Component Configuration Management

Component-based programming is now a widely recognized approach in software development. There remain many open problems related to both technical and non-technical aspects of the components. In this paper, we point out the problem of component identification. Since the components are usually binary units deployed in the system at run-time, we do not have the same insight into their characteristics as into those of the software units which we manage at development time. This problem could be solved if the components had this information integrated together with the binary code, which can be achieve by defining a standardized identification interface. As such interfaces do not exist in standard component models today, this concept can only be used with components built inhouse. For external components, extensive tests can, to some extent, compensate for the lack of information. To perform a successful testing efficiently we must limit the number of test cases. Which parts of our system can be affected by the introduction of a component, or by its updating? We can answer this if we can keep track of changes introduced in the system and their impact on the system. These problems are similar to the problems arising at development-time solved by Software Configuration Management (SCM) disciplines. In this paper we point out these problems and make proposals for their solutions at run-time using SCM principles.