An experiment to determine software reusability factors (abstract)

Software reusability has been proclaimed as the common sense solution to many software development problems. The concept of reusability promotes productivity because it avoids “reinventing the wheel.” Using existing components which are similar to the current needs can be much faster than creating components from scratch. Reusability can also be viewed as promoting reliability since reused components have the benefit of both experimental and field testing.

However, reusability has not fulfilled its potential for revolutionizing the software development industry. Identifying the factors which cause current reuse efforts to fail is essential to its later success. Likewise, identifying the factors that seem to promote successful reusability is equally important. Furthermore, practical ways to eliminate the detrimental factors must be developed.

An experiment designed to ferret out the causes of software reuse success and failure must consider several important issues: (1) The experiment must consist of actual development and reuse. Questionnaires and subjective measurements about whether to reuse, etc. are necessary but not sufficient. (2) The experiment must be greatly controlled to avoid extraneous factors from skewing the results. Factors which might influence the outcome must be deliberately tested for, or controlled such that they do not bias the experimental data. (3) The components to be reused must be determined. Reusing requirements and designs has been suggested, but with little success. On the other hand, reusing test cases has been greatly successful. In between is code. Current experiments should still concentrate on the ability to reuse source code. You must walk before you run. (4) Finally, the factors being tested must be established and they must consider two main tangents. First, specific factors concerning the code characteristics, the organization of components, and the development environment must be considered. Other concerns deal with the human factors. Predisposition, ego, training and skill must be taken into account for an accurate study of reusability.

A current reusability experiment concentrates on the use of an object-oriented organization scheme, reusable code characteristics, and several human factors. The experimental subjects actually design and implement code under varying conditions. Subjects are divided into groups that must reuse whenever possible, may reuse if desired, and cannot reuse at all. Comparing the results of the various groups will lead to a better understanding of the problems faced in software reusability.