Support software evolution with abstration rules and programming knowledge patterns

Reviewer: Ed Harcourt

Teng and Liu have the ambitious goal of providing a language and methodology for deriving formal specifications from legacy software systems. The language, reengineering wide spectrum language (RWSL), is intended to be a domain independent target language. The authors posit that a translator exists for each legacy programming language (for example, C++). They then propose a process where a legacy system is automatically translated into a RWSL program, and then processed using abstraction rules to derive a specification in a temporal logic. Additionally, the RWSL program is analyzed manually, to extract programmer intent, by applying a set of knowledge patterns that capture how programmers solve common problems. The process is clearly described, and follows the general outline of other reverse engineering methodologies. The unique aspect the chapter introduces is the use of RWSL, and the conversion to temporal logic. It is difficult to ascertain whether the process can be applied to real legacy systems, since no RWSL translators exist, and the translation in the case study was done by hand. Furthermore, the example case study is rather simple, and clearly not indicative of the complexity one would encounter if the process was being applied to a real large-scale legacy system. However, there is anecdotal evidence that these tools can be written, since the object-oriented community has been providing, for quite some time, reverse engineering tools that generate unified modeling language (UML) models from legacy source code. While the use of the temporal logic to capture specification is sound, the set of knowledge patterns that the authors provide seems capricious, especially since the authors don't mention the vast amount of work in reverse engineering that can be found in the program refactoring literature. This more mature refactoring community has identified a wide array of knowledge patterns that do not appear to coincide at all with the authors' patterns. The chapter also suffers from a variety of grammatical problems, and needs editing. Why the chapter appears in a work dedicated to computational neurobiology is unclear. On the whole, a software practitioner interested in reverse engineering legacy systems is probably better served by refactoring research (see http://www.refactoring.com). On the other hand, researchers interested in reverse engineering will find here a general overview of a viable process, but with lots of missing details. Online Computing Reviews Service