A study of a tool platform for developing formal specification

Since computer systems are pervasive and have a major impact on society, such systems must be built safely and reliably. One of the existing approaches to this issue is to formally model (the designs of) such systems and formally verify that they have desired properties. Computer systems are often modeled as transition systems. If a computer system can be modeled as a finite transition system, model checking techniques may be the most useful. Otherwise, theorem proving techniques should be used. In our approach, a computer system is modeled as an OTS (Observational Transition System), a kind of transition system, which is described in CafeOBJ. It is then verified that the systems have desired properties by writing proof scores in CafeOBJ and having the CafeOBJ system execute (or rewrite) the proof scores. We have demonstrated the efficiency of our approach by performing several case studies. However, proof scores have been entirely written by hand, which means that human errors might have occurred. To minimize human errors, writing proof scores should be mechanically supported. We think that the difficulty to develop tools treating CafeOBJ specifications causes lack of such tools around CafeOBJ. Since CafeOBJ’s users can speficy that their own syntax of terms such as _+_ or if_then_else_fi (An underscore _ indicates the place where an argument is put), we cannot parse specifications with parser generators such as Yacc. To solve this problem, we propose an XML scheme called CafeOBJ/XML as a tool platform and present a tool called Buffet which converts the original specification to the XML version of it. The main advantage of this idea is that we can use common XML parsers to parse specifications, and most programming languages have such parser. This advantage makes easy to develop such tools. In this paper, we also presents three applications to demonstrate that the proposed scheme is a suitable for such platform. One of these applications can generate proof scores automatically, it is guaranteed that generated proof scores cover all cases, excluding human errors.