Abstract. As development techniques, paradigms and platforms evolve far more quickly than domain applications, software modernization and migration, is a constant challenge to software engineers. For more than ten years now, the Sodifrance company has been intensively using ModelDriven Engineering (MDE) for both development and migration projects. In this paper we report on the use of MDE as an efficient, flexible and reliable approach for a migration process (reverse-engineering, transformation and code generation). Moreover, we discuss how MDE is economically profitable and is cost-effective over the migration through out-sourced manual re-development. The paper is illustrated with the migration of a large-scale banking system from Mainframe to J2EE.

Processes are a major concern of information system. They need thus to be defined as precisely as any other artefact of the system. Meta-modeling techniques have proved their relevance for addressing such issues. However, a process meta-model may not be restricted to the definition of a set of concepts, relations and assertions. As the purpose of a process is to be executed, the underlying execution rules have also to be made explicit. In this paper we attempt to identify the main characteristics of executable meta-models. This study is illustrated with a concrete example, a Petri nets meta-model.

No abstract

A new information system landscape is emerging that will be more model-centered than object-oriented, characterized by many models of low granularity and high abstraction. These models may describe various aspects of a system such as products, structural organization and processes. This latter category may be refined into manufacturing, software and business processes for example. Each of these sub-domains share common concerns such the definition of work items, performers, inputs and outputs, etc. However, as they may be represented using separate meta-models, there is therefore a risk of facing an uncontrollable number of meta-models that all redefine similar modeling constructs. It would be a wise initiative to put all these contributions together and to study what may be common and what should stay specific. In this paper we underline some similarities between process meta-models and we sketch out how they could be organized.

No abstract