Re: Examples of SQL anomalies?

From: Cimode <cimode_at_hotmail.com>
Date: Sat, 5 Jul 2008 05:38:14 -0700 (PDT)
Message-ID: <981bd264-4807-4c8c-862e-78026ed0139f_at_y38g2000hsy.googlegroups.com>


On 5 juil, 04:26, paul c <toledoby..._at_ac.ooyah> wrote:
> To explain a little more, my point of view is that of somebody who is
> interested in the construction of systems. As much as I admire the
> efforts of D&D, for their precision, diligence and faithfulness to their
> logical principles, I've always been sceptical, for reasons I can't
> explain very well, of the practical possibilities for combining a
> language that operates on relations with one that allows type definition
> by users.
>
> I realize that unlike me, not everybody here has the RM in mind when
> they talk about values. But as far as the RM is concerned, I think it
> is only concerned with determining equality or inequality of values.
> From a practical point of view, I would prefer an engine that uses a
> relational language to operate on relations and some kind of type
> language (not one of the current OO languages) to define types. I would
> hope this would be implemented 'outboard' of a relational engine (or
> layer if you like) and use some form of early or late bind to provide
> the 'equal' or 'not equal' answers the RM needs. Am sure this must be
> extremely controversial to many who know the RM better than I do, as a
> change in software could change the facts, eg., the answers a given db
> gives!
Hi paul

Mmm..I see my comments introduced some confusion. Apologies fo that. I should have written *next best thing is extrapolation*...There is no doubt in my mind that extrapolation is a smart hack but one that allows at least to establish limited relevance of the output. The alternative is chaos. Compared to flagging the result as *suspect* it is limiting in some cases the absurdity of the error, extrapolation in temporal based relations can be beneficial.

I understand your point and fully agree with your concern. Missing information ought to be dealt with relation decomposition on RM standpoint or any subsequent implementation. The db engine I am working does it systematically by giving the programmer/designer the choice of column subset on which the sum would be accurate then have the system return the corresponding correct result.

Regards Received on Sat Jul 05 2008 - 14:38:14 CEST

Original text of this message