On Apr 15, 11:05 pm, "Brian Selzer" <b..._at_selzer-software.com> wrote:
>
> That's because you appear to have bought into the [...]
Anyone who's been reading this newsgroup for any length
of time is familiar with the fact that "correlating tuples"
is a pet issue of yours. Let's not rehash those old arguments;
it was not productive the first n times, and there is strong
reason to believe it won't be productive time n+1.
Much of the discussion rests on the importance of transition
constraints. I would like to discuss that. Are they important?
Fundamental? What theoretic basis do they have? What
are we to make of the fact that they are tuple oriented
when the RA isn't?
Perhaps the best way to frame this discussion is in terms
of a use-case. Please note that such a use case *cannot*
be stated in terms starting with "Assume you need to
have a transition constraint." That's called assuming the
conclusion, a particularly nasty logical fallacy. Instead
we should hope for something that will be a recognizable
business requirement, that could be enforced, say, in
application logic, but which could not be enforced centrally
except via a transition constraint. That I would find illuminating.
Anyone?
