Abstract
Component software technologies and architectures have grown with great popularity along with the advances of web and internet technologies. The successes are coming from several compound factors. A try in the analysis of these factors might include abstractions of object-oriented concepts which separates the implementation from interfaces, open and distributed component architectures which enables building of cooperative applications, re-usable components which allows managements of software complex, the possibilities of component compositions at runtime which makes software more flexible and extensible. While enjoying great successes, component technologies are with great challenges ahead. This is in part due to an era of the new arrival of computer architectures including the distributed enterprise architectures built with individual components linked by messaging subsystems, the autonomous and grid-like architectures in the web, and the distributed and pervasive architectures assembled by a group of small devices and connected by heterogeneous wireless environments. In this dissertation we argue that additional properties of component software need to be annotated in addition to the separation of implementations from interfaces. We think this is important for component software and architectures to meet the new era of distributed systems. Specifications on software components will allow the adoption of components best adaptive to given environments. Due to object-oriented techniques, the component resources have common information associated with them, the interfaces they implement. We argue that additional descriptions can be added. These properties, we called metadata, address the difference between the implementations of the interface. It helps us search not only the kind of components, but also specific realizations. In spite of the focus on the performance benefits brought by specifying and dynamically changing components, this dissertation also presents many works on providing advanced improvement on commonly used component communication, such as RMI in Java and Remoting in .NET. For example, to enable the optimization of runtime component composition described above in a distributed system, the communication must be handled carefully as well. It is possible that the composition of components occur involving components in different sites. Thus the communication between components should act seemly with external changes. A software layer has been proposed in this dissertation to monitor data transportation. It leads to the work on the roaming support built-in existing component communication layers. We further provide many other features to current component communication methods, which includes embedding data streaming under the programming paradigm of common component communication, as well as using network processors as switching support for component communication.