Re: boolean datatype ... wtf?

On 06/10/2010 2:44 AM, Erwin wrote:
> Even Brian acknowledged that NULLs are (what he called) "indicators",
> or iow, boolean flags, or yet iow, 2VL truth values. They say whether
> data is really present yes or no. That sounds very boolean to me.
>
> Now guess what. Would the fact that I think it is better to eliminate
> booleans from a logical database design by vertical decomposition,
> have anything to do with the fact that I also think it is better to
> eliminate nulls from a logical database design by vertical
> decomposition ?
>

Reminds me of Codd who frequently mentioned 'marks' of various kinds. Not to put words in his mouth, but that often gave me the feeling he was straying into the hardware realm, as if they were a hardware signal to the logical system to 'switch gears'. Many of the machines Codd used in the 1950's and 1960's had 'variable word' sizes, with reserved byte configurations that would signal physical exceptions. Some of these were called 'field marks' and 'word marks' and their use even persisted into the 1990's on certain IBM terminals. When hardware designers in their wisdom moved away from such 'modal' features the notion survived, eg. I think it was Kernighan and Ritchie who put the string delimiter in the C language.

It was some years after Date's first books before he started to treat projection precisely. That was when I began to wonder how a projection operator could possibly deal with a table containing nulls unless it specified two sets of attributes, both the ones to result and the ones to be projected away. Either that or (I thought) some exception ought to be raised, not that I'm in favour of lots of exceptions. Received on Wed Oct 06 2010 - 16:18:34 CEST