This paper is well written and clear. However, it is not a survey paper and seems quite out of touch with current expert systems work. The only systems “surveyed” are EMYCIN, OPS, PROLOG, and a couple of quite less known system. In particular, only “rule-based” schemes are considered and other approaches, an important one being the blackboard model, are ignored.

It is difficult to see what the point really is of this paper. The author’s classification of control mechanisms seems rooted in state-space search and theorem proving work, and thus has very little to do with most expert systems. He claims, unrealistically, that an expert system builder can follow a “checklist” approach to defining control structures for a particular system. This shows a lack of appreciation for the differences between very small, simple, and well-defined tasks and large, complex domains of application such as medicine. In summary, I do not recommend the paper.