Re: Multiple-Attribute Keys and 1NF

From: David Cressey <>
Date: Thu, 30 Aug 2007 14:11:45 GMT
Message-ID: <BeABi.8794$CO5.3795_at_trndny01>

"JOG" <> wrote in message

> Ok, well the wire example was just that - a contrived example, and as
> it turns out not a particularly good one given that I can hardly argue
> against the fact that wire patterns are codes, and I can't really
> imagine why anyone would be interested in splitting them up into
> constituent colours (perhaps for an investigation of colour-blindness?
> Tenous).
> Another example that occured to me in the past were the propositions:
> Some cows have black fur and white fur.
> Some cows have brown fur.
> Some cows have red fur.
> Its exactly the same example in structure, but it is a lot harder (for
> me) to be comfortable with a 'code' view of the repeating roles. (and
> yup, again in RM I would use two relations and a surrogate for this,
> as opposed to an rva). Is everyone else still comfortable with a code
> interpretation? Dank je wel, J.

The big problem with contrived examples is that there is generally no answer to the question, "what do you intend to do with the data". One could contrive the answer, along with the rest of the contrived example, but most people generally don't.

The example of the cows sounds contrived to me. What do you intend to do with the data?

It seems to me that modeling as such cannot be meaningfully done in an environment where you can't distinguish between the essential features of the thing being modeled and the ignorable features of the same thing. That applies to data modeling as well.

In the absence of some way of distinguishing the essential features of the data, the data model has to either incorporate all the features of the data, or it has to incur the risk that some essential features are omitted. If some essential features of the data are omitted from the data model, the data model is defective.

If no features of the data are omitted from the data model, the data model is just as hard to understand and work with as the data itself. Hence the data model will not add any value.

May I suggest that you come up with an example that is not contrived, so that questions about what features are essential to the data model will have an answer? (My apologies in advance if the cows example was not contrived.) Received on Thu Aug 30 2007 - 16:11:45 CEST

Original text of this message