In this article the author discusses the problems of software construction by configuration and explains why this is a challenging area for software engineering researchers. The author believes that the software maintenance research community has a key role to play in this research. Although much maintenance research has been focused on code and code analysis, the community has addressed issues of dealing with software without specifications and where the internal workings are not readily understandable. Some maintenance tools and techniques, such as program visualization, point the way to effective support for construction by configuration. The author concludes by setting out a research agenda that allows the research community to play an active part in addressing the 21st century software engineering problems.