Re: Pascal/Date

On Jan 3, 7:11 pm, "s.j.lagoe" <s.j.la..._at_googlemail.com> wrote:
> Bruce C. Baker wrote:
> > In re Fabian Pascal: Does anyone have any info on the status of
> >http://www.dbdebunk.com/index.html? It hasn't been updated in a /long/
> > time! :-(I read this link and the site is excellent. It is sad that it has no
> more updating. I especially found the article called 'MUCH ADO ABOUT
> NOTHING' between Date and Codd and was interested in this below:
>
> 'Now suppose the system supports two kinds of null, "value unknown" and
> "value does not apply" and four-valued logic (4VL), and suppose
> employee Joe's job is unknown. What do we do about Joe's commission?
> It surely must be null--the information is surely missing--but we don't
> know whether that null should be "value unknown" or "value does not
> apply." Perhaps we need another kind of null, and five-valued logic
> ... This argument clearly goes on for ever, leading to an apparent
> requirement for an infinite number of kinds of null. '
>
> I understand why 4VL causes 5th null, but I do not understand why 'this
> argument clearly goes on forever'. What would 6th null represent? If
> someone can help explain why it is 'leading to an apparent requirement
> for an infinite number of kinds of null.' they have thanks from me.
> Sigmund.

My take: the problem here is that application-level issues are leaking down to the system level. Since application requirements may well "[go] on forever" then this may go on forever as well.

The better choice is to simply stick with 2VL (since that's necessary and
sufficient,) and use whatever normal mechanisms there are for defining new types and new values, and use *regular* logic on those. Rather than making up a new logic specific to your application's needs for undefined
or inapplicable or whatever.

Keep the levels distinct.

Marshall Received on Thu Jan 04 2007 - 06:03:14 CET