This report consists of three parts: (a) a summary of an empirical investigation on current software development practice, (b) the presentation of a new software process model based on its results, (c) a discussion of the consequences for managing projects which follow an evolutionary, object oriented approach.
In an interdisciplinary field investigation, software developers, managers and users were asked about their work procedures and conditions. This research aimed at obtaining a status of current software engineering practice and getting insights for future work design and organisation. Among other results, a significant discrepancy was found between the officially prescribed phase-oriented software life cycles and the work processes actually practiced. Formal project planning turned out to be less important for project success than a flexible and anticipating task management.
In order to re-harmonize theory and practice and also to cover new requirements resulting from a paradigm change towards object-oriented techniques, we propose a model for
evolutionary, object oriented software development (EOS). This model is based on a framework of cycles and activities bound to the
objects
of development as, for example, classes and system components. It is shown that management procedures built on this framework allow a multi-level, differentiated planning and control. This implies new challenges on project managers but it enables them to plan and to act in a flexible, dynamic and more adapted way.