Browse Books

Reviewer: Andrew Robert Huber

With the current emphasis on family values, its about time someone brought family-oriented values to software development. Weiss and Lai describe and advocate a method specifically designed for developing software product line families. The goal of the family-oriented abstraction, specification, and translation (FAST) approach is to provide a systematic way of analyzing families of software products and to develop processes for generating these products. This results in an engineered environment for producing software family members. That is, FAST separates development into two steps: domain engineering, which creates the production environment, and application engineering, which invokes the production environment once for each software product. The payoff comes in reusing the engineered domain. Benefits include more efficient production of products, the ability to produce and support more product variants, improved quality, and a competitive advantage in the application domain due to lower costs. Chapter 2 discusses the three fundamental assumptions of the method: The redevelopment hypothesis—most software development is redevelopment, that is, creating variants of existing software with more commonality with than differences from the existing software. The oracle hypothesis—the changes needed to a system over its life are predictable. The organizational hypothesis—the software and the development group can be organized to take advantage of the predicted changes, so that producing a new version of the software consists of making independent changes to a few software modules. Knowing these assumptions goes a long way toward telling you whether the approach will work for you. The next chapter discusses the basic economic assumption of FAST: the investment in domain engineering is paid back by more efficient application engineering. A critical question is when the payback occurs. The authors present a model and state that, in their experience, the payback typically comes after the production of three family members. This is undoubtedly beyond the attention span and discipline of many development organizations. The first part concludes with an extensive example of a software family for a series of weather station buoys. This includes all the steps in FAST, from constructing a simple economic model to justify the rest of the process, through defining the common and varying parts of the system, to designing and implementing the tools that generate the final software and documentation from the family specifications. Details such as the commonality analysis, the code generating environment, and the generated code of a family member are included. The second part of the book describes the process and artifact state transition abstraction (PASTA), a state-based method for process modeling. The PASTA model is based on roles, artifacts, artifact states, operations, process states, and analyses. Though it is useful for general process modeling, the authors use PASTA to construct a complete model of FAST to aid in its communication and adoption. The second half of the book—a full 200 pages—details the FAST PASTA model. Chapters show the artifact definitions, the activity definitions, the role definitions, the analyses, and the relations using the PASTA forms and templates. These chapters fully describe all aspects of the FAST PASTA process. This section serves as a complete reference guide, rather than as text to read linearly. An accompanying CD-ROM also includes the contents of these chapters. It is much easier to view these forms online, and interested organizations can use the electronic versions as a basis for their own version of FAST PASTA. The CD versions would be more useful if they were more extensively cross-linked. Many cross references are not hypertext links, negating some of the value of online material. For example, three top-level figures show the FAST artifact tree, activity tree, and role tree, but there are no links from any of the nodes of these trees to the forms and templates that describe them in detail. Even more useful would be having the complete text on the CD, with links between the relevant sections of the book and the forms, making the CD a fully integrated part of the book. The CD also contains Java source and class files for the floating weather station example, although there are some differences between the books sources and the CDs sources. A 45-item bibliography, a glossary, and an index complete the book. This book will interest those involved in process modeling, development, or improvement. An open question is how widely applicable the FAST method is, given its assumptions, its payback time, and the impatience of many software developers in todays Internet environment.