Re: A Normalization Question

> The real problem here is that Neo has conflated the concepts of duplication
> and redundancy.

With respect to dbs, normalization is the process of eliminating or replacing duplicate things with a reference to the original thing being represented. Within the context of a db, duplicate references are not considered redundant because among other reasons:

1. they are unrelated to the thing being represented.
2. they are implementation specific.
3. they typically aren't/shouldn't be exposed to the db-user.
4. they are automatically maintained by the db.

In the most general sense, normalization is the process of removing redundancy. If you need to hear this from an authority in order to accept this common sense idea, see C.J. Date's "An Intro to Database Systems", 6th Ed, Chapter 10, Further Normalization I: 1NF, 2NF, 3NF, BCNF, pg 280 where he begins with "what is wrong with this design: redundancy", "redundancy leads to several problems", "so perhaps a good design principle is 'one fact in one place' (ie. avoid redundancy). The subject of further normalization is essentially just a formalization of simple ideas like this". In the summary of the same chapter, he restates "normalization ... the purpose of such reduction is to avoid redundancy".

In addition, see pg 291: "By now the reader might well be wondering whether there is any end to this progression and whether there might be a 6NF, a 7NF, and so on ad infintum. Although this is a good question to ask, we are obviously not in a position to give it any detailed consideration as yet. We content ourselves with the rather equivocal statement that there are indeed additional normal forms not shown in Fig 10.2, but that 5NF is actually the 'final' form in a special (but important) sense."

Because RM is a limited data model and because some forms or normalization are either impossible or impractical, some RM users have a limited understanding of duplication, redundancy and normalization preventing them from recognizing it in a db containing the word "brown" serveral times in several tables. Received on Thu Jul 01 2004 - 20:53:04 CEST