Goals

The goals for the first edition remain goals for this edition as well. This book is still meant to be an easy-to-read introduction to the UML and how to apply its notation and semantics to the development of real-time and embedded systems. At the time of this writing, it is one of two books on the UML and real-time systems. I am also the author of the other, Doing Hard Time: Developing Real-Time Systems with UML, Objects, Frameworks, and Patterns (Addison-Wesley, 1999). Doing Hard Time is a more in-depth look at the fundamentals and vagaries of real-time systems, with emphasis on analysis of object schedulability, the use of behavioral patterns in the construction of statechart models, and how to use real-time frameworks effectively. Doing Hard Time is a deeper exploration of real-time systems that happens to use the UML to express these concepts. In contrast, Real-Time UML is primarily about the UML and secondarily about capturing the requirements, structure, and behavior of real-time systems using the UML.

In addition to these original goals for the first edition, the second edition adds two more: (1) bring the book in conformance with the recent changes in the UML standard, and (2) enhance the book's effectiveness based on feedback from the first edition.

The UML has undergone a couple of revisions. The first revision, 1.2, as almost exclusively editorial, with no significant modification. The UML revision 1.3, on the other hand, is a significant improvement in a variety of ways. For example, the «uses» stereotype of generalization of use cases has now been replaced with the «includes» stereotype of dependency, which makes a great deal more sense.

Similarly, the notion of an action in UML 1.1, relied heavily on the use of "uninterpreted text" to capture its details. The UML 1.3 has elaborated the metamodel to encompass a number of different kinds of actions, making behavioral modeling more complete. The action semantics metamodel, and how it relates to object messaging, is discussed in Chapter 2 and 4.

There have been a number of changes to the statechart model in the 1.3 revision as well. The first edition of Real-Time UML devoted a lot of space to statecharts, and this second edition expends even more effort in the coverage of behavioral modeling with statecharts. Much of this space is used for the new features of statecharts synch pseudostates, stub states, and so on. This resulted in a significant rewrite of Chapter 4, which deals with object behavioral modeling.

Recent consulting experience in fields ranging from advanced medical imaging to the next generation of intelligent autonomous spacecraft, in addition to reader feedback from the first edition, is reflected in this second edition. For example, numerous consulting efforts have convinced me that many developers have a great deal of difficulty with understanding and applying use cases to capture requirements for real-time and embedded systems. To address this need, I developed a one-day course called Effective Use Cases, which I have given at NASA and elsewhere. Those principles that have proven their effectiveness in the field are captured here, in Chapter 2. Similarly, the techniques and strategies that have worked well for capturing object models or state behavior, have wound up expressed in this book as well.

Another change in this book is the elaboration of an effective process for using the UML in product development. I call this process Rapid Object-oriented Process for Embedded Systems (ROPES). The most common questions I have been asked since publication of the first book have been about the successful deployment of the UML in project teams developing real-time and embedded systems. Thus, Chapter 1 has explains this process, and identifies the work activities and artifacts produced during different parts of the iterative lifecycle. In fact, the ROPES process forms the basis for the organization of the book itself from Chapters 2 through 7^[1].

^[1] More information on the ROPES process can be had from the I-Logix Web site, www.ilogix.com, as well in another book, Doing Hard Time: Developing Real-Time Systems with UML, Objects, Frameworks, and Patterns (Addison-Wesley, 1999).

Despite the goals of the UML in terms of providing a standard, there has been some fractionalization as vendors try to differentiate themselves in the marketplace. While progress will naturally involve vendors providing new and potentially valuable model constructs above and beyond those provided by the UML, several vendors have claimed that their new features will be part of some new yet-to-be-announced UML for Real-Time. Interestingly, some of these vendors don't even participate in the OMG while others provide mutually incompatible "enhancements." By spreading this FUD (Fear, Uncertainty, and Doubt) among the developer community, I feel these vendors have done a grand disservice to their constituency. Developers should understand both the benefits and risks of using single-source modeling concepts. These features may make the system easier to model (although in many cases these so-called "enhancements" fail in that regard), but it also locks the product development to a single vendor's tool. Another risk is the inability to use model interchange between tools when the models no longer adhere to the UML standard. This can greatly decrease the benefits to the developer that the use of the UML was developed to provide. In an effort to dispel some of the FUD, I've added Appendix B to outline what it means to make changes to the standard, why no single vendor can claim they own the UML standard (it is, after all, owned by the OMG), and what changes are likely to be made to the UML over the next several years.

Finally, I would suggest that interested readers visit the I-Logix Web site, www.ilogix.com. There you will find a number of papers on related topics, written by myself and others, as well as the UML specifications, tool descriptions, and links to relevant sites.

Bruce Powel Douglass, Ph.D.
Spring, 1999