On the sequential nature of unification

Reviewer: Louis Hodes

Since unification is the basis of logic by resolution, it is central to most current work in logic programming. This is a readable work on the complexity of unification and should be read by everyone involved in that area, especially if they are attempting an implementation on a parallel machine. The authors conclude that unification cannot be significantly improved with parallelism. Their work makes use of earlier efforts in complexity theory, since their proof consists mainly of a reduction of the monotone circuit value problem to unification. They also deal with a special case of unification, term matching, which they show can be improved by parallelism.