The authors propose the use of expert system technology in the medium access control for a LAN (Local Area Network) system. It is a combination of CSMA/CD (Carrier Sense Multiple Access with Collision Detect) and Bus Token techniques under the control of a set of expert rules. The system is event driven and controlled by statistical measurements that are made on the bus traffic.

This problem and its associated domain knowledge are complex enough to be considered for expert system methods. There are other nonexpert methods discussed by the authors that can at least partially solve the problem, but these methods usually have shortcomings and limitations that must be accommodated.

The paper is quite long because of the tutorial discussions on LANs and expert systems. Considering the length, the authors perhaps could have included at the beginning a small Table of Contents; then the various types of readers could skip over parts they are already knowledgeable about and would not have to hunt through the paragraph headings. After the tutorials, the LAN access problem is discussed in good detail with a number of excellent analogies concerning human experience with communication and how we avoid or resolve collisions in our conversations under various conditions.

Next, the problem is described in detail, and a sketch of the model is given as an event-driven, rule-based control that runs in one of three modes: Contention (pure CSMA/CD), Resolution (rules that govern collision resolution and delay estimation), and System Control (rule-based token control). One design consideration was where to locate the expert. The decision was to make each node autonomous with its own expert. The expert is a simple blackboard architecture. Under “details of operation,” the Resolution mode (delay functions, collision management, and “spokesman’s token control”) and the Control mode (overload recognition, token maintenance, and recovery methods) are discussed. Then the knowledge base (facts and rules), the blackboard, and the inference engine are very briefly described.

The last section is a discussion of the proposed system characteristics (apparently it is not built yet), problem areas, and suggestions for further research. The anticipated limitations are also discussed; e.g., the possibility of the size of the expert system causing unacceptable interference/degradation in the system performance.

The best parts of this paper are the problem description and the details of operation. Since the system has not yet been built, detail on its construction or performance is not included. This application of expert system techniques to the LAN access control is a novel one, and a number of good ideas are described. The paper is good, easy to read, and presents a number of good ideas that make one want to set up a system or simulation and try them out]