A Method for Porting Software Using Formal Specifications

Abstract: Formal specifications are mathematically based techniques with which a system can be analyzed, and its functionalities be described. Case studies have shown that using formal specifications can help reduce bugs and other inconsistencies when implementing a complex system; they are more likely found during the software design phase rather than later. During the process of porting code, testing has been used to verify that the port has the same functionalities as the original. However, testing alone has been deemed necessary but not sufficient to accomplish this. This thesis questions if formal specifications could be used during the process of porting code to create an accurate model of the system, and thereby provide higher degrees of certainty that the final product conforms to the original. A step-by-step methodology is presented to answer this question. The methodology ascertains the behavior of a port target through testing and a formal specification model based on these tests is created. This model is then used to create the port. The result indicates that the methodology used has some potential since it provided a high level of certainty that the ported code adhered to the original. Since the methodology puts a high emphasis on the specification and has several layers of verification, it is likely that it is suitable for projects with several modules and interdependencies. When using it for porting a trivial or non-complex system, the overhead of the methodology may prove high in comparison to the value gained. It was also found that one must take into consideration the implicit functionalities a language provides. Strict reliance on a model could thereby lead to a less flexible process where creativity and consideration of the specifics of the target language may have produced a different result.