Distributed systems are inherently complex, and therefore difficult to design and develop. Experience shows that new technologies-such as components, aspects, and application frameworks-can be effectively used for building distributed applications. However, our experience also shows that most of the applications built in that way are difficult to be re-used, documented, and maintained. Probably, one of the major reasons is the lack of a clear separation between the concepts used at different levels (application domain, application architecture, supporting application platform, programming language, etc.). In this paper we present our experience with a platform we developed for building distributed applications using components and aspects. In particular, we show how many of the (conceptual) problems we hit when trying to document, re-use, and implement it in different contexts can be naturally solved with the adoption of the MDA concepts. In addition, we describe the process we followed for identifying and separating the entities that live in different "models" (in the MDA sense), and the required transformations between them. MDA offers a good framework for specifying different views of our model, and mappings to platform-specific profiles. In this way, we are able to address the particular needs of different stakeholders: from the designer interested in developing new applications following our (component and aspect-based) modeling approach, to the software vendor that wants to implement a proprietary version of our supporting middleware framework in CORBA, EJB or .NET.