Re: Relation Schemata vs. Relation Variables
Date: 4 Sep 2006 17:02:14 -0700
Message-ID: <1157414534.666540.38270_at_i42g2000cwa.googlegroups.com>
Brian Selzer wrote:
> [snip]
> This is a contradiction. A transition constraint must either be able to
> pair the elements of two relations for comparison or be given a set of pairs
> to compare.
Ok, what on earth is a transition constraint then? A transition occurs
when an 'entity' changes. In contrast a fact cannot 'change'. It is
either true or not, and that is that.
> In order for the transition constraint to pair tuples, there
The rest of this argument is based on a confused assumption AFAICT.
Received on Tue Sep 05 2006 - 02:02:14 CEST
> must exist at least one set of identifying attributes (a key) that is
> *guaranteed* to remain constant throughout an update--any update. One way
> to do that would be to specify which key will remain constant during a
> particular change, but without tuple identifiers it's possible for all keys
> to change at the same time. Do you treat such an occurance as a
> delete...insert? Just as there may be constraints that apply only to tuples
> that are about to be removed, there may be constraints that apply only to
> tuples that are about to be added. How can the system know whether or not a
> particular transition constraint applies? Or do you just allow the change
> based on the assumption that the user knows what he's doing?