous systems interoperate. Not everybody was convinced however that all fundamental problems of scalability were already solved for asynchronous systems. A number of new resource sharing techniques needs to be developed by the distributed systems communication that need to be applicable to make both synchronous and asynchronous systems so they can address the issues of scale in a more fundamental way.

The session ended with a heated discussion on the capabilities of humans to be able to collaborate electronically and how humans deal with technology and its failures.

Return to the Workshop on Middleware Homepage.