Re: Normalizing the ER model

Greg Boland wrote:
>
>"Jan Hidders" <hidders_at_REMOVE.THIS.uia.ua.ac.be> wrote in message
>news:3df142a5$1_at_news.uia.ac.be...
>> Greg Boland wrote:
>> >
>> >I like ORM because of the contraints one can apply to relationships.
>>
>> Good. I like ORM because it has a good linguistic and philosophical
>> foundation and its semantics are formally described. :-)
>>
>> >But in my experience, these constraints are usually implemented by
>> >stored procedures or some other procedural logic. So what does
>> >normalization have to do with it?
>>
>> A fact type is a relation so all of normalization theory applies to it
>> directly. For example if you have a fact type with the roles Street,
>> Number, City and Zipcode then it is not in BCNF if Zipcode only
>> determines Street and City. However, AFAIK there is no graphical notation
>> for functional (or other) dependencies in ORM.
>
>I agree, and you stated my point. What I'm saying (or attempted to say) that
>ORM relationship constraints cannot be implemented in the relational model .

That's nonsense. All ORM constraints can be implemented in the relational model.

>The fact type is simply a relation, and no constraints(except for
>cardinality) can be applied to it.

That's also not true. You can also have key-constraints and inclusion and exclusions constraints and even more.

>So I still contend that normalization for the fact type does not enforce
>constraints.

It does in the same way and for the same reasons as when you normalize directly in the relational model.

• Jan Hidders