On the Relative Merits of Software Reuse

Using process simulation and AI search methods, we compare software reuse against other possible changes to a project. such as reducing functionality or improving the skills of the programmer population. In one case, two generations of reuse were as good or better than any other project change (but a third and fourth generation of reuse was not useful). In another case, applying reuse to a project was demonstrable worse than several other possible changes to a project. Our conclusion is that the general claims regarding the benefits of software reuse do not hold for specific projects. We argue that the merits of software reuse need to be evaluated in a project by project basis. AI search over process models is useful for such an assessment, particularly when there is not sufficient data for precisely tuning a simulation model.