'. . . and nothing else changes': the frame problem in procedure specifications

The first aim of this analysis is to outline a certain general problem which arises in all formal specifications using the pre/postcondition notation, and which is related to a longstanding problem in the field of AI, called the frame problem (J. McCarthy and P. Hayes, 1969). The authors then present examples illustrating this problem, which becomes particularly acute for large object-oriented specifications where inheritance plays a central role. The examples are intended to demonstrate that failure to deal with the frame problem compromises a formal specification language with respect to its notational suitability and its capacity to support a methodology for formally proving properties of specifications. How existing specification languages have endeavored to cope with the problem are reviewed. A novel approach is presented based on recent work intended to solve the frame problem in planning applications within AI.<
>