Re: A new proof of the superiority of set oriented approaches: numerical/time serie linear interpolation

"Cimode" <cimode_at_hotmail.com> wrote in message news:1178389295.897789.225590_at_l77g2000hsb.googlegroups.com... [snip]
> I agree undred percent. On a purely logical standpoint the risk is
> indeed very high. The reason I flirt with such dangerousity is
> identify some key issues that I may integrate for further refining for
> building a computing model that supports a systematic treatment of
> missing information *without* projecting logical decomposition
> (described by Darwen). I have reasons to believe that projecting such
> method on a physical standpoint is a dead end because it would break
> the primal principle of separation of the two layers. So I am
> reviewing a few computing methods that would allow systematization of
> treatment by the dbms. Among these methods interpolation may be
> applied in the specific context of numerical and datetime
> implementation of mathematical series. My hope is that if I can avoid
> enough traps, that may prove helpful for the refining of the computing
> model.
>

I'm liking the idea of the "active default" constraint that I mentioned before. Since a domain can be anything at all, why couldn't it be a set of scalar expressions, sort of like the formulas in cells in a spreadsheet? A scalar value is a degenerate scalar expression. Expressions have a type, yes? Of course this might be stretching the limits of what a possible representation is....

Short of that, I don't think you can avoid both nulls and logical decomposition. Even with a derived relation, there has to be something from which it is derived.

[snip] Received on Sat May 05 2007 - 21:31:40 CEST