Browse Books

Reviewer: Krishnarao Gururao Kulkarni

The two editions of Ullman's earlier book on Principles of database systems [1,2] have both been popular texts since their publication. The book under review is the first of two volumes intended to replace them. Most of the material in this volume is concerned with classical database systems, although one chapter is devoted to describing some aspects of so-called knowledge-base systems (KBS). Another welcome addition to this volume is the clear discussion of the emerging area of object-oriented database management systems (OO-DBMS). As in his earlier book, the author emphasizes theory, particularly the relational theory and first-order predicate logic. The book is divided into ten chapters. The first chapter provides the usual introductory material on database systems and includes an excellent introduction to both OO-DBMSs and KBSs. The author describes KBSs as “systems implementing logic, in one or another form, as a programming language” and OO-DBMSs as “systems supporting object identity, abstract data types or classes and type hierarchy.” Chapter 2 discusses the three classical data models (relational, network, and hierarchical) and a typical object-oriented model. Chapter 3, which provides a good introduction to logic programming, discusses data models that use first-order logic as a way to represent knowledge. This chapter begins by considering the meaning of logical rules. The author mentions three ways of interpreting logic programs: proof-theoretic, model-theoretic, and computational. His comparison of these three approaches would have been better if he had used examples. This chapter also covers the evaluation of rules (both nonrecursive and recursive), negation handling in rule bodies, and the impact of the closed-world assumption. Much of this discussion employs “datalog,” a restricted version of Prolog developed by the author. Chapter 4 covers the query languages associated with relational systems. SQL, which has recently been accepted as an ANSI standard, is discussed in greater detail than in the previous editions. Chapter 5 discusses data definition and manipulation languages associated with network (DBTG) and hierarchical (IMS) systems, as well as the OO-language associated with Gemstone, a commercial OO-DBMS. I find it strange that the author describes the language associated with the DBTG and IMS systems as object-oriented. Though he explains his reasons, the characterization is potentially controversial. The remaining five chapters deal with physical data organization, the design theory of relational databases, integrity and security aspects, transaction management, and distributed database management. The second volume is expected to cover query optimization techniques (for both classical database systems and KBSs) and “universal relation” systems. While the material on KBSs and OO-DBMSs is new, the rest of the material in this volume has been taken almost verbatim from the earlier book. Each chapter includes a good set of exercises and bibliographic notes. The notes have been extensively revised to include more recent publications. The book's major strength is the way it handles data models, languages, and relational design theory. Though the tone is rather formal, the material is clear and easy to understand. The same cannot be said about the material on physical data organization and transaction management. Some recent publications [3,4] cover these aspects much more clearly than this work does. The book also lacks a clear discussion of the shortcomings of the value-based (relational and logic-based) data models. Though it is true that value orientation leads to declarative languages, which have the advantage of extensive query optimization, value-based data models are forced to depend on user-defined identifier keys in order to establish the identity of an object. The problems with this approach have been well documented by Khoshafian and Copeland [5]. Indeed, much of the interest in object-oriented data models has come about as a response to this important shortcoming in relational systems. In conclusion, this book is well written and up to date, and it will be especially valuable in new areas like KBSs and OO-DBMSs. It would make an excellent text for a senior or graduate-level course.