Re: SUPPORT FOR DECLARATIVE TRANSITION CONSTRAINTS

On Sep 22, 7:10 am, Erwin <e.sm..._at_myonline.be> wrote:
> On 22 sep, 03:31, Brian <br..._at_selzer-software.com> wrote:
>
>
>
>
>
> > On Sep 21, 12:52 pm, Erwin <e.sm..._at_myonline.be> wrote:
>
> > > states transition: {t1 t2} ---> {t3 t4}.
> > > tuple transition: {t1--->t3   t2--->t4} gets accepted.
> > > tuple transition: {t1--->t4   t2--->t3} gets rejected.
>
> > > You explain that (the difference between the rejected update and the
> > > accepted one) to a user (e.g., me) without having to resort to
> > > explanations that really are tuple-level.  That is, without using the
> > > word 'tuple', and relying exclusively on words such as 'set' and
> > > 'relation'.- Hide quoted text -
>
> > > - Show quoted text -
>
> > You haven't provided enough information to make that possible.
>
> Yes, I have, as I will show.
>
> > Constraints are a matter of semantics--that is, what the data means.
>
> And what the data means is the logical conjunction of the propositions
> that the tuples represent, or in my abstract example, "P(t1) AND
> P(t2)", and "P(t3) AND P(t4)", respectively, and this regardless of
> what the P() is.
>
> Nothing less, and most certainly nothing more.  And I do not need to
> specify the predicate for that claim to be true.
>
> Meaning that the transition itself, is not part of the semantics of
> the database.  Meaning that transition constraints are, contrary to
> what you claim, NOT about semantics.  And therefore NOT about business
> rules either.
>
> Also meaning that if you have one possible transition from "P(t1) AND
> P(t2)" to "P(t3) AND P(t4)" that gets rejected, and another one from
> "P(t1) AND P(t2)" to "P(t3) AND P(t4)" that gets accepted, then that
> means that your model is driven by something more than merely
> semantics, and that in turn means that your model is flawed.- Hide quoted text -
>
> - Show quoted text -

Again, you're not getting it. When defining state constraints, when a particular state obtains is not relevant, so in that context it's OK to abstract the universe to the point that it contains nothing but arbitrary objects independent of time. For queries, there can only ever be one current database, so in that context it's OK to abstract the universe to the point that it contains nothing but arbitrary objects independent of time. But when a transitions occurs is relevant, so in that context it would not be OK to abstract the uinverse to the point that it contains nothing but arbitrary objects independent of time. Your argument above is lacking because it doesn't take into account that since P(t1) and P(t3) were judged to be true at different times, P(t1) and P(t3) can mean the same thing at different times, or different things at different times. The context of P(t1) is different than the context of P(t3) because each database represents exactly what has or had been the case only during a particular interval in time and because no two databases can represent exactly what has or had been the case at the same time. The intervals during which P(t1) and P(t3) are true DO NOT OVERLAP! Received on Wed Sep 22 2010 - 15:52:39 CEST