Re: Auto increment

"Costin Cozianu" <c_cozianu_at_hotmail.com> wrote in message news:b9kodd$ka6pj$1_at_ID-152540.news.dfncis.de...
> Marshall Spight wrote:
> > "--CELKO--" <71062.1056_at_compuserve.com> wrote in message
news:c0d87ec0.0305101930.65eb8cca_at_posting.google.com...
> >
> >>[a bunch of good points.]
> >>
> >>There are better ways of creating identifiers.
> >
> >
> > Do these same problems apply to *any* system-supplied
> > key? It seems that your "The entire, whole, completed
> > set is presented to Foobar all at once, not a row at a time"
> > objection would apply to any such.
> >
> > But it seems pretty common that I just want to have a key,
> > and I don't care what the specific values are (as long as
> > they are unique,) and I'd like to have the system supply
> > it for me. Is there no clean way to do this?
> >
> >
> > Marshall
> >
> >
> >
> >
>
> In the worst cass scenario, the whole set of columns is your key.
> If the whole set of columns doesn't form a key, than you surely are
> making a design mistake.
>
> So you'll always have a natural key.
>
> Costin
>
>

I would tend to agree, except that I must accept the fact that there might not be enough discriminating attributes to uniquely identify what is conceptually termed an "entity" or a "relationship". There can be a wide variety of reasons in the practical world where this might be so. Unless you think, for example, that everyone should be capable of uniquely identifying a "person" by recording geo-spatial information and what-not as Celko has posited in the past?

Regards,

Dan Guntermann Received on Mon May 12 2003 - 07:15:14 CEST