The purpose of this collection is to assemble significant research contributions so they are easily accessible to anyone interested in database research. It is appropriate for use as an introduction for students or professionals from industry, and as a reference volume to anyone active in database systems. . . . It is intended to serve as a core of material that any DBMS professional should be familiar with. Moreover, any industrial practitioner or graduate student who wishes to be current on the important research themes would be well advised to read these papers.

--From the Preface

I can find no reason to disagree with the above remarks. This collection of 40 papers is organized into 9 chapters, each of which provides a solid introduction to a major area of database research.

Chapter 1, “The Roots,” contains five papers on the beginnings of relational database systems. The first of these is, appropriately, E. F. Codd’s seminal 1970 paper “A Relational Model of Data for Large Shared Data Banks.” The remaining papers describe the two most important early relational DBMS implementation efforts, IBM’s System R and the University of California at Berkeley’s INGRES. Two papers discuss each system, one written around the time the system was being developed, and the other being a retrospective look at the system some time later.

Chapter 2, “Relational Implementation Techniques,” contains eight papers. These include key papers on query optimization and transaction processing (concurrency control) mechanisms, as well as papers on the DBMSs’ operating system interfaces and buffer management.

Chapter 3, “Distributed Database Systems,” leads off with an overview of a complete distributed DBMS--IBM’s experimental R* system--and continues with four papers on techniques for implementing distributed relational DBMSs. These papers deal with extensions of the query optimization and transaction processing issues discussed in chapter 2 to include distributed and multiple-copy data. They also deal with recovery from various types of failures that can be encountered in distributed systems.

Chapter 4, “Performance and Database Machines,” contains four papers about various aspects of providing high performance and reliability in database systems. Two of these describe the well-known Wisconsin and TP1 (debit/credit) performance benchmarks. The third paper surveys various approaches that have been used to provide highly available database systems, while the fourth describes a specialized database machine architecture, the Wisconsin GAMMA machine.

Chapter 5, “User Interfaces,” contains three papers. In the first of these, Joachim Schmidt’s classic paper on Pascal/R, the user is a programmer. This paper describes a DBMS/programming language interface and can be considered a basis for much of today’s research in “persistent” or “database” programming languages. The second paper also describes a database programming approach, but this time at a higher level by “filling in the form.” The third paper describes a database interface for nonprogrammers based on entity-relationship techniques.

Chapter 6, “New Data Models,” contains seven papers that describe work on data models intended to be advances beyond the relational model. These include Peter Chen’s original paper on the entity-relationship model and representative papers that describe work on semantic and object-oriented data models and Codd’s own extended relational model, RM/T.

Chapter 7, “Extendibility,” contains two papers that cover approaches to extending a DBMS to handle data types and structures required in “non-business” applications, which conventional relational DBMSs cannot easily handle. The first of these describes Stonebraker’s own POSTGRES system, which attempts to provide extensibility within a relational framework. The second is on the Wisconsin EXODUS system, which takes a more radical approach and allows extensions to the basic underlying data model.

Chapter 8, “Integration of Knowledge and Data Management,” contains two papers that illustrate representative work on integrating “knowledge processing” (such as that performed in expert systems) with database management. One of them covers strategies for recursive query processing (queries over ancestor relationships or part hierarchies are frequently of this type), while the other covers the mechanisms for handling rules provided in the POSTGRES system.

Chapter 9, “Storage Management Issues,” contains four papers dealing with database implementation techniques, at the level of access methods and storage management, that go beyond the conventional B-tree approach. The contents include Witold Litwin’s paper on linear hashing, two papers on storage structures for spatial data, and a paper that describes a complete subsystem at this level, the POSTGRES storage manager.

This solid book is well suited to its intended purpose. The papers certainly provide the core material that the preface describes, and most of them are very readable as well. The introductory commentary that Stonebraker has provided for each chapter is also important. Natural questions that newcomers to the DBMS research literature might have on reading particular papers from the past (even the recent past) are “Yes, but what do people think of these ideas now?” and “How much of this is actually reflected in real DBMSs?” Stonebraker’s commentary deals with such questions very well and provides the reader with a valuable perspective both for evaluating the papers in the book and for pursuing other relevant work. The commentary also indicates some of the issues that were or are still controversial in the field (as well as some issues on which other researchers might be expected to disagree with Stonebraker’s commentary). This discussion can help guide the reader who is encouraged by this book to dive into more recent literature.

Anyone interested in database technology who does not have ready access to the papers in this collection would be well advised to buy it. Stonebraker has indicated that he intends to update the book every couple of years. Such updates would be worth looking forward to (perhaps the publisher should consider a loose-leaf edition).