Software Product-Line Engineering: A Family-Based Software Development Process (Paperback)

~ (Author), Chi Tau Robert Lai (Author)

Editorial Reviews

Product Description

"It is in the "how to do it" part that this book excels, because it illustrates a process that has been successfully applied to reduce costs for organizations that develop large programming systems. With the help of this book, many more can learn how to exploit the idea of program families and bring about a substantial improvement in the state of practice in the software industry." --David Lorge Parnas

Many organizations have mastered the practice of software development, yet few have become truly efficient at software production. With the adoption of an efficient, systematic software production method, organizations can gain significant competitive advantages, including reduced time to market, better schedule predictability, more reliable code, and decreased costs. Software Product-Line Engineering provides the actionable information and proven tactics necessary to effect organizational change and make your future software projects more successful.

The authors outline a systematic method for rapid software production through the FAST (Family-Oriented Abstraction, Specification, and Translation) process, a revolutionary commercial product developed at AT&T that continues to evolve at Lucent Technologies. FAST uses practical domain engineering to decrease the time and effort necessary to develop, deliver, and maintain software. Any software development projects currently using C, C++, or Java can easily incorporate the FAST model and quickly reap the benefits of a more efficient software methodology.

From the Inside Flap

Industries and marketplaces often suffer radical changes in seemingly brief periods of time. The software development industry, in all its forms, has the opportunity to undergo such a change now. The hallmark will be a conversion from software development processes that are characterized by develop-ing an individual system and then creating variations of it, to software development processes that create product lines and families of systems. Creating variations on individual systems takes continual investment in understanding new requirements, and in redesign, recoding, and retesting. Creating product lines and families, on the other hand, invests in support for understanding new requirements and rapidly cre-ating new family members with little or no redesign and recoding and with reduced retesting. Changing from the first strategy to the second means changing the software development techniques that you use and changing your organization. Fortunately, you can make both changes incrementally if you know what you are trying to achieve.

We wrote this book to show you what we think your target in improving your software development process should be: a software development process focused on building families. The process we describe is based on our experience with creating soft-ware families, experience that extends back to the middle 1970s. Most recently we have seen improvements from applying family-based processes at Lucent Technologies, showing decreases in development time and costs for family members of 60%n70%. Our comparison is based on measuring the time and effort to create variations on a product before a family-based process is introduced and again after it is used.

Our intent is to identify the key ideas whose combination can radically alter the way software developers do their jobs, with attendant major gains in their productiv-ity and in the quality of their products. By focusing software developers on building software families, these ideas can be woven into a process for software development that is much more effective than the processes in common use today. The result is a paradigm shift in software development that involves creating two new types of organizations: one devoted to defining families of programs and creating facilities needed for rapidly producing family members and a second one devoted to rapidly producing family members by using those facilities. Creation of members of a family is akin to a production process and is enabled by an investment in tools and pro-cesses for the entire family. The result is to make it possible to create high-quality software applications much faster than with current processes.

Most of the ideas we useosuch as abstraction, separation of concerns, information hiding, formal specification, and model buildingohave permeated the research lit-erature in software engineering for many years but have not been widely applied in engineering practice. A process that incorporates these ideas into a practical, family-oriented software production process was introduced at AT&T in 1992 by David Weiss; its roots can easily be traced back through 30 years of research in software engineering conducted by a variety of people. Called the Family-Oriented Abstraction, Specification, and Translation (FAST) process, it is now in use at Lucent Technologies, where its evolution is continuing. Because of its focus on producing family members, we often refer to FAST as a software production process rather than a software development process.

This book introduces the ideas that software development organizations need to know to evolve into software production organizations. Such an evolution is easiest if an organization can create for itself a software production process. Accordingly, the FAST process is really a pattern for doing so, and we think of any process that conforms to that pattern as a FAST process. Put another way, FAST processes form a family. Within Lucent we have created several members of the FAST family of processes, using variations on the basic pattern as people and circumstances demand. With the help of this book we hope that you can begin to create your own FAST processes.

Planning and structuring for change is a central theme of FAST processes. It has also been a key theme in software engineering for many years. Characteristic of this theme is a continuing search for better abstractions. Finding and applying appropri-ate abstractions in software design and in programming languages has been a major tool for software engineers. FAST processes further develop the theme of abstraction by asking software engineers to find, for each family, abstractions that are useful in defining the family and describing its members. For each family we incorporate such abstractions into a language for specifying and modeling family members. The description of a family member in the language can be analyzed for completeness, consistency, and other properties, and, with a sufficient investment, engineers can build tools that generate the software for the family member from its description. The ability to perform each of these steps represents a further step in an organiza-tionis evolution from a software development organization to a software production organization. The ability to do all of them represents a step in the evolution of the software engineering community to use abstractions to better advantage.

As an organization evolves from software development to software production, so will its FAST processes evolve. We have started this evolution within Lucent Technologies.To understand and track our progress, we need a way of precisely describing the pro-cesses we are using and have used. Our mechanism for doing so is the Process and Artifact State Transition Abstraction (PASTA) process description method.

PASTA allows us to describe the artifacts that we use in our process, the activities that are performed during the process, the operations that we use to manipulate the artifacts, and the roles played by people during the process. It allows us to describe activities that may proceed concurrently and activities that must be performed sequentially as well as situations when backtracking may occur. A PASTA model of a process is also a good basis for developing automated support for the process.

This book contains a PASTA reference model for FAST processes. Having such models gives us a record of the evolution of FAST. The nature of PASTA and its sup-porting tools help us to understand the possible effects of a change to FAST and make it easier for us to make changes to the model. PASTA thereby facilitates changes to the FAST process. We have left certain aspects of the model incomplete because they vary considerably from one organization to another. For example, the configu-ration management process is usually highly specialized for an organization.1 We have done no more than sketch how change reporting might be handled for a family. We hope that you will use the model in this book as a starting point for creating your own reference model for your own FAST process.

We live in a time when business enterprises of all sorts appear to be undergoing continuous change. Such change usually relies on altering the processes that the enterprise uses. In manufacturing industries, the idea of redesigning product lines and processes so that a product is easy to produce using its production process is known as concurrent engineering. Both product and process are designed together. FAST and PASTA together can be viewed as concurrent engineering for software: FAST processes help software engineers to design both a family whose members are based on predicted changes and a process for producing those members based on the predictions. PASTA helps a software development organization to deploy and enact FAST as a production process, providing a way to create guidebooks and toolsets to support the process.

A variety of industries have adopted the notion of continuous change in the interest of gaining a competitive advantage--namely, to be able to produce customized products rapidly. The software development industry is no different. Competitive demands are pushing software developers to create products in greater variety faster than ever before. FAST is designed to help software engineers respond to this trend and, by appropriate investment and planning, to take advantage of it. 1. See R. E. Grinter, "Recomposition: Putting It All Back Together Again," ACM Conference on Computer Supported Cooperative Work (CSCW '98) (Seattle: November 14-18, 1998), for an interesting view of how configuration management varies among software development organizations. Acknowledgments

This book was in the planning, creating, and re-creating process for an agonizingly long time. We thank our wives--Joanne Glazer Weiss and Ya Chien Chuang--for their patience and willingness to delay other projects in favor of the book. We also thank Deborah Lafferty for her patience and encouragement through many missed deadlines. Many people helped to shape the ideas that underlie FAST and PASTA and worked to polish them enough so that they were ready for use. Prominent among them are David Parnas, Grady Campbell, Stuart Faulk, Rich McCabe, James Kirby, Jr., James O. Coplien, Omer Aiken, and Steve Yau. We would never have had the con-fidence to write this book without the assistance and support of many people who were willing to experiment with FAST and PASTA and who helped to shape them into their current working form. Prominent among these people are David Cuka, Mark Ardis, Lloyd Nakatani, Bob Olsen, Lynn Paulter, Paul Pontrelli, Doug Stoneman, Richard Braatz, Lizette Velazquez, Jan Sharpless, Ivy Mackowiak, Steven Nolle, Carl Chang, Wei Tek Tsai, Tom Case, Ken D. Shere, Lina-Na Hwa, Martha Lin, Chu-San Sam Shen, C. W. Chen, Ya-Chien Chuang, and Herman Chuan-Ming Hsiung. Art Pyster, Eric Sumner, Jan Sharpless, and Mary Zajac provided management support and encouragement. Deserving of special mention are David Cuka, who did much of the development work on the SPEC example and who has been an early and tireless advocate of FAST within Lucent, and Dan Hoffman, who did much of the develop-ment work on the Floating Weather Station example. James O. Coplien occupies a special niche in our thoughts, first for his encouragement to us to write this book and second for his explorations of the ideas of commonality and variability and how to use them to create a multiparadigm approach to software development. These explorations are captured in his own book, which we commend to your attention. Special thanks to Tina Murthy for helping to create the FAST PASTA model and to Ching-Peng Chu, Chung Hong Chien, Ting-Shun Cuang, Jimmy Truong, Tong-Gao Tang, and Wei-Li Johnson Liu for helping to implement the PASTA toolset, which we used to create the FAST PASTA model. Joanne Glazer Weiss helped with the tedious task of verifying table formats. David dedicates this book to Jack and Bertha Weiss. Robert dedicates this book to Tsung-Yen and Chiu-Lan Lai. David Weiss
Lucent Technologies
Naperville, IL

Chi Tau Robert Lai
International Software Process Constellation
Reston, VA

See all Editorial Reviews

Product Details

• Paperback: 448 pages
• Publisher: Addison-Wesley Professional (August 22, 1999)
• Language: English
• ISBN-10: 0201694387
• ISBN-13: 978-0201694383
• Product Dimensions: 9.5 x 7.7 x 1.1 inches
• Shipping Weight: 2 pounds (View shipping rates and policies)
• Average Customer Review: 3.0 out of 5 stars  See all reviews (2 customer reviews)
• Amazon.com Sales Rank: #1,265,792 in Books (See Bestsellers in Books)

Customer Reviews

Most Helpful Customer Reviews

11 of 12 people found the following review helpful:

3.0 out of 5 stars A fair introduction, a good reference..., April 15, 2000

By	Ed (Northern Virginia) - See all my reviews

This book really deserves two separate ratings. The first five chapters introduce domain engineering in general. If you are new to the idea of family-based software development, or need some convincing of its intended benefits, this section isn't bad. The two case studies presented gloss over some of the complicated concepts while providing too much technical detail in others. For example, the important but non-trivial job of creating the "decision model" for the domain is given minimal coverage. On the other hand, every class in the JAVA code used to illustrate one example is explained individually. I'd give this section a 3-star rating as an introduction to domain engineering.

The rest of the book is an incredibly detailed reference into the authors' implementation of the FAST process. They spend two chapters bringing the reader up-to-speed on the process modeling technique called PASTA. Then they provide almost 200 pages of reference information (also included on CD-ROM) on the specifics of FAST. If you are looking for a recipe to use to try domain analysis and implementation in your company, this section covers it. If nothing else, you'll see a very well defined process modeling notation in action, one that can be used in many applications. Without any personal experience implementing the process, I hesitate to rate this section. But if the process works as well as this section is detailed, I'd give it at least 4 stars.

Its hard to say from their examples how well the FAST process will scale to industrial-size domains. But the authors have presented an easily read book with enough detail that a motivated software engineer may be able to win management support and give domain engineering a try.

3.0 out of 5 stars Just another methodology book, May 11, 2006

By	wiredweird "wiredweird" (Earth, or somewhere nearby) - See all my reviews (TOP 100 REVIEWER)

Amazon Verified Purchase

I've just about never worked on commercial software that was shipped in only one form. There have been variations in the platform configurations, in the functional options, or in the generation of product. This book addresses the process of building families of software products, the kind that derive generally similar sets of functionality from a shared code base, i.e., the majority of commercial software in my experience.

The authors start with an obvious premise, but a premise that's overlooked far too often. They start by saying that their approach does not work in all cases, and may not work in yours - if only such honesty were more common! Although it makes the third, or fourth, or tenth member of the product family easier to develop, it incurs design and implementation costs up front. Despite long term economies, the first few family members might actually cost more than with one-off design techniques. The first step in this methodology is to decide whether the next steps will actually do you any good.

After that, though, I find this to be only an ordinary presentation of software design methodology. Although it adds steps for analyzing product families to the standard software development cycle, those steps are only a small part of the whole. At least 90% of the cycle is the same for a one-off project as for one where massive reuse is a major design goal, but the authors present the entire cycle in detail. I'd even say excruciating detail. The second half of the book, 200 pages or so, presents the methodology one step at a time, often one step per page, with forms to be filled out at each step: formal pre- and post-conditions, lists of involved participant, artifacts generated or modified at each step, and on and on and on.

Despite it's overwhelming mass of detail, I find it systematically lacking in one major respect. The first pass over the process is only a small part of any long-lived project. After that, every successive design cycle takes the entire result of the previous cycle as input. In some sense, acknowledgement of the past is the first and biggest requirement, where the past consists of all code and documentation artifacts from the previous effort. "Green field" construction is a rarity in the software world, but a premise of most design methodologies, this one included. The second major weakness in this book may just be a matter of editorial choice: it addresses the organizational aspects of the project only, with few (if any) tips for the technologist creating a line of products. The techniques of design for flexibility, especially for variation over time, cover whole ranges of topics that aren't part of this dicsussion.

So we're left with a mass of imposing organizational detail, with reams of families of forms and formalities. These could become oppressive if not handled in a context-sensitive way, but the authors say little about tailoring the process to different environments. And, anong what's presented, only a small part is specific to the product-line problem as opposed to any other software development. I'm still looking for stronger technical advice and a more flexible approach.

//wiredweird