A painstaking and utterly theoretical exercise in formal modeling, this work is almost unreadable in the usual sense. At its core, this dated dissertation (finished in 2004 and only available on demand) contains 92 definitions, 60 requirements, and only six derived theorems in 100 pages, all specified within the (unexplained) Rigorous Approach to Industrial Software Engineering (RAISE) specification language (RSL); think pseudocode supplemented with set-theoretic and logic extensions like functions, relations, and logical connectives.

The main topic is framed by an introduction to aspects of domain engineering (which are dispensable and not required later on) and two case studies demonstrating the application of workflow model theory to well-known process patterns and a rather simplistic model of a clinic (comprising only four types of processes). The book concludes with 60 pages of RSL specifications, translated into the (unexplained) prototype verification system (an automated theorem prover), and an almost-20-page bibliography. The missing index renders the book practically inaccessible--a reader looking for an earlier definition (and you will need to do that often) will have no option but to skim through all previous pages.

I work for a software company that invented a business process modeling language (ARIS) back in the 1990s that is still in use today. I really did try hard to find some pearls amid this tiring repetition of RSL snippets containing definitions, requirements, and a few theorems, annotated (too) briefly by plain-text motivations.

Practitioners within the workflow management systems and software engineering domains will undeniably want to avoid this work, as it merely reflects the thoroughly outdated state of the art from around the year 2000.

Furthermore (and despite its claim), the book definitely is not about software architecture in any common understanding of this term. I am an information technology architect, and initially failed to spot any of the usual architectural artifacts, references, language, or diagrams. Only by carefully reading the definitions on a word-by-word basis did I discover that whenever the formal RSL model requires a function for computing a task such as checking whether a resource is available, that function is supposed to be provided by a suitable software component. However, the author spares us any component diagram.

The theoretical computer scientist who still persists through this review might rightfully wonder about the six theorems. Sadly, they fall short of revealing insights and merely demonstrate how certain subsets of definitions can (sometimes trivially) be combined to yield others. For instance, one theorem states that “if a process is sound, that it does not reach deadlock.” This is simply a rewording of the notion that not reaching deadlock is an element of the very definition of soundness. This also pertains to the concept of “workflow correctness” featured in the blurb, which is (superficially speaking) tantamount to a workflow that meets all its time, task, and resource dependencies. Proofs of all theorems are only given in the form of (unintelligible and totally unexplained) prototype verification system traces.

The blurb also claims that the work is based on “domain engineering” and on “algebraic” methods. The former claim is true, albeit in the (restricted) sense that only the domain of workflow modeling and execution systems is modeled. Proper techniques for eliciting requirements are mentioned in the introductory sections on domain engineering, but do not appear at all in the later modeling sections. The 60 so-called “requirements” of the formal model merely represent definitions motivated by fairly simple domain considerations. The “algebraic methods” collapse into a short, unexplained list of ten introductory definitions of model theory (signatures, model morphisms, and such concepts), which are never used later on.

Finally, while I am confident that much effort and work has been put into this theoretical compilation, given the age of the work, I would advise readers to look for other, more recent contributions by the author to judge whether to pursue this particular line of his reasoning in the current format.