Normalization Theory

On Sunday, February 1, 2015 at 3:29:28 PM UTC-8, Derek Asirvadem wrote:
> Please be advised that when I teach RDb Design, Normalisation is a major subject, and I approach it as science. The first thing I do is excise the abnormal "normal form" definitions which are marketed by the schizophrenics that wrote them, retaining the three Normal Forms that undamaged human beings need. Then I teach the science, complete with exercises that use several tables (frauds use single "relations" to construct their Straw Man arguments; papers; mathematical poofs). Then I teach the two scientific Normal Forms that are /informally/ but scientifically (at least to those who have their feet on the ground) defined in the Relational Model (the concrete Codd RM, not the 42 religious artefacts).

There are 2 sides of the issue. Is database normalization theory polished enough to be presentable to an average undergrad? My answer is no. However, is normalization topic a gem of database theory? Yes. Has it progressed since 1970s? Let's postpone this answer a little.

> The fact that the theorists continue to define schizophrenic "normal forms", the purpose of which is to justify Record Filing Systems that are non-relational, in forty five years, stands as evidence that the theorists in our field understand only Record Filing systems.

The fact that you are so upset by Normal form definitions indicates that you pay no attention to what happens underneath. I'm little uncomfortable with those voluntaristic definitions as well, but happily ignore them because there is something more fundamental going on. In case if you are wondering how normalization theory progressed since its early days:

1. The first foundation that was explored quite exhaustively in last couple decades was information measure (entropy). Is entropy "not scientific" enough for you?
2. Algebraic approach was also studied quite in depth. Remarkably, the major tool in algebraic approach is algebra of binary relations, and one of the main results is that relation with multivalued dependency is formally described by commutativity of relational composition.

To conclude, a typical database introductory course is quite odd. The first half studies relational algebra, but later almost nothing from that kit is utilized when teaching normalization theory.   Received on Mon Feb 02 2015 - 19:18:07 CET