From: BR

To: Editor

 

I wish to congratulate you both on this website. I greatly appreciate the articles and discussions that you have provided.

 

My introduction to relational database theory was through reading Chris' introductory book on relational theory. Over the years it has provided me with much help in the database design work that I have done. My only regret over the years has been the inability to put the theory into practice in all the work undertaken due to the ad hoc nature of the projects involved with and the constraints others have placed on the respective database designs.

 

Keep the fight going - it's the only way some of us who care will get to see true RDBMS systems come into existence. I am avidly going through your site extending my knowledge with the information you have supplied.

 

I am in the process of purchasing your books PRACTICAL ISSUES IN DATABASE MANAGEMENT and THE THIRD MANIFESTO, which won't be available to me till the new year. I am looking forward to reading both of them over the coming months.

 

I was at a presentation yesterday given by Borland regarding their "database product" Interbase 7. I asked a "silly" question relating to using their system with a fully normalized (to 5NF) database design - it was an attempt to see the reaction of the participants to mentioning relational database design. I have to admit that the reaction to my question brought forth mostly negative reactions - as if I was a fool to be even considering proper design. The reactions included many of the examples that I have read of in the various articles you have on your website.

 

Only one other person even considered that data integrity (in all its many requirements) was worth more than a passing glance. Most of the reactions seem to be on the lines that data integrity was an application issue not a database design issue. In a brief discussion with this fellow, he mentioned 6NF.

On my way back from the presentation to the office, I had time to discuss a database project that a couple of older gentlemen were working on and were having considerable problems with. It became obvious during the discussion that both of them were self-taught and really knew little about database design of any kind. I tried to (I hope) diplomatically direct them towards learning about relational database theory, but I don't think it was at all successful - I can be too obtuse at times. My wife has been trying for many years to get me to be clearer with I want to say.

 

My question to you both is in relation to a definition for any NF above fifth. Chris, I am currently reviewing your book "An Introduction to Database Systems" (3rd Edition) to review my understanding of the mathematical underpinnings of relational theory. It doesn't discuss anything after 5NF. Are there any available reference material that I can download before I get my copies of your books?

 

I am considering using your material to develop a training program in Relational Database Design for the local TAFE near where I live. TAFE is Technical and Further Education. Earlier this year I completed the appropriate certificate to allow me to design and give such training courses.

 

Chris, in relation to what you and Hugh have discuss in your book - are you basing your type and inheritance work on Lattice Theory. I have for a number of years (as time permits between God, family and work) been working on creating a type-based programming environment. The basis of this work is research done in the late 70's to mid-80's on type domains and algebras that was based on Lattice Theory and Denotational Semantics. The approach of one of your articles in one of your collected writings books (belonged to a friend) seemed to indicate that your were heading in that direction as regards the domains available for use in the Tuples of Relational Database Theory. I may have (and probably did) misunderstood your direction at the time.

 

Thank you both for your efforts in the field of Relational Database Theory.

 

 

From: Fabian Pascal

To: BR

 

Good for you. Enjoy.

 

You are not telling me anything I don't know. This has always been the state of the industry and the fault is not only with the people, but also with the system: not only does it not reward proper foundation knowledge, but also it actually punishes it, as you experienced yourself.

 

Two comments:

 

·   You don't have to worry about >3NF, except in certain situations when you have composite keys;

·   There is an article by William Kent on the five NFs, if you can locate it on the Net;

 

 

Chris Date Responds: Normal forms--Classical normalization theory admits of no level of normalization higher than fifth.  Fifth is final, for reasons explained in AN INTRODUCTION TO DATABASE SYSTEMS.  (By the way, I'm sorry to see you're reading the third edition!  The seventh edition came out in 1999, I think it was, and I'm currently working on the eighth.)

 

That said, I must add that the picture is muddied in at least two ways:

 

1.      Certain "database professionals" (I use the term somewhat tongue in cheek here), following the teachings of a certain well-known, influential, but not very technically aware industry figure I choose not to name, have always used the terms "Nth normal form" for various values of N > 3 in sloppy, incorrect, and misleading ways.  I well remember an argument I had on one of my seminars with one of these people who challenged the definition I'd given for fourth normal form ("That's not fourth normal form, etc., etc.").  I even wrote a short article about the issue (Will the Real Fourth Normal Form Please Stand Up? --see my book RELATIONAL DATABASE WRITINGS 1989-1991.  The problem was that these folks had come up with certain design recommendations of their own--no problem there--but, thinking it somehow made their ideas more academically respectable, decided to dress up those recommendations with the completely inappropriate terminology of "Nth normal form" for some N > 3.

 

2.      I'm guilty myself of recently introducing a new sixth normal form (6NF)!  However, I can of course defend my use of that term; it goes beyond what I called "classical normalization theory" above, and extends that theory in a way that really is "academically respectable."  See the book TEMPORAL DATA AND THE RELATIONAL MODEL, by myself, Hugh Darwen, and Nikos A. Lorentzos, published last month by Morgan Kaufmann.

 

 

Reference material:  I'd naturally like you to look at the seventh edition of the book, which includes coverage not only of normalization theory as such but also of a complementary theory, developed by David McGoveran and myself, called orthogonal design.  You might also be interested in the following articles by myself:

 

·   The Normal Is So ... Interesting (in two parts), DBP&D 10, Nos. 11-12 (Nov-Dec 1997). 

·   The Final Normal Form! (in two parts), DBP&D 11, Nos. 1-2 (Jan-Feb 1998). 

·   What's Normal, Anyway?, DBP&D 11, No. 3 (March 1998). 

·   Normalization Is No Panacea, DBP&D 11, No. 4 (April 1998).

·   Principles of Normalization

 

All but the last of these will I hope be included in my next WRITINGS book. The last one summarizes five normalization principles (not corresponding to five normal forms!).

 

Posted 03/07/03

 

 

 

[ABOUT] [QUOTES] [LINKS]