Re: Has E/R had a negative impact on db?

David Cressey wrote:
> "Jon Heggland" <jon.heggland_at_idi.ntnu.no> wrote in message
> news:e2hslc$sfq$1_at_orkan.itea.ntnu.no...
> > -CELKO- wrote:
> > > One problem is that a relationship in a ER diagram is a line between
> > > boxes. That puts us in a binary-only world. ORM, on the other hand,
> > > allows n-ary relationships.
> >
> > Are you talking about Chen's original ER? Because most ER variants I've
> > seen definitely allow n-ary relationships---and the relationship aren't
> > the lines, they're the diamonds.
>
> Agreed.
>
> And another way to depict n-ary relationships in an ER model is called
> "reification". That's just a fancy word for expressing
> a relationship as if it were an entity. The reified relationships play the
> role of diamonds in the above.
>

And what exactly defines a relationship as not being an entity in the first place? This is a rhetorical question of course, as I cannot seem to discern any boundaries. It appears to me that this is no more than an arbitrary, and artificial, decision.

> The UTexas outline of modeling suggests doing this for resolving
> many-to-many binary relationships. They call it "association entities". I
> don't doo what they suggest, but I have done it for n-ary relationships.
>
> A second parallel question is whether attributes can be attached to both
> entities and relationships, or whether they can only be attached to entites.
> I've followed the first practice.

Bill Kent (a great loss imo) would have disagreed with the second line of thought wholeheartedly. Relationships may always have attributes themselves and indeed, if one views them as just links then it's a lot tougher to change diagrams in the future when one discovers that they do in fact have attributes that require modelling. Why then not just think about them as n-ary relationships (or associative entities if preferred) at all times?

>
> In my own mind, without developing any formal concept for it, I've had a
> tendency to view entities as "unary relationships". I realize that's an
> oxymoron, but it's intentional.
> This collapses two concepts, entity and relationship, down to a single concept, relationship.

I am unclear here David why one would want to view an entity as a unary as opposed to an n-ary relationship. An 'entity' has numerous items in it all playing a role in that association, just as a 'relationship' has. Received on Mon Apr 24 2006 - 15:21:32 CEST