A compositional approach for analyzing dependencies in component-based systems

Software components have become a mainstream technology in the software industry. However, the development of component-based systems poses some interesting problems, including how to effectively perform system-level dependence analysis.; In systems built from components, for example, one cannot easily determine whether a component can affect other components and the system as a whole. As with many semantic questions about software systems, exactly how part of the system can affect and/or be affected by another part generally cannot be identified. It is possible, however, to recognize some conditions that may lead to these effects. Those conditions are related to the set of dependence in the component-based system, which is the main focus of this dissertation.; Understanding the set of dependence is essential for several tasks during the life cycle of a component-based system. For instance, dependence knowledge can be used to evaluate the potential side effects of system modification and to reduce the scope of system testing.; This dissertation contributes a novel methodology for analyzing dependence in component-based systems. The main aspects underlying this methodology relate to (1) providing mechanisms to acquire dependence of a software component and to deploy the results with the component in the form of metadata, and (2) performing system dependence analysis based on the deployed metadata of each of the components that compose the system rather than on the component itself. System-level dependence analysis is important to establish the conditions for proper and reliable development and maintenance of a system. Using dependence metadata rather than the component itself to perform such system-level dependence used to analyze its dependence, often are not available during deployment or composition time, when an effective system dependence analysis should be performed.; The unique characteristic of the methodology developed here is the use of the Pomsets (partial-order multi-sets) formalism to represent the possible effects of the component's services. Pomsets descriptions are automatically captured from the component implementation and included in the dependence metadata. Pomsets is also the basic formalism used by our methodology to combine the dependence metadata of the diverse components of a system.; The expressiveness and practicality of our methodology is explored through three case studies. Those studies are used not only to empirically evaluate the approach developed, but also to provide an analytical basis to confirm the claims of this dissertation. The focus of the evaluation effort is to validate our methodology with respect to three main points: (1) applicability, (2) usefulness, and (3) scalability. Those points are examined based on carefully formulated hypotheses.