Re: Help on normalization

>I am now working on a project on how far should normalization go.

Ron Fagin (http://www.almaden.ibm.com/cs/people/fagin/) determined back in the 70's that fifth normal form is both necessary and sufficient to prevent update anomalies using non-loss decomposition by projection.

Conceivably, one could define whole other normalization processes using other non-loss decompositions based on other operators. I do not recall whether anyone has ever attempted such a thing.

Additionally, Chris Date strongly suggests that every relation in a database should have a unique predicate. At one time, he was calling this "The New Design Principle".

For a comprehensive treatment of this issue, I would recommend reading Date's work. Received on Fri Feb 23 2001 - 02:21:36 CET