Re: Generalised approach to storing address details

Neo wrote:
> > > I'd say that any data
> > > model besides the well-described relational one (not that I'm suggesting
> > > another one) would be right to follow the IP, otherwise it is hiding
> > > information which seems crazy to me.
>
> IP simply states that a relational db expresses it's content via
> relationA's tupleB's attributeC's valueD. Thus it would be ideal if db
> users would utilize only this system to encode all his information. In
> reality, db users routinely violate this system as in the T_Hierarchy
> and T_Person examples above. In some cases, db users knowingly violate
> this system in order to achieve desired goal

Just because they do this doesn't make it correct. A user's failings do not reflect on the system they are using. When you employ EAV to circumvent a schema, you are not using the RM as the logical model, full stop.

> (ie extreme flexibility to
> handle known and unknown data structures with minimal impact on
> schema/code) that is otherwise unachievable with RMDBs. What this
> points out is that RM's scope, while large, is not unlimited.

As ever this is nonsense. RM's scope is unarguable. All information can be recorded as propositions, and all propositions can be represented in the relational model. If you have some other complaint against the RM not to do with its scope, at least try to avoid expressing it in a nonsensical way.

> Once
> application requirements fall outside its scope, one has to resort to
> unsystematic methods (ie EAV) to achieve them. If you have a non-EAV
> solution that meets the OP's requirements, I would like to see it.
Received on Sun Dec 10 2006 - 21:53:20 CET