Re: A Logical Model for Lists as Relations

Bob Badour wrote:
> Jay Dee wrote:
> > Bob Badour wrote:
> > > Since it is relatively easy to write a query that extends a relation
> > > with a rank per any explicit order, I am not even sure the ordinal
> > > attribute is required.
> >
> > True. I wasn't exactly sure what MS wanted to happen, for example,
> > to the fourth element in a list when the second element is elided.
> > Does it become the third? If so, your proposal is a good
> > solution. But...
> >
> > What if he envisions a list in which duplication of elements is
> > significant?
>
> If one has a numeric index that differs for each tuple, one never has
> duplication. If one has duplication, one wonders how to refer to the
> duplicates. As Codd observed long ago, once one has said a thing is
> true, what does saying it again achieve?

Wise words from the Codd. However, I think people confuse (and by people I essentially mean myself) trying to encode statements such as:

There is a person called Ernie who likes beer. There is another person called Ernie who also likes beer.

Ex likes(x, Ernie, beer) &&
Ey likes(y, Ernie, beer) &&
x != y

The two Ernies are obviously different people but I don't have enough information to harness liebniz equality and so distinguish them, I just have the fluff of 'another' saying that those distinctions do exist. So I obviously use a surrogate to represent the missing information. Something seems sliiightly artificial about this though - I just can't put my finger on what at the moment. (perhaps in terms of logic I should just be encoding "There are 2 people called Ernie who like Beer"?).

>
>
> In that case - and I'm supposing that's what MS
> > had in mind - a relation of element values wouldn't work.
> >
> > "More requirements, please!"
> >
> > In the end, though, I'm confident that the lists MS desires can be
> > implemented in a relational design without inventing anything beyond
> > what already exists.
>
> I still do not know what sort of requirements would cause me to want a
> list in the first place.
Received on Thu May 11 2006 - 15:40:07 CEST