Re: SUPPORT FOR DECLARATIVE TRANSITION CONSTRAINTS

On Sep 21, 1:02 pm, Erwin <e.sm..._at_myonline.be> wrote:
> On 21 sep, 04:47, Brian <br..._at_selzer-software.com> wrote:
>
> > I can't see how it does.
>
> See my other reply.
>
> > Since it is obvious that
> > keys cannot be relied upon to provide that precise determination,
>
> I would have hoped it was equally obvious to anyone that nor can
> objectid's, or whatever other kind of meaningless identifying
> construct one wishes to use.
>
> After all, wasn't it you who suggested that Codd wrote that even
> 'regular' keys are really just surrogates just like any other kind of
> identifier ?  And shouldn't that be sufficient to prove that any other
> kind of identifier would suffer from exactly the same "deficiencies"
> as regular keys do, when it comes to enforcing transition constraints
> in the situations you complain about ?

I think you misunderstand. Permanent surrogates do not suffer from exactly the same "limitations." If every relation has keys that are permanent surrogates, then it is possible to determine which transition the user intended, unless it is significant that a tuple in the relvar as it was prior to the update also appears in the relvar as it was . The problem is that not all keys are permanent surrogates, and not all relations have keys that are permanent surrogates. In order for all relations to have keys that are permanent surrogates, some relvar headings must be altered. I don't think it should be necessary to alter relvar headings in order to enforce transition constraints, and it clearly isn't. In what I propose, key stability does not bear on whether it is possible to determine precisely which transition the user selected. Received on Wed Sep 22 2010 - 05:39:38 CEST