From: JG

To: Editor

Date: Thu, 13 May 2004

I appreciate the points you make, and their value – but I was taken aback by one remark in the conclusion to [PRACTICAL DATABASE FOUNDATIONS paper #2] What First Normal Form Means Not.  You say:

Because MV files are not relations, and MV products are not TRDBMSs, normalization is essentially irrelevant.

Since all widely used DBMSs permit the construction of tables that are not faithful to the definition of a relation, and no widely used DBMS is a TRDBMS, would it not follow from your premises that normalization is essentially irrelevant everywhere?

While acknowledging that I am here veering off at a slight angle from your original intent, I would argue that the application of normalization skills is just as important – perhaps even more so – in the design of databases that are not implemented in an RDBMS, as in the design of those implemented in DBMSs with pretensions to R-hood (i.e., those built around SQL).  Whether the target platform is flat files, indexed files, XML, IMS, SQL, Excel, or other, an understanding of the meaning and value of the normal forms is critical to making the system flexible, extensible, comprehensible, maintainable, etc.  N’est ce pas?

From: Fabian Pascal

To: JG

I was trying to debunk a common misconception that normalization is all there is to relational fidelity. By "irrelevant" I meant that with MV products, full normalization does not recover full relational benefits and guaranteed logical correctness in that sense. Furthermore, in some instances--e.g. indexed files--it may also complicate your work considerably, which may counterbalance some of the benefits from it.

Posted 08/13/04