Session summary : Session on Architectural Modelling
Jeff Magee, Imperial College London
Three papers were presented in this session. The first paper on “Analysis of SW architectures in High- and Low-volume electronic systems” presented by Henk Obbink of Philips Research instigated a discussion on the need for multiple viewpoints when modelling software architecture. The Soni architectural framework presented by Henk identified four main views of a software architecture: Conceptual architecture, Module architecture, Code architecture and Execution architecture. It was agreed that multiple viewpoints are essential when modelling sogtware architecture, however, no consensus was reached on what those viewpoints should be. A number of participants felt that the structural view of a software architecure described in terms of components and connectors was the central view and other views where subsidiary however, other participants felt that the central view depended on the class of system being modelled.
The second paper on “A Practical Unifying Active Architecture Concept” presented by Richard Juren of MacDonald DettWiler raised the question of what constituted an architectural style. It was felt that the architectural model described by Richard in his presentation was really an architectural style ( or more accurately a set of styles). This led to a discussion of what analysis could be done with respect to an architectural style as opposed to a specific instance of that style. Richard felt that the architecture presented could be used to informally analyse many of the extra-functional properties required such as flexibility, scalability and extensibility. Many participants agreed that this constituted the main value of this sort of architectural model.
The third and last presentation was on “Framework for documenting design decisions in product families development” by Alejandro Alonso of Universidad Politecnica de Madrid. This focused discussion on architectural modelling of product families. What constituted the common architecture for a family of products and the relationship with the architecture of variants was discussed together with the drivers for variation. It was felt that it many cases because of the adhoc process of generating variants which constitutes current practice, a common “family” architectural model does not usually exist. However, the value of such a model was agreed by all participants.