Re: Question re: Practical Issues in Database Management

Joe Thurbon wrote:
> I've been reading some of the books recommended to me recently by c.d.t
> readers, and have a question regarding the Pascal's "Practical Issues in
> DB Management."
>
> My question is: "Have I misunderstood the Pascal definition, or is it
> incomplete/incorrect."
>
> In 5.2.2.2 'Multivalued Dependencies' there are two definitions
>
> Multivalued Dependencies: 'An MVD between two columns exists when sets
> of values in one column are each associated with values in another column'
>
> 4th Normal Form: 'If no multivalued dependencies exist between columns,
> a table is in 4th normal form.'
>
> This definition seems too restrictive. In particular, under those
> definitions, the following tables (AFAICT) would not be in 4th normal
> form. (Although in his book he offers this as an example of 4th normal
> form.)
>
> EMP# Project | EMP# Activity
> ==== ========= | ==== ========
> 1 Services | 1 DEBUG
> 1 Education | 1 SUPPORT
> 2 Services | 2 DEBUG
> 2 TEST
> 2 CODE
>
>
> From other reading that I've done, the Pascal MVD definition seems a
> little different to the standard ones. e.g.,
>
> http://www.utexas.edu/its/windows/database/datamodeling/rm/rm8.html
>
> has a definition that involves a third column, as do several other sites.
>
> Many thanks,
> Joe

Joe, I'm guessing that Fabian P meant to say something like "If no trivial multivalued dependencies exist ...".

As Date puts it, "The MVD A->->B is trivial if ... or the union AB of A and B is the entire heading.

I'm also guessing that what Date is driving at is the same as the authors of the def'ns that involve a third column, ie., there is no point in trying to decompose a two-column table (at least one that has no RVA's).

p Received on Sat Mar 24 2007 - 17:47:20 CET