Re: The Relational Model & Queries That Naturally Return Duplicate Rows

On 11 okt, 03:51, David BL <davi..._at_iinet.net.au> wrote:
> On Oct 10, 6:36 am, Erwin <e.sm..._at_myonline.be> wrote:
>
> > I'm also curious as to why he's asking to be pointed out what I think
> > should be obvious, but then again recent discussions on other subjects
> > have revealed to me that there is no predicting when people will fail
> > to see the obvious, merely because of certain kinds of mental block
> > they happen to be suffering from.
>
> I think the underlying problem is that we often fail to distinguish
> between belief and what's formally provable from axioms or perhaps
> more importantly to realise that it's an emphasis on the latter that
> leads to real progress in any branch of mathematics, and surely
> database theory should be treated that way (and not like a religion).
>
> The phrase "an ideal RDBMS" doesn't appeal to what's provable.  Can
> "ideal" be defined uniquely and mathematically?  I doubt it.

The universal meaning of "ideal" can perhaps not be pinned down, but if the OP speaks of "an ideal _R_ DBMS", then I think it is fair to infer from that that he speaking about a DBMS that scrutinously and faithfully follows all prescriptions of the relational model.

And the relational model says that a relation is a _set_ of tuples.

The relational model also defines very precisely what a tuple [value] is, and when and when not tuple values are to considered equal.

And a set is defined, in any version of mathematics I know of, to contain no duplicates, that is, no two distinct things that are equal.

So I think it is completely fair to say that the answer is indeed obvious that "an ideal RDBMS" should, in the case of his example, return a relation holding only a single tuple. Received on Mon Oct 11 2010 - 11:57:50 CEST