Re: Nested Sets vs. Nested Intervals

amado.alves_at_netcabo.pt wrote:
> Some semantic domains are better modelled with graphs and trees than
> with relations. The idioms for trees and graphs in SQL is extremely
> convoluted to say the least (the word "pathetic" comes to mind). The
> original problem in this thread is an concrete example.

This is because the domain is poorly understood. I would challenge the idea that XML people have clearer understang of graph problems. Tags don't magically solve problems.

> Relational star people themselves recognized shortcomings of the
> relational theory for semantic modelling. This was a major factor in
> the creation of the Entity-Relationship model. And guess what: the E-R
> is a graph-based model. (Not to mention UML class diagrams, but I know
> what you'll say: these are for programs, not data.)

Well, the join operation is another binary relation that can be interpreted as edges in a graph. Excuse me, but that is not a particularly deep idea. I fail to see it utilized anywhere except query optimization.

> There's graph-based and graph-based. As you I also don't subscribe to
> the socalled "hierarchical" and "network" models as they are often
> presented in relational circles. They seem to be a heritage of Codasyl
> and alike systems of the 1960's. Fortunately others kept researching
> and have come up with much better graph-based models e.g. Resource
> Description Framework and even XML.

It is not a matter of "do hierarchical and network models make sence". It's a matter of how nice is the underlying theory. So far relational people seems to be unimpressed. Received on Wed Nov 09 2005 - 19:09:59 CET