An experimental examination of the role of re-engineering in the management of software quality

This paper reports on the design and the results of a randomised, controlled experiment estimating the effect of predetermined changes in module complexity on the maintainability of different program versions seeded with equivalent logic errors. The experiment measures maintainability which is a defining sub-attribute of quality. The hypothesis "low module complexity results in high maintainability" is tested experimentally by monitoring and recording the time taken to identify and correct the seeded errors. Prior to the experiment programs are statically analysed to produce measurements of internal sub-attributes of the fundamental attribute of complexity. A first program version is modularised according to established rules giving a new version with a larger number of modules and with smaller individual module complexity. The results of this work can be used to design tools capable of providing an indicator, or factor, for re-engineering whereby a given program can be restructured in such a way that quality improvement can be quantified or at least estimated. As maintainability is a defining attribute of quality the insights gained can be further applied in understanding the underlying processes involved and ultimately lead to quality improvements