Summary form only given. Adapting a component-based application means adapting one or more of its components, and in general, adapting a component at run-time means disconnecting it from the application and connecting a new version of this component. Many approaches like OLAN and DCUP are based on component models in which each component is constituted by a functional part and a control part. This last part provides a set of services to administrate the component. In our approach we aim to separate the control part outside the component. The benefit is the possibility to apply our solutions to existing models that have not been designed to support the dynamic adaptation. Through a first experimentation on the JavaBeans component model, we define an adaptation process composed of several steps: (1) defining the mapping rules between the old and the new component (correspondence between properties and services); (2) passivating the old and the new components; (3) transferring the state of the old component into the new one according to the mapping rules defined before; (4) disconnecting the old component and connecting the new one; (5) activating the new component. These steps may constitute the basic activities of an adaptation machine.

Building large software applications from Commerical Off The Shelf tools (COTS) is not an industrial reality so far. This work presents a new approach to solve the different problems found when building a federation of COTS.

A software process model is a complex entity describing activities, products and quality aspects of the "real" software-production process. The ever-changing nature of the software process imposes a continuous control of its evolution; maintenance should adapt the process model to accurately reflect the state of the real process. Presently, process model maintainers have to make evolve these complex models under time pressure and with very little assistance. In this paper we present a preliminary approach for improving maintainability of Petri-net-based process models using the concept of view. A view can be seen as the sub-model capturing the perception a participant has about the process. We show how views can structure a model as a set of more abstract modules which are easy to make evolve. After evolution, views can be composed using a set of formally defined operators to obtain a new process model in an automatic way. Our study is applied to the Process Weaver modelling language. After a short presentation of the Process Weaver formalism, we illustrate view definition with an example of a review process model. View evolution and view composition are also introduced through a scenario of process model change based on the above example.