Marshall wrote:
>
> The idea of transition constraints is inherrently non-set-theoretic,
> because it assumes tuple-identity. Since it is possible to
> model non-tuple-identity scenarios in the RM, there will be
> some scenarios in which transition constraints do not make
> sense. This is not a flaw in RM; rather it is an indication that
> transition constraints are a lower-level albeit useful hack
> on top of some subset of RM schemas, that are applicable
> only some of the time.
>
More precisely, the idea of *tuple-level* transition constraints is
inherently non set-theoretic. Fortunately, logical operations of any
kind at tuple-level are not possible or permitted in RM. The only
applicable way to define a transition constraint is in set-based
fashion at the logical level. Truly relational transition constraints
would therefore suffer none of the problems mentioned.
--
David Portas
Received on Sun Aug 20 2006 - 19:22:42 CDT
