Re: A searchable datastructure for represeting attributes?

> The structure in his system is slightly different. He put attributes
> with different types in different tables, each type in a separate table.
> I think this helps keeping values more selectable (raises selectivity)
> and automatically checks domain rules.
>

Depending on what you call a "type", and what you call a "domain", your answer might begin to make sense.

If there can be two different domains with datatype "char (5)", say 5 digit zip codes and employee id's, and
we put each of them in its own table, as the PK, then each table serves as a reference point for its own domain.

But if you do that, then it's unnecessary to store metadata in user columns. You are now back to the relational data model (at least in this regard). Hurrah!

If, however, you store all the "char(5)" keys in the same table, even when they don't refer to the same class of object, then you have to use metadata in a user column to disambiguate. If you don't want your data to be self describing, then maybe you don't want to put your data in a database, after all.
  --
Regards,

    David Cressey
    www.dcressey.com Received on Thu Feb 28 2002 - 20:16:07 CET