Re: Does Codd's view of a relational database differ from that of Date & Darwin? [M.Gittens]

David Cressey wrote:
> "Jan Hidders" <jan.hidders_at_REMOVETHIS.pandora.be> wrote in message
> news:jDyqe.116442$646.6800443_at_phobos.telenet-ops.be...

>> The relational model is more a *prescriptive* theory than a 
>> *descriptive* theory. In such a context these words are not that
>> far apart. Moreover, as a mathematical statement it would simply be
>> wrong to say that nested relations are impossible.

I couldn't agree more. Although I would immediately add that this doesn't mean that it is therefore in any way less valuable or less interesting, just that the types of claims that are made and the certainty with which they are made, are different.

>>> What I'd be curious to know is whether it's an undesirable burden
>>> to place on the DBMS implementor, an undesirable alternative to
>>> place in the hands of the database designer,  an undesirable
>>> avenue for further exploration of the theory,  or something else.
>> 
>> They tend to lead away from the goal that is so important in large 
>> shared databases: data independence.

The usual. If you nest in relation A the subrelations B and C then this is very convenient (i.e. easy to implement efficiently by the DBMS, easy to query by the user) if most of your acces patterns navigate from A to B or from A to C. But if the access patterns change, for example most navigation now goes from B to A to C, then you might want to choose another nesting such as B contains A contains C. But the whole point of data-independence was that applications would be shielded from such changes. The structure of the relations (their types) should not have to change because of changed access patterns.

• Jan Hidders