Implementation of Component-based Systems by Systematic Refinement and Translation Steps

The component paradigm promises to address many of the productivity and quality problems currently faced by the software industry. It is hindered in doing so by the "binary-module" model of components that prevails in contemporary component technologies as well as by the limited support for systematically implementing UML models of higher level components. This paper describes a technique, known as SORT, which is part of a larger method for component-based software development, known as KobrA. KobrA is based on the development and deployment of component-based frameworks in conjunction with a high-level, UML-based representation of components. It provides a reasonable degree of support for implementation that is based on the concept of strictly separating refinement steps from translation steps, and on the provision of patterns to guide the application of these steps. The method also defines a profile of the UML, known as the Normal Object Form (NOF). The NOF defines the appropriate level of abstraction at which refinement should end and translation into a tool-comprehensible form should begin. In addition to the usual source code, therefore, the implementation of a component will typically also contain refined, "implementation-level" UML class and object diagrams to describe the "as is" implementation of the application.