>Playing devil's advocate:
>
>One of the benefits touted for Relational Database Model at its introduction was that the keys could be based on data as opposed the pointers used in the Hierarcical and Network Database Models.

No, the actual benefit that was touted is the idea that parent and child relationships are based on keys instead of fixed in the model. Keys are, in fact, a type of pointer.

The difference is that in the relational model the "pointers" are part of the data in the record and can therefore be used in a variety of ways, while in the other models the pointers are part of the underlying schema and are set at design time and it is difficult to rearrange the view of the data at runtime.

Surrogate keys are no less "part of the data" than natural keys are. They are a field in the record and not some system pointer.

>>As these database models are much faster than the Relational Database Model, shouldn't we consider using the Hierarcical or Network Database Model? After all the Object Databases that are so new and exciting are Hierarcical or Network Databases wrapped in new terminology.

Yes, they can be faster but they are limited in flexibility. They work reqally well for getting phone bills to 20,000,000 custoemrs, but the phone company surely has either a relational or data warehouse for business analysis because teh hierarcical structure makes it very difficult to analyze the data from various perspectives. It (hierarcical) is very to do analysis of the data i teh view that was used to set the hierarchy.