Stefan Haustein

A Model-Driven Runtime Environment for Web Applications

A large part of software development these days deals with building so-called Web applications. Many of these applications are database-powered and exhibit a page layout and navigational structure that is close to the class structure of the entities being managed by the system. Also, there is often only limited application-specific business logic. This makes the usual three-tier architectural approach unappealing, because it results in a lot of unnecessary overhead. One possible solution to this problem is the use of model-driven architecture (MDA). A simple platform-independent domain model describing only the entity structure of interest could be transformed into a platform-specific model that incorporates a persistence mechanism and a user interface. Yet, this raises a number of additional problems caused by the one-way, multi-transformational nature of the MDA process. To cope with these problems, the authors propose the notion of a model-driven runtime (MDR) environment that is able to execute a platform-independent model for a specific purpose instead of transforming it. The paper explains the concepts of an MDR that interprets OCL-annotated class diagrams and state machines to realize Web applications. It shows the authors’ implementation of the approach, the Infolayer system, which is already used by a number of applications. Experiences from these applications are described, and the approach is compared to others.

Towards Information Agent Interoperability

Currently, many kinds of information agents for different purposes exist. However, agents from different systems are still unable to cooperate, even if they accurately follow a common standard like FIPA, KIF or KQML. Being able to plug agents together with little effort and exchange information easily, would be of a great use for several reasons. Among others, the agents could profit from each others’ services. In addition, certain aspects of multi-agent systems could be evaluated without needing to build a complete system. Testing agent systems with standard components would allow simpler comparison. Furthermore, building different agent-based applications would be simplified by combining new software with “off the shelf”-components. In this paper, we explore the feasibility of practical software development and integration of existing systems, without developing “yet another abstract agent architecture”.