Re: 1NF

On Jan 21, 3:35 am, "DBMS_Plumber" <paul_geoffrey_br..._at_yahoo.com> wrote:
> paul c wrote:
> > I would say SU doesn't mean the same thing as S. Maybe this isn't a
> > problem to anybody else, but it troubles me if you are saying that the
> > meaning can be preserved by unnesting alone, eg., without the
> > introduction of other attribures than P#. I agree.
>
> In the first place, the "unnesting" here loses information;
> specifically, how many of each part was shipped. That might be OK for
> the purposes of defining the operation but I observe that it will
> significantly complicate data modeling.
>
> In the second place, it's entirely reasonable to define a model which
> organizes the same information and doesn't require the same kind of
> mixing of sets and types.

I'd generally agree, with the addition that any such model will have no query bias, and quite happily nest results in a particular view to be served as results to a user if so desired. However I'd note that there are a couple of examples that have cropped up in previous threads with proposition sets which do require nesting to model in the RM.

>
> I'm a fan of the simple, old-fashioned definition of 1NF. Each
> attribute of a tuple has exactly 1 element value from a domain.

As good a definition as I've heard. Received on Sat Jan 27 2007 - 03:38:44 CET