Re: Normalization and DBMS

From: Dawn M. Wolthuis <dwolt_at_tincat-group.com>
Date: Wed, 12 May 2004 09:23:44 -0500
Message-ID: <c7tc24$tk3$1_at_news.netins.net>


"Alfredo Novoa" <alfredo_at_ncs.es> wrote in message news:40a1e2c1.4991727_at_news-read3.maxwell.syr.edu...
> On Tue, 11 May 2004 18:25:19 -0300,
> =?iso-8859-1?q?Leandro_Guimar=E3es_Faria_Corsetti_Dutra?=
> <leandro_at_dutra.fastmail.fm> wrote:
>
> > 1NF was initially applied to clean our databases of such
> >horrors as:
> >
> > CREATE TABLE
> >(
> > subject,
> > year,
> > grade_jan,
> > grade_feb,
> > ...
> >)
>
> Yes, but this table is in 1NF.
>
> > These were known as repeating groups. Now there is a debate
> >about nested tables which I haven't looked into properly.
>
> I prefer: relation valued attributes :)
>
> The conclusion is that they are useful in very rare circumstances.
>
> For instance a good use could be:
>
> var Keys real relation { RelVar Char, Attributes relation {
> AttributeName Char } } key { Id, Attributes }
>
> > But I do
> >maintain that getting rid of repeating groups is The Right Thing To
> >Do, and one of the reasons why relational is so much better.
>
> Yes, but relations with "repeating groups" are still relations thus
> they are in 1NF.

According to the new definitions of normalization by folks such as Date, relation-valued attributes are now acceptable. So, the OO, XML, PICK and/or others who have accepted non-1NF data structures in spite of tremendous pressure from the SQL-DBMS folks to do otherwise just might have had some little influence in helping relational theory be a bit more useful. Of course, it is also possible that relational theorists, when looking at the theory itself saw that there was nothing logical about excluding "repeating groups" and simply added it in (thus significantly changing the def of normalization) without any outside influence from the peasants. smiles. --dawn Received on Wed May 12 2004 - 16:23:44 CEST

Original text of this message