On the generation and use of quantitative criteria for assessing software maintenance quality
Robert S. Arnold · 1983
Whenever software metrics are used in assessing the quality of software maintenance, their values must be interpreted relative to the metrics' intended use. Further, these interpretations should lead to concrete suggestions for improving maintenance quality. This dissertation presents a method, Criteria Application (CA), for performing these interpretations and improvements. The principle feature of this method is the development of criteria for modelling software/maintenance quality concepts. A criterion is a testable condition, concerning software metrics' values, with an explicit judgment attached to the condition's outcome. The criteria are used in (1) establishing goals for software maintenance, (2) quantitatively determining whether the goals are met, and (3) guiding remedial work if the goals are not met. In this dissertation we describe when Criteria Application should and should not be used. To illustrate its usefulness, we apply CA to some actual software maintenance data, from the maintenance of scientific application software. We evaluate CA on its potential usefulness for actual software maintenance. Finally, we show how CA can be applied beyond software maintenance to software development.