Customer Reviews

Databases, Types and the Relational Model (3rd Edition)

5 star:		(1)
4 star:		(1)
3 star:		(0)
2 star:		(0)
1 star:		(1)

 

 

In this 3rd edition of The Third Manifesto, the authors have, in their own words, stripped the 'polemics'and added exercises to facilitate tuition.

This is a further refinement and expansion of their blueprint for a truly relational Database Management System (DBMS). It is fair to say that they are not influenced in the slightest by vendor concerns, and that is a refreshing but sadly diminishing trait in these computing times. Consequently, they will always call a spade a spade, and will not be shy is so stating. Therefore NULLs and duplicate tuples, for example, will always be abominations to be repulsed totally in any database that aspires to the 'relational' label.

Throughout all of their work concerning the Relational Model (RM), the first and foremost inspiration for which is of course E.F. Codd, there prevail several fundamental and immutable themes that are given full treatment in this volume: strict adherence to the mathematics of the RM (no NULLs, duplicates, etc.), clear separation of the logical and the physical (so often muddled together), the orthogonality of data types within the RM (data types do not impinge on or compromise the RM, and vice versa), proper and comprehensive declarative integrity constraints within the DBMS (no current SQL DBMS provides anything like full integrity constraint support), the many deficiencies of SQL (bearing in mind that Darwen and Date produced what is perhaps the definitive SQL reference in 'The SQL Standard') and therefore necessitating a better language (which they have devised, and called it 'D'). No physical aspects of implementing the authors' recommendations are considered in this book, and nor should they be.

Anyone who has clocked up any degree of significant experience on SQL DBMSs will surely have had 'fun' with duplicate, triplicate, quadruplicate, etc., rows, or having totally incorrect results returned because NULLs have not been handled correctly; not to mention severe data integrity issues, or the quirks of SQL.

I have no doubt that had DBMSs from their inception adhered to principles similar to those as laid forth in this volume, DBMS professionals would have had much less 'fun', and end users would have had much more confidence in the data returned to them. It may be very late in the day where the major DBMS vendors are concerned, but hope springs eternal (almost).

This boos is a very good and very important one. Unlike books on Object-Orientations, this book is properly detailed (most if not all Object-Orientation books are either not detailed or poorly detailed). However, there are things that prevent me to rate this book as 5-star. First, the use of the word Variable as the four core concepts. Not all objects are variables. There are constants and non-valuable objects (such as objects of type module / package / unit). Thus, for my own purpose and teaching, I replace variables with objects, and properly define 'Object' in my Doctoral Thesis. Second, the use of both terms Value and Literal. The term literal is not necessary. It breaks the accuracy of claim that the four core concepts is sufficient. The authors must either include literal as one of 5 (instead of 4) core concepts, or not use literal altogether. Third, the use of the word operator. While the authors admit that both terms operator and operation can be used interchangeably, they prefer operator. I prefer operation. Using the word operator is too mathematical for programmers. Fourth, the use of the word tuple. Again this is bad. I prefer the word record. Fifth: the concept of possrep (possible representation) is unclear; it is a new idea but unclear and worthless. Why not using the module's interface and module's implementation? Sixth: Finally, Tutorial D is not a good programming language because it does not properly implement modular programming. Seventh: the books does not associate encapsulation with modular programming. Terrible mistake.

This book and its related site, "The Third Manifesto", is nothing but utter nonsense. The author and his associates are attempting to build a relational database model that doesn't have to deal with NULL values.

The end result is a complicated mess in which even a simple one-line query baloons into a dozen lines or more. Needless to say, these theories have never been put into practice in any real application.

I highly recommend that anyone interested in actual database theory to stay as far away as possible from this book.