Re: Do Data Models Need to built on a Mathematical Concept?

> I don't think anybody is claiming that the Relational Model
> (as we know it now) is the final answer.

IMO, Mr. Cozianu seems to imply that the final model for representing data is the current relational data model. Although the relational data model is sufficient for the majority of applications (currently), it is not sufficient for all applications. I am certain he is wrong on this account.

> However, what is important to understand is that the whole body
> of modern mathematics is based on the "relational model", in a sense.

If by "relational model" you simply mean that things can have relationships, then I would agree.

If by "relational model" you mean Codd's "relational data model" where each tuple in a relation has to be of the same degree and each value of a domain has the same type, then I would disagree with the use of the quantifier "whole body". The arbitrary restrictions in Codd's "relational data model" prevent it from being the basis for the WHOLE body of useful mathematics that man can invent. The arbitrary restrictions make the implmentation easier but also limited its flexibility.

> As for using the human brain as a model:
> There is a theory called Bayesian Networks that handles
> fuzzy reasoning well (it is actually based on a statistical model).

By the human brain model, I meant only that area related to representing things. I know for certain that it is not restricted by data type and the need to have same number of attributes for each thing.

> Maybe the elements of BN can
> be merged with RM in due time...?

For certain there is great overlap.
For certain they cannot be fully merged without breaking each model rules (which were invented by us). Received on Sat May 03 2003 - 20:04:22 CEST