Dynamic component and code co-evolution

Abstract

This papers presents a radically new approach for the dynamic evolution of long-lived systems that can not easily be shut-down for maintenance and restarted afterwards. Conventionally, the source code of a software system is viewed as a static entity and separated from the system at runtime. This seems intuitive as a single piece of code is usually associated with multiple components at runtime. Obviously, this viewpoint is a major obstacle for dynamic evolution during runtime as it raises difficult consistency issues concerning the relationship between static code and the dynamically executing system. The evolution approach presented in this paper takes a completely different direction by seamlessly integrating static code with dynamic execution. By this and sound concepts for component categories, incompleteness and dynamic completion, software can be generalized and adapted during runtime in a highly flexible way.