Re: Oracle 7.1 Enhancements

Wolfgang Roeckelein (wolfgang_at_wi.WHU-Koblenz.de) wrote:
: No! Read Codd, who defined what key points a relational database should
: have and one thing that he clearly states is that NULL has to be different
: from anything else. NULL denotes a missing value (unknown), whereas a
: empty string is a string. E.g. a telephone number: NULL means unknown (no
: information), empty means the guy has none (information).

I feel that this null business is an example where relational mathematics goes to far from the practical life. Since mathematics tell that there are different meanings of NULL, and that nulls may or may not be eqal to each other, our rdbms vendors take the easiest way and do nothing. And we pay the bill.

Suggestion: a null is a null is a null - whatever mathematics say.

First thing to do would be to give us some way to define that, in OUR installation, it is preferrable to considers NULL as something different from anything else. A simple INIT.ORA parameter could be fine.

This would make it possible to include null values in index trees and compare null with other values, for example. I don't care so much where nulls are bigger or smaller that others.

Regards,

        Hessu

-- 
---------------------------------------------------------------------
Heikki Rissanen, United Papermills, Rauma, Finland
Email: heikki.rissanen_at_ra.upm.yhtyneet.mailnet.fi
---------------------------------------------------------------------