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

paul c wrote:
> Jon Heggland wrote:

>> I think this is oversimplifying. Constraints are associated with
>> variables, so we need the variables for something more than value
>> aliases.
>> ...

A constraint is a boolean-valued expression, yes. But expressed in terms of variables---otherwise it would be a proposition, and not all that useful.

I don't think we disagree all that much. But the concept of "variable" is very simple, and very familiar. I think your talk about language devices and relations having one value is confusing rather than clarifying. Consider this:

'Integer variables are nothing more a language device, for talking about integers, whether the language is Java or English. They merely serve to save us specifying an integer everytime we refer to it. The name of an integer variable stands for a given integer in a given conversation. All assignment does is allow one to replace one integer with another. A given integer has one value now and forever. Someone realized that such names could be used to stand for different integer values at different times, ie., the values of different integers that happen to have the same characteristics. I think it's okay to talk about mutations and such when one is talking about a programming language but there is no such notion when one is talking about integers. Maybe that someone opened the door to such myticism when he mentioned "time-varying integers". Unfortunate phrase, if you ask me. Words like "state" are akin.'

This is obviously a much too convoluted (and imprecise) way of explaining integer variables, and relation variables really aren't significantly more complicated than those.

-- 
Jon