A number of journal articles C. J. Date has written over the last five years are collected here with a few additional comments. The book begins with a summary of the major advantages of the relational approach--simplicity, power, and ease of use. In “Why Relational?” Date spells out the many benefits of the relational data model.

In “What Is a Domain?” Date describes a more refined concept of domains in a relational database. As he defines the concept, it begins to sound more and more like an object class in the fully object-oriented sense. After some mild hedging about lack of agreement on object-oriented terms and definitions, Date admits that domains and objects are probably the same. This article is very important because, if this recommendation is followed, it will extend the relational model to subsume object-oriented databases.

In the next chapter, “Defining Data Types in a Database Language,” Date turns to the specifics of describing and implementing a user-defined data type (or domain or object class). He goes from the semantics of the new data type through to the definition of operations. Although this article is really about defining a date (chronological) data type, it is a model for any user-defined data type. In a related article later in the book, Date criticizes IBM’s implementation of date and time data types.

In “Why Duplicate Rows are Prohibited,” Date provides theoretical, semantic, and pragmatic reasons why the relational model does not allow duplicate rows. This issue is important because the Relational Model version 2 does not allow duplicate rows. This article provides some, although not all, of the justification for this position.

Date examines the issue of nulls and three-valued logic in “NOT Is Not ‘Not’!” and “EXISTS Is Not ‘Exists’!” In the first chapter, he examines a proposal for three-valued logic (true, false, and unknown), which provides a more systematic treatment of missing information. He brings up several cases in which the relational model would require greater intelligence in order to be logically consistent. For example, “where x = x” should always be true. If x is unknown, however, this statement is not true. Date describes further logical flaws in “EXISTS Is Not ‘Exists’!” These two articles are important in light of the four-valued logic that Codd introduced in the Relational Model version 2, which still does not solve all the problems.

In “What is a Distributed Database System?” Date provides 12 criteria for a distributed database system. These criteria would be extremely useful for anyone who is evaluating a distributed database or comparing products, but one has to wonder what is so magic about the number 12.

“SQL Dos and Don’ts” might more appropriately be called “Elements of Database Style.” It is an informal survey of good techniques for designing and exploiting databases. A database designer or developer might get a better understanding of good database design from this article than from all of the relational literature to date.

One-to-one relationships are not as simple as they may seem. Often they are one-to-zero or zero-to-one, yielding nontrivial design problems. “A Note on One-to-One Relationships” discusses some of the subtleties of this problem and its implications for database integrity.

Other chapters include “Referential Integrity and Foreign Keys” (Parts 1 and 2), “A Contribution to the Study of Database Integrity,” “Further Relational Myths,” “Support for the Conceptual Schema: The Relational and Network Approaches,” and “What’s Wrong with SQL?”

The final section of the book is a collection of chapters by Andrew Warden called “Adventures in Relationland.” ( Andrew  Warden is a pseudonym for Hugh Darwen, although the rationale for the nom de plume escapes me.) The chapters highlight some technical subtleties of the relational model and some weaknesses in SQL. The points are presented in an allegorical style using monsters, princes, beautiful maidens, and wise old men. I like the idea of allegory for conveying complex concepts and think that this industry needs more storytelling. I also think that Warden makes some strong technical points. Unfortunately, I do not think that the allegorical style works well with technical points. Nonetheless, these chapters are well worth reading and make a significant contribution to the book.

This book is important reading for those who wish to keep abreast of the current thinking about the relational data model. While E. F. Codd is the founder of the model, C. J. Date is certainly its banner carrier. Interestingly, Codd and Date do not always agree. This book provides an interesting counterpoint to a few of the issues brought out in Codd’s recent book on version 2 of the relational model [1], and should be read to get some alternative views on key issues.