Re: The fable of DEMETRIUS, CONSTRAINTICUS, and AUTOMATICUS

Marshall Spight wrote:
> "Laconic2" <laconic2_at_comcast.net> wrote in message
news:yeGdnZXYRLK1jOjcRVn-tQ_at_comcast.com...
> >
> > A lot of the current discussion seems to be turning on the issue of
> > constraints, and basically
> > whether you express them as DDL, or as metadata, or as code that
enforces
> > them.
> >
> > But it's my impression that in the "ultimate normal form", all of
the
> > constraints follow from the definitions of the keys and of the
domains.
>
> This was roughly my thought as well. The *best* way to enforce
> a constraint is structurally, so that in the language of the
application's
> schema, it is *not possible to express* corrupt data.

That's great when possible. But some constraints just aren't expressible structurally - examples given in earlier posts. For those, what you want is a complex declarative constraint (aka an "assertion") - something many DBMSs can't do, unfortunately. Received on Tue Oct 19 2004 - 18:05:51 CEST