An understanding of the main features of the Dexter model is necessary in order to appreciate the finer points of this paper (see Halasz and Schwartz, also in this special issue of CACM [1]). The authors used this model as the basis of an open hypermedia system for experimental system development and cooperative design. The system, called DeVise Hypermedia or DHM, uses a shared object-oriented database and is intended for use in application areas such as large engineering projects. The paper reports on the authors’ experiences and makes a number of detailed recommendations and comments on the Dexter model.

Overall, the authors found that the Dexter model provided a robust and workable basis for their system. They found, however, that they needed to extend the link model to handle dangling links and to clarify the notion of link directionality. Problems also arose in the provision of composite objects, the relationships between anchors and composite objects, and the integration of third-party applications. The authors make a number of useful suggestions for improvements in these areas.

The paper provides thoughtful discussions of several important issues. I recommend it as useful material for anyone seriously interested in the design of hypermedia systems.