Re: object algebra

> Nulls in SQL are the result of a design decision
> made by the inventors of SQL.

The above is incorrect. Even if you remove the SQL layer, RDM will need to deal with NULLs. The original source of NULLs is not SQL. It is RDM that creates NULLs under some situations, unless one resorts to generic modelling which in the extreme case requires a two-column table which is impractical.

> They (wrongly, IMHO) thought that 3VL would be a better way of handling
> missing information than having a separate table with just a foreign key
> and the optional info. It is certainly possible to make a different
> decision in designing future DBMSs.

Since it is always possible to enter a tuple that lacks an attribute defined by the header at design time, the possibility of incurring a NULL can't be elminated unless you build upon a two-column (of which one column is automatically supplied by the db) table which is impractical.

Unless RDM changes its model, NULLs are forever an integral part of RDM as Codd himself has realized. Why not prove him wrong by providing us will an implementation that does not require NULLs in the general case? It is impossible, unless one resorts to generic modelling which in the extreme case, would result in a two-column table and is impractical.

When/if you are able to implement www.xdb1.com/Example/Ex076.asp you will begin to see that pattern arising. Received on Thu Feb 26 2004 - 16:54:28 CET