Re: Can we solve this -- NFNF and non-1NF at Loggerheads

"paul c" <toledobythesea_at_oohay.moc> wrote in message news:YYgOd.334094$8l.51433_at_pd7tw1no...
> Dawn M. Wolthuis wrote:
>
> ...
>> compared to other models. I think someone was right on when they
>> suggested that if Date & Darwin want to redefine 1NF, then they should
>> call the new model by another name.
> ...
>
> that's like saying 'atomic theory' should be renamed merely because the
> term pre-dates the discovery of muons, pesons and so forth.
>
> i'd say D&D are simply REFINING their definition, which seems in the
> spirit of progress to me. it takes guts to admit one's past imprecisions.
>
> my personal cut, having once listened to Codd explain databases in terms
> of Christie Brinkley's 'phone number, is that incisive minds like his have
> little patience for revisionists. nobody can prove it now but i'd bet
> that his original point about 'atomic' values was merely a kind of
> place-holder while he was on his way to making his other points. he was
> trying to cover a lot of ground against tremendous resistance such as the
> powerful IMS marketeers in his own company.
>
> pc

I don't have his original paper in front of me, but I seem to recall that he explicitly left room open for RVA's. But didn't he state that he was avoiding them for simplicity of the model's sake. That is the power of RM to me. Simplicity with power. I find it interesting that if it was a placeholder, he never bothered to go back and fill in that whole in the twenty year's hence. However, he did try to extend the model "to capture more meaning".

With normalization we have a framework and a set of criterion for good design. A lot of that goes out the window with RVA's. For example, what is the criteria for determining when an RVA is appropriate or not in contrast to a flat model? What is "bad" about someone arbitrarily declaring the following:

R(S(T(U{A, B}))))

• Dan