Key points are not available for this paper at this time.
A key problem in component-based software development (CBSD) is that developers have incomplete knowledge of components. In many cases, the only available source of such information is experimentation. In this paper we argue that the provision of tool support for automated and repeatable experiments can provide significant value to designers. Such tools, which we call exploration harnesses, promise to help enterprises to exploit prefabricated evolving third party components. We evaluated the exploration harness concept by building a prototype and using it to support the exploration of large components in the design of a dynamic fault-tree analysis tool called Galileo. Galileo employs package-oriented programming, in which shrink-wrapped packages such as Microsoft Word and Visio Technical are used as large components. Using our exploration harness helped us to discover a range of relevant but undocumented properties of such components, across several versions, which enabled us to make better informed design decisions.
Copenhafer et al. (Mon,) studied this question.