The Re-entrance Problem and Its Solution

In this paper we take a formal look at the nature of the re-entrance problem. We see the essence of the problem in a conflict between the assumptions components make about each other’s behavior. We argue that to ensure safe composability of mutually dependent components in an open system one needs to reason about them modularly. We propose an approach for extending component specifications with rely/guarantee conditions, explicitly expressing the assumptions components can safely make about each other. Accordingly, we define how to refine components in accordance with the rely/guarantee conditions. By extending the refinement calculus, we define a customized theory of components necessary for reasoning about composability. Then we formally prove that refining components taking into account the rely/guarantee conditions permits to reason about component systems modularly and in this way avoid the re-entrance problem.