Re: Is it a prorperty or entity

On 14 Jun 2002 23:10:14 +0200, hidders_at_hcoss.uia.ac.be (Jan.Hidders) wrote:

>In article <lgajgu4vg6pu0htb6jr471tce43peau96l_at_4ax.com>,
>GoranG <no_at_spam.net> wrote:
>>
>>Are rules here simply:
>>a) if meaning of other properties depend on a certain property (within
>>problem specifications)
>
>If the meaning of a property depends on something then it is not really a
>single property but should be split into a separate property for every
>distinct meaning. This may mean that you get optional properties, i.e.,
>properties that are sometimes undefined. If this happens and you get big
>clusters of properties that are always together defined or undefined then
>this is a sign that a subtype should be introduced. E.g., if entity E has
>properties a, b, c, d and e, of which a, b, and c are always defined and d
>and e are always together defined or undefined then you introduce the
>subtype E' with properties d and e and let E have only the properties a, b
>and c.

Agreed. This binding would indicate existance of functional dependancy at a new level - therefore a new entity woul be modeled.

But what is a 'big' cluster - more than two? Or some common sense and project dependant - or even entity dependant constant? :)

>>or
>>b) if meaning (or even type) of a relationship depends on a certain
>>property (within ...)
>>then divide entity on this property?
>
>Also here: if the meaning is different you have to model it as a different
>relationship. Don't forget you are doing conceptual modelling here, and not
>implementation. Mixing things with different meanings into the same column
>or table is for later. Conceptual modelling is about making things clear,
>not confusing them.

 Actually, I asked a wrong question, what I am actually puzzled about is:
What would be indicators of taking normalization too far (in conceptual and/or phisical model)?

Anyway, thank you very much for taking time to answer...

( GoranG79 AT hotmail.com ) Received on Mon Jun 17 2002 - 11:32:42 CEST