Re: 3vl 2vl and NULL

On Sat, 10 Dec 2005 00:05:31 +0100, Hugo Kornelis <hugo_at_pe_NO_rFact.in_SPAM_fo> wrote:

[snip]

>"Show me all family members who are older than Aunt Marge" - should
>Uncle Vernon be included or not? Why?

     No, because we can speak only of what we know. Think of the answer as starting with "As far as we know".

     As far as we know, the following family members are older than Aunt Marge: <list>

>"Show me all family members who are NOT older than Aunt Marge" - should
>Uncle Vernon be included or not? Why?

     No, because . . .

     As far as we know, the following family members are not older than Aunt Marge: <list>

     <list> would include Aunt Marge.

>"Show me all family members who are older than Uncle Vernon" - should
>this return noone, everyone, or only some? Why?

     No one. We can not tell based on what we know.

"Show me all family members who are NOT older than Uncle Vernon" - should this return noone, everyone, or only some? Why?

     This one is interesting. Necessarily Uncle Vernon is not older than Uncle Vernon, but this involves the inner meaning of the predicate. We assume that older-than makes sense w.r.t. to Uncle Vernon, but with another predicate, it may not make sense.

>In all these cases, you run into the problem that Uncle Vernon's age is
>not in the database. With or without databases, fact remains is that you
>don't know if Aunt Marge is older, younger or the same age as Uncle
>Vernon. This proves that, as soon as you allow values to be missing in
>your database, True and False are no longer the only possible results of
>a predicate. Reality forces 3VL on you.

     No, it does not. "There is no answer" is also a possible answer. That is at a different level than the 2VL.

     What is one divided by zero?

     Have you stopped beating your wife?

[snip]

Sincerely,

Gene Wirchenko Received on Mon Dec 12 2005 - 21:34:34 CET