My assignment was to evaluate a certain number of production type workflow engines for my client and, if the bill was too high, estimate what it would take to develop a custom engine. Since I knew - and still do - very little about workflow engines, obviously I had to turn to the existing bibliography. As a reference though, I think that 90% of what I learnt, I actually picked from studying the excellent documents produced by the Workflow Manufacturers Coalition (www.wfmc.com).
Anyway, my initial impression, when I went through my first diagonal reading of this book was "what a total waste of time". However, going over it more thoroughly, I did find a few chapters that brought interesting insights.
To sum it up, I find the title misleading "Business Process Implementation: Building Workflow Systems" transmits the idea that after reading the book, you will be able to BUILD a workflow (engine). Well, forget it. The content can only qualify as a basic description of requirements, at the initial step of the life cycle: analysis. What I feel was a wrong initial decision by the authors, is that a book offering to help you build workflow systems should assume that its readers will not be knowledgeable in data structure as a minimum. But the book dedicates fully 2/3 to 3/4 of its content teaching entities, relationships and a few basic concepts about object orientation, as if the reader started out without any knowledge about Information Technology.
All in all, chapters 4, 6, part of 7, and part of 8 actually deal specifically with worflow concepts. About 65 pages out of 227 are relevant to the field (Unless you accept the authors'premise that you also need to be taught how to design your ERD).
As for the relevant part, I liked the scope of the definitions given by the authors for what should be expected of a workflow engine. The relationships between the lifecycle, stages and tasks states are clear enough so you can start sweating over writing your own algorithms. There is a much too short mention about rules, enough to let you detect a black hole there, but not enough to know what to do about it. The treatment of backtracking is useful as well as that of relevant stages for a task (have not found it mentionned in any wkf engine) and so are the general ideas of when to start and end a task to minimize the cost of backtracking.
On the down side, don't look for block diagrams, structured map, component diagrams, UML definitions, state transition diagrams, sample source code, etc. that you would need to actually build a workflow system.
But then, the question becomes more general: "Can you expect to find in a $45 book enough information and specifications to start building what it took others $100,000 or a $1,000,000?". After all, the book is based on the acquired knowledge of developping a real commercial engine, so maybe I was a little ingenuous in thinking I would find it all ready to take home. I really don't know the answer to that one. Still, as an example among many, I have a book (Simulation Modeling and Analysis, by Law and Kelton) that give you the source code for a basic simulation engine in Fortran, Pascal and C. You really know what simulation is about after reading it. Maybe MM.Jackson and Twaddle could have been slightly more detailed with their experience!
So, if you are a manager and want to understand what the salesmen from FileNet or Oak Grove, are talking about or you are suddenly developping an interest in the subject but never heard of System Architecture I, then this book is for you. To be fair, I am still looking for the right book for what I need!
