steX bv

I consulted and mentored for an SME that develops client-specific laser systems. The challenge of this project was to improve the –limited– available skills and resources (one employed software developer), to obtain a more reliable software development practice and a practical evolution path to organizing their collection of machine control applications into a software product family.

• Scalable design perspective: I developed an organisational framework for the software, as well as a roadmap, which support step-wise evolution from a set of independent products with substantial overlap, to a product-line where shared functionalities are reused, and maintained once.
• Technology-people-value perspective: This was an interesting example of balancing technology, people and value; for an organization with very limited software development resources, the right techniques and methods had to be gracefully introduced, aligned with the objectives and expectations of the organization.

The IDEALS project had the aim to develop a software design methodology that realizes the structured composition of software from separate modules, while handling system-wide interacting aspects of a problem domain. ASML was the ‘industrial carrying partner’. I was supervising the contributions from the University of Twente. One of the big successes of this  project was that already during the project, we achieved a technology transfer, by performing a feasibility study and proof of concept for an aspect-oriented weaver (for the C language); the proof-of-concept was eventually redesigned in-house at ASML, and (early 2008) is in use by ca. 50 developers, involving ca. 2MLOC, and in use for all their lithography machines. This also led to a spin-off project between ASML and the University of Twente (‘StarLight’), to turn a research prototype into a production-quality aspect weaver for the .NET platform. On another topic, we addressed the gap between design specifications and implementation. We came up with an innovative solution (patented by the University of Twente), that was demonstrated (through controlled experiments with 24 ASML software developers) to offer substantial improvements in effort and error reduction.

• Scalable design perspective: the above-mentioned 'weaver', addresses the issue that certain types of functionality, such as tracing and parameter checking, do not scale up well for large systems, as they become repeated and replicated over the system. An aspect weaver enables modular implementation of such functionality, which results in a much cleaner design, that is virtually unaffected by the size of the system.
• Technology-people-value perspective: the implementation gap between design and implementation is only a problem because of the limitations of the human mind; in addition, advanced formal specification methods existed that could solve the problem, but these were too complicated to use to gain acceptance among developers. Finally, there was a challenge to avoid over-specification of requirements, which would break down as soon as the system evolves. The proposed solution was positioned to balance these forces.

I consulted at Philips Medical on a large project with the goal of building a family of X-ray systems. Important issues were managing complexity, reuse, soft real-time constraints and reliability. My responsibilities in this large project included: the specification of an OO design process that is tailored to their organization, product, development environment, and so forth. Further, I mentored a pilot project that applied the design method, and was a member of the software architecture team, offering OO modeling and architectural experience.

Panfox, now a subsidiariy of Ordina, is well-known in the Dutch IT market for architecture-based design of large administrative systems. I have been involved in their efforts to develop an architecture for a component-based environment for developing and maintaining large data-intensive applications. The approach was based on their own method for analyzing and modeling information systems. My role was to offer knowledge about object-oriented modeling, composability of software components, and to co-design the software architecture. I have also given an in-house course on various aspects of object-oriented software architecture design.