Towards a Compiler for Business-IT Systems -- A Vision Statement complemented with a Research Agenda

Business information systems or enterprise applications have continuously evolved into Business-IT systems over the last decades, directly linking and integrating Business Process Management (BPM) with recent technology evolutions such as Service-Oriented Architecture (SOA) concepts and Web services. Interestingly, many of these technological evolutions include revivals of areas that have been in the focus of academic research in the past. For example, business rules closely relate to expert systems, Semantic Web technology uses results from description logics, attempts have been made to compose Web services using intelligent planning techniques, and the analysis of business processes and Web service choreographies relies on techniques originating from model checking and protocol verification.

As such, many of the problems that arise with these new technologies have been solved at least in principle. Research in the BPM/SOA space, sometimes also called Service-Oriented Computing (SOC), has tackled many of the problems and proposed a variety of solutions. However, if we try to apply these "in principle" solutions, we are confronted with the failure of these solutions in practice. Either a solution technique does not scale to the real-world requirements or it relies on assumptions that are not satisfied by Business-IT systems. This failure of existing solutions is limiting a successful application of these new technologies, in particular for smaller business players.

As has been observed previously, research in this area is fragmented and does not follow a truly interdisciplinary approach. To overcome the fragmentation of the area we propose the vision of a compiler for Business-IT systems that takes business process specifications described at various degrees of detail as input and compiles them into executable IT systems. As any classical compiler, the parsing, analysis, optimization, code generation and linking phases are supported and we describe a set of 10 research problems that we think must be solved in order to bring our compiler vision to reality. We argue that our vision provides a unique technological basis and foundation that enables a multi-disciplinary approach where existing techniques are combined with novel solutions filling in technology gaps. We position our vision within the life cycle of BPM and SOA applications to address the two interleaving trends of commoditization and innovation of business processes.