Re: Multiple-Attribute Keys and 1NF
Date: Thu, 30 Aug 2007 01:01:10 GMT
Message-ID: <qFoBi.106010$fJ5.63845_at_pd7urf1no>
mAsterdam wrote:
> JOG schreef:
>> mAsterdam wrote:
>>> JOG wrote:
...
>> What I am really struggling with is whether requirement of a surrogate
>> key to achieve the information need (as a result of enforcing 1NF) has
>> a theoretical basis. Is 1NF just there to provide a structure that
>> fits a relation, or does it have a deeper purpose that I haven't yet
>> grokked.
> > The need is a intersubjective topic /ergo/ not formal. > ...
I thought that the reason for "surrogates" was entirely practical. I wouldn't put it quite like "structure that fits a relation", maybe instead "added structure that allows us to express some relations we couldn't otherwise". Also suspect it isn't quite that simple, for the Information Principle seems also to be involved. I'm not a mathematician but that still disturbs me a bit, because I'm not able to "separate" the two concepts of 1NF and IP in this case.
I also think 1NF is a compromise, but a practical one. It allows certain operators such as natural join to be defined for all relations without respect to the intricacies of whatever domains they use. I guess it's also inherent in 1NF is also that any relation name, names a relation that can be expressed in a "regular" or "rectangular" form, ie., where all propositions in one relation give values for the same set of predicate variables, sorry if that makes me sound like a mediocre math fan.
The subject of "or" was mentioned somewhere. It may be that different people take JOG's musing in different ways according to their interest. I admit mine has very little to do with modeling philosophy which for me is nothing more than agreement about an application and nothing to do with scriptural interpretation (of a mathematical metaphor) as Bob B calls it. Once people agree, the machine takes over.
I took JOG's intent to be asking what happens if an attribute always refers to a set of values of the domain it is associated with in a particular relation. I also presumed that he wants to maintain the use of predicate logic. The intricacies of a set are limited compared to those of all the domains people are capable of thinking up. For me the main question in wondering about an algebra for this would be to decide where and when the idea of subset would apply, eg., for negation/not, union/or, conjunction/and.
Just my two cents,
p
Received on Thu Aug 30 2007 - 03:01:10 CEST