This is a good book. It covers a wider range of topics and provides more details than any other workflow book I read. The authors target technical people who are interested in building, evaluating, or using workflow systems. The book doesn't provide all the implementation details you would need to build a workflow system, though. But once you read it you'll have a better idea about what it takes to do it.
The first part of the book discusses concepts relevant to workflow management. The authors cover a wide range of topics, from business processes to workflow basics to advanced workflow features like dynamic modification of workflows. However, some concepts don't fit in as nicely as others, which makes it seem unfinished. For example, in Chapter 4 the authors present in great detail a formal workflow metamodel. But 63 pages later they don't show how the reader can use this formal treatment: "One of the advantages of representing a metamodel in a formal manner is that one can prove statements formulated in the corresponding language. In our case, properties of process model graphs can their instances can be derived." The chapter ends with a 1/2 page (sketch of a) process termination proof.
The second part of the book focuses on techniques for implementing workflow systems. Here the authors discuss objects, transactions, advanced workflow functionality, and the architecture of workflow systems and workflow-based applications. These chapters seem more biased by IBM's MQSeries Workflow system (with which both authors have been involved) than claimed in the preface. Therefore, the presentation revolves around the techniques used in IBM's product and shadows other alternatives.
Many discussions in this book are followed by examples in the Flow Definition Language used by MQSeries Workflow. There are two problems with this approach. First, unless the reader uses the same workflow system, I question its value. Second, the FDL examples focus on specification rather than implementation. Probably the majority of readers are interested in the latter (which is missing) rather than the former. In fact, throughout the 2nd part the presentation of different techniques remains on a conceptual level.
Object technology seems an afterthought. While you can find the word "object" quite often, this book doesn't show how workflow management systems can benefit from object-oriented technology. The authors regard objects as components that merely encapsulate the implementations workflow activities. Their treatment ignores the other key features of object-orientation (inheritance and polymorphism). Incidentally, Chapter 6 (Workflow and Objects) is one of the shortest chapters in the book.
Some additional comments:
- There are a few places where the authors provide basic background information which probably should be left out. For example, Section 10.1.1 on Availability introduces MTBF and MTTR. In my opinion, unless this discussion is tied into workflow (which is not), it doesn't belong to this book. But this may not be the case for someone who hasn't encountered these concepts.
- The travel reservation example from Appendix A is too simple. As in the previous chapters, the authors spend a great deal with the FDL specification. I would have liked to see a more elaborate example that convinces the reader about the benefits of workflow technology. For example, does the 2nd part of the travel reservation process really require workflow, or merely a batch system?
To summarize, Production Workflow has an intrinsic bias that stems from the authors' experience with the FlowMark and MQSeries Workflow systems. Both these systems use persistent queues and relational database technology. The uninitiated reader may be led to believe that there are no other technologies for building workflow systems, when in fact this is not true. However, the experienced reader should be able to filter out the bias and project the ideas discussed in the book into different contexts.
As far as the quality of the book is concerned, it looks like a rushed job from the publisher's side. I have found quite a few typos, errors, and inconsistencies which should have been fixed by the copy editors.
