Re: Implementation of boolean types.

From: dawn <dawnwolthuis_at_gmail.com>
Date: 19 Jul 2005 12:07:50 -0700
Message-ID: <1121800070.904775.9810_at_g43g2000cwa.googlegroups.com>


Marshall Spight wrote:
> Drago Ganic wrote:
> >
> > > Symbolic programming is a good solution for NULL values....
> > should be
> > Symbolic programming is a good solution for UNKNOWN values ...
>
> I figured that was what you meant.
>
> However, I have to say, I question whether unknown values is really
> the right problem to be working on. I think it it's. I think
> the better, easier approach is to model the empty set,

Yup! I've worked with tools that have a null value (yes "value") that models the null set and I've worked with SQL and it is so much easier and more intuitive to work with a 2VL and null value = empty set model. Design, development, trouble-shooting, and maintenance are all significantly easier with this model.

This is one of several features where those attempting to implement relational theory took what was working in databases in the 70's (such as 2VL) and mucked it up, contributing to greater cost of ownership for data-centric applications. Taking just this one issue, I suspect the path to getting the industry from here to there is via XML/XQuery, perhaps? However, I did read that SQLServer has an option of switching from 3VL to 2VL.

cheers! --dawn Received on Tue Jul 19 2005 - 21:07:50 CEST

Original text of this message