The development of a successful approach to the integration of computer-aided design and computer-aided process-planning will require solutions to several fundamental problems in planning and geometric reasoning. In this article, we consider two such problems, both of which arise from interactions among machinable features. The first problem is that interactions can lead to several alternative interpretations of what the machinable features are. We address this problem by developing an algebra of features. The second problem is that interactions among features can cause problems in optimizing the process plans. For this problem, we have developed a branch-and-bound optimization algorithm with good heuristics to guide the search. This article presents the theory and mathematical framework for both problems and also the details of the implementation and experimental results.