Re: Why not many2one with pk array type

"dawn" <dawnwolthuis_at_gmail.com> wrote in message news:1138578631.065163.264020_at_g43g2000cwa.googlegroups.com...
>
> David Portas wrote:
> > Sandy.Pittendrigh_at_gmail.com wrote:
> >
> > > I disagree. Relational design is inescapably subjective.
> > > Normalization rules are not rules, they are guidelines.
> > > No two programmers model the same complex data problem the same way.
> >
> > There are subjective design decisions to be made in hierarchical models
> > too of course,
>
> agreed
>
> > but RM has enormous and proven advantages over the
> > hierarchical ones.

> I have looked for these proofs. I can see trade-offs where one
> approach has advantages over another approach, but can you point me to
> something that nets it out and really proves that the RM had advantages
> over any other model?

When you compare the models what criteria do you have in mind ? I have heard that when one try to formalize a network DBMS (in other words to extract a network model of data from it by making abstraction of the implementation details) one get a complex model that is subsumed by the simpler relational model but I have not checked that myself. There is also the question if the implementation of the relational model follow the assumptions used in the above proof.

> > >
> > > That is now and never will be possible with relational systems.
> >
> > An RDBMS can or could expose exactly as much metadata for a GUI mapping
> > tool as a hierarchical one can - for example through business rule
> > constraints exposed in catalogue views. Are you claiming that there is
> > metadata inherent in a hierarchical database that cannot be represented
> > relationally? That would seem like a startling claim. Can you
> > substantiate it with an example? If not, it seems that your program
> > must have exactly the same opportunity to map a GUI from both
> > relational and hierarchical systems.

> I suspect this has more to do with the available operations than with
> whether the data could be mapped. cheers! --dawn

Experiment with the relational division operator. Received on Mon Jan 30 2006 - 11:06:22 CET