Industrial software evolution processes are, in general, complex
feedback systems. Recognition of this opens up opportunities to
achieve sustained process improvement. The system dynamics
approach was used in the FEAST projects to model behaviour of
key process and product attributes, yielding foundations for a
paradigm to achieve dynamic models of software evolution
processes. It is based on top-down model refinement of a high
level abstracted view of the process. Resulting models are
relatively simple, when compared to others in the literature. Their
simplicity, however, facilitates understanding, progressive
refinement, subsequent validation, and industrial take-up.
The paper illustrates the paradigm with a model of a process
highlighting a particular issue. The results of executing this model
indicate how one may analyse and eventually optimise division of
resources between progressive activity such as functional
enhancement and anti regressive activity such as complexity
control. The approach and model are based on empirical
observations and interpretation of the evolution phenomena. When
appropriately refined and validated to reflect a particular process,
a model such as this can more generally be used for release
planning and process improvement. The approach, together with
its models, provides the basis for management technology and
tools to achieve sustainable long-term evolution of complex
software and systems.