Search

This article is a reply to comments of computer professionals Paul B. Moranda regarding the paper "A Review and Evaluation of Software Science." The authors of the article say that whether a complexity measure should be intensive or extensive is an interesting philosophical question. There are many good and valid reasons for the intuitive appeal of intensive measures. However, it was not the purpose of the paper to evaluate the intuitive appeal of one complexity measure or another. Rather, the authors were simply saying "Here is a theory of software complexity that has been applied to several large real world data bases and subjected to carefully controlled experiments and it still holds up. We think the theory is worthy of more careful scrutiny and analysis." They leave to some other ambitious author the task of answering "why" one theory works and another does not. It is true that there are many ways to compute the effort measure E. It may be true, as Moranda suggests, that one may choose among alternative E measures, but the subjectivity of the choice does not contradict the objectivity of the measures.

This article presents comments of the author regarding the paper "A Review and Evaluation of Software Science," by A. Fitzsimmons and T. Love, published in the March 1978 issue of the periodical "ACM Computing Surveys." The article raised two questions in author's mind: whether the effort measure is a good measure of complexity and whether the correlation coefficient is a reliable tool for validating the conjectures of software science. Article's abstract states that software science's effort measure (E) gives an objective measure of complexity. E is an extensive measure that is, one that depends on the bulk or size of the program. All previous measures of complexity have been intensive that is, they depend on the internal structure. Examples include the relative frequencies of branching statements and calls, or a normalized to unity spectrum of program's listing across indenture levels, or a cyclomatic measure of program's loop and nesting structure. Indeed, T.J. McCabe, a software engineer has expressed skepticism of extensive measures.

This article presents comments of computer professional T.P. Baker regarding the paper "A Review and Evaluation of Software Science." He was disturbed by the seemingly uncritical nature of the evaluation given software science. When combined with the definition of effort, the definition of language level leads to an inconsistency.