Point/Counterpoint

We have the technology to produce software that has predictable behavior, but doing so requires a better understanding of the economics of confidence and better integration of architecting and programming. I have a long-standing interest in understanding how software components (for present purposes, implementations with interfaces) influence software design. From 2002 to 2008, several colleagues and I at the Software Engineering Institute explored how to combine software architecture and software components such that a system design specifies what architects must know and trust about the components, how this determines the kinds of architectural analyses they can perform, and what measure of confidence they can associate with the results. We referred to this combined capability as predictability by construction (PBC).