Towards Integrating Multiple Perspectives by Distributed Graph Transformation

In order to support multiple perspectives in software development one needs a scheme which expresses explicitly all the views held by the various stakeholders like requirements engineer, software architect, client, user etc. The ViewPoints framework has been developed in the past as a conceptional framework for expressing such a multiple perspective setting in software development projects. In this contribution we describe how this framework is formally described by distributed graph transformation and we demonstrate the applicability of our approach by presenting a non-trivial sample system. Introduction and Related Work System design is a staged process as for example the spiral model of software development suggests. The various development stages are visited more than once. In such a setting multiple participants collaborate to construct a wide range of development artifacts. In addition, multiple components of the software system under construction must interoperate effectively to achieve the desired behaviour. Especially for large projects various participants with different needs and even conflicting views are involved. In order to support such a setting the ViewPoints framework was developed which pursues the idea to express the various views of the stakeholders explicitly. This also involves to tolerate inconsistent information in related ViewPoints until it seems necessary or appropriate to check and (re)establish consistency at least in some parts of the system [4]. The ViewPoints framework has been used quite successfully and has been documented in the literature [10]. Especially in industrial interdisciplinary projects where engineers of different disciplines have to work together such a framework seems to support the actual development process quite well. E.g. the design information a production automation engineer creates has to be set into relation to the software architecture a software designer has to produce in order to control the automation plant. This involves to loosely integrate quite different notations and processes. The question which is addressed here is how tool support can be constructed to effectively represent the loosely coupled approach: some local development within a ViewPoint is followed by interaction with related ViewPoints via consistency checks. The approach of distributed graph transformation supports the idea of loosely coupled ViewPoints as outlined above quite naturally. It realizes the separation between the independent development of single local ViewPoints and the configuration and connection of a set of related ViewPoints in a structured way. The concepts as well as the formal definition of distributed graph transformation are based on the double-pushout approach to algebraic graph transformation [1] where basic concepts from category theory are applied. Distributed graph transformation is introduced formally in [11]. The ViewPoints framework was devised by A. Finkelstein et al. [5] and B. Nuseibeh [10] to describe complex systems. An overview of other approaches related to multiple perspectives in software development can be found in [6]. In [10] a general overview wrt inconsistency management is given. This work used a logic approach to describe inconsistency check actions between ViewPoints. In contrast to our approach, this logic approach is only used for the definition of ViewPoint check actions. Our approach uses graph transformation to provide the entire ViewPoints framework with formal underlying semantics. Thus, not only check actions are formalized, but general benefits of formal specification are achieved for the entire ViewPoint architecture (cf. conclusions). In the chapter Basics we will introduce briefly the basics of our approach: first an overview wrt. the ViewPoints framework will be given and then we will sketch how it can be formalized by distributed graph transformation. Based upon this we will present an example in the chapter A Sample System: Integration of Software Architecture and Performance Evaluation.