Re: Normalization by Composing, not just Decomposing

From: Eric Kaun <>
Date: Mon, 12 Apr 2004 20:58:52 GMT
Message-ID: <gqDec.7650$>

"Alan" <> wrote in message news:c5erh9$qt9b$
> Normalization rules are Codd's rules, not God's rules. They are a _guide_
> distributing data among entities, not a dogmatic recipe.

They're more than a guide, though certainly not divine law. They're based on some fairly black-and-white rules, once you can decide on what propositions you're modeling, and give far more solid advice than other modeling approaches (entity-relationship, anything XML-ish, etc.), which are far fuzzier.

> You seem to want to
> project a certain amount of dogmatism on everything, as if life were black
> and white.

Dawn? I never got that impression from her - more the opposite (she's far more situationally-minded than I am, for example).

> It isn't, it's an infinite number of shades of gray (well 16,384 at

Yes, there is ambiguity in life, and there are choices in data and systems design. That doesn't imply all of them are of equal merit.

> Here's the general rule of thumb: Normalize to 3NF, and then see if that
> works for you in your situation.

So what's the situation? The app you're working on? The needs of the business? How long a timeframe?

> If it doesn't, then denormalize or normalize further. Iterate.

Uh... ok. Sound advice (cough).

  • erk
Received on Mon Apr 12 2004 - 22:58:52 CEST

Original text of this message