Towards a benchmark for the evaluation of software testing techniques

Despite the existence of a great number of software testing techniques we are largely ignorant of their respective powers as software engineering methods. It is argued that more experimental work in software testing is necessary in order to place testing techniques onto a scale of measurement, or classify them in such a way that is useful to the software engineer. Current experimental practices are examined using a parametric framework and are shown to contribute little towards a cohesive and useful body of knowledge. The idea of a benchmark repository of faulty and correct software is explored enabling unification of diverse experimental results. Such a unification should start the process of moving towards an evaluation taxonomy of testing methods.