On Thu, 2002-08-08 at 06:47, Curt Sampson wrote:
> On 7 Aug 2002, Hannu Krosing wrote:
>
> > > Theory: Sure. But this is much harder to express in a turing machine
> > > isn't it?
> >
> > You got it ;) The claim was that it is easiest to express it using
> > inheritance, a little harder using pure relational model and much harder
> > using a Turing machine.
>
> Ok. I agree that it's much harder with a turning machine. I do *not*
> agree that it's harder with the relational model. In fact, since you
> *must* use the relational model for some things, I argue that it's
> harder to switch back and forth between the relational and OO models,
For me they are _not_ two different models but rather one
object-relational model. Same as C++ in _not_ a completely new language
but rather an extension of plain C.
As you seem to like fat books, check out :
"Object Relational Dbms: Tracking the Next Great Wave" by Michael
Stonebraker, Dorothy Moore (Contributor), Paul Brown
ISBN: 1558604529
I'm sure you find the requested arguments against Date there ;)
> and understand the effects of each on the other, than it is just to do
> it in OO form in the first place.
>
> In fact, I'd argue at this point, as far as table inheritance goes,
> we don't even have a real model here.
The table inheritance _implementation_ in PG is in fact broken in
several ways, most notably in not enforcing uniqueness over all
inherited tables and not inheriting other constraints.
But as you often like to emphasize, model and implementation _are_
different things.
--------------
Hannu
