Re: Little question for RDM theoristes

From: Cimode <>
Date: 16 Jun 2006 06:27:40 -0700
Message-ID: <>

I am aware of that definition and I do not quite agree with it because it leads to confusion...

CJ Date is well known for using ambivalent terminology for vulgarization purposes towards SQL audiences...I personally admire his patience with SQL people given the bad it did to RM on the past 20 years...
For instance he uses Tables for relvar projection (which typically creates confusion with SQL tables) while Pascal prefers R-Table...

I am more a follower of the FP, McGoveran approach who advocate a tighter commitment to terminilogy ...

A relation is BOTH a relvar which represent the abstract structure of the relvar and the relvalues which represents its matter at a specific point in time.

Think about the implication of stating that a relation1 = (set of relvalues)1 (relvalues drawn from relation1 domain1 of possible values)

if relation1 = (set of relvalues)1 of domain1 and relation2 = (set of relvalues)1 of domain1

You can conclude that relation1 = relation2 IF AND ONLY IF you assume all values have the same location in a multidimensional representation...Admitting such axiom would lead to accept that all relvalues in a specific relvar are always located at the same position at any time...Think about the implications: All projections would be then necessarily ordered sets of values...a relation is not an unordered set of values at logical level....This would be totally silly

Therefore defining a relation only through its values is unsufficient to allow to both represent faithfully and operate the relation...But SQL people need this kind of confusion to make sense of what a relvar is...

Erwin wrote:
> > Your question implies relations = relvalues...which if I follow this
> > false premise reasonning would lead to relations that have similar
> > relvalues being equal which is totally false...2 relvar with same
> > relvalues are NOT necessarily equal.
> TTM Chapter 4, RM prescription 10 :
> "A relation value (relation for short) ..."
> Therefore at least to Chris Date, 'relations=relvalues' is most
> certainly true. I'd say that's a strong indication of just how much
> "false premise" there is within.
> > This question is totally irrelevant if you consider a relation as being
> > equal to a relvalue...
> This question is not irrelevant at all since the heading is regarded as
> the definition of the applicable relation type. And for values to be
> equal, they must most certainly be of the exact same type, inheritance
> issues notwithstanding of course.
Received on Fri Jun 16 2006 - 15:27:40 CEST

Original text of this message