Re: Pizza Example

"Jan Hidders" <jan.hidders_at_REMOVETHIS.pandora.be> wrote in message news:KVZec.70676$MG1.4809827_at_phobos.telenet-ops.be...
> Eric Kaun wrote:
> >
> > OK, I expressed myself badly. Logically, the relational user sees values
in
> > relations. The types of those values can be anything, including lists
and
> > such, but there shouldn't be operators in the data model to manipulate
> > those - rather, those are user-defined operators for the specified
types.
>
> So, one would expect that the NEST and UNNEST operators of the nested
> relational algebra would not be allowed, wouldn't one?

Do you mean GROUP? As far as I know, those are merely shorthand, not something new. And there's certainly no

> I know I would.
> What else could "logically expose" mean for a relation-valued column?

Operators on an attribute/column of type T are exposed, same as operators on any other type are exposed for use on values of that type. But one could certainly argue that the "nesting" in values/subvalues in Pick are simply exposed operators of a 2/3 level type. Hmmm. Mayhap I've argued myself into a corner. Or maybe it's just late.

> Or the nested relational calculus, which doesn't have such operators, but
> still allows you to operate on (and combine) the nested sets, would that
> be Ok?

Again, unless I'm thinking of the wrong operators, they're mere shorthand for combinations of other relational operators. Much like intervals, I think.

> > Is that better? Date's paper "What First Normal Form Really Means" talks
> > about this in great detail...
>
> Ah, well, let me say here and now that I'm not a big fan of Chris Date,
> to put it mildly, and the arrogance of dbdebunk makes me physically
> sick.

I can certainly see that, and I don't claim to be an expert - from what you've written, I'm fairly sure you're much more knowledgable than I on relational matters. But I find their site interesting, and useful as a bulwark against the wave of "novel" new data management techniques. While I'm not familiar with the "deep" research, I don't see much understanding of relational in common practice, and think it's certainly better than the ad hoc approaches being advocated.

> I know the paper you refer to, and I find it rather unconvincing.

I actually thought that was the point - to show the relativism inherent in normalization, including 1NF. I found that paper to be non-absolutist.

> -- Jan Hidders
Received on Wed Apr 14 2004 - 02:49:12 CEST