Re: Codd provided appropriate mathematics ... (was Re: Relational and MV (response to "foundations of relational theory"))

On Fri, 27 Feb 2004 08:01:43 -0600, "Dawn M. Wolthuis" <dwolt_at_tincat-group.com> wrote:

>> The Relational Model permits relation typed attributes, so
>> "multivalued" does not have any advantage. Collection typed attributes
>> don't violate 1NF.
>
>Yes, so the theory in RDM is evolving to the point where it might become
>more practical in its application.

The understanding of The Relational Model is better now than in the 70's, but the model is always the same.

But relation typed values in base relations seem to have little utility. The only good practical use I have found to the moment is to use them to implement a candidate keys relation variable of a catalog.

> But while each RDBMS vendor seems to
>have a means now of storing an array (or relation or other collections),
>querying against this data is done differently by each vendor, it seems,
>although SQL-99 has a means of doing this (and perhaps more db vendors are
>employing that standard, but when I checked in 2002 there was no
>consistency).

As you heared hundreds of times, SQL is not relational. You will not find the term relational in the SQL specifications.

>> Relations are in 1NF by definition. With a RDBMS you can not violate
>> 1NF even if you want.
>
>You can violate any of the other normal forms, but not 1NF.

Because you have no way. Relations are in 1NF by definition, otherwise they are not relations. If a DBMS does not use relations then it is not a RDBMS, so if you can violate 1NF then you are not using a RDBMS. A simple tautology.  

Regards
  Alfredo Received on Fri Feb 27 2004 - 18:19:31 CET