Re: Modeling question...

On Nov 13, 9:39 am, paul c <toledobythe..._at_oohay.ac> wrote:
> Roy Hann wrote:
> > paul c wrote:
>
> >> paul c wrote:
>
> >>> Aren't all identifiers artificial? ...
> >> Ie., we make them up.
>
> > You need to be more precise. All identifiers are made up somewhere, but
> > not necessarily within the enterprise of interest. "We" might not
> > need to make up any. Or we might need to make up a few--ideally the
> > fewest sufficient.
>
> Sure. But as for "not necessarily within the enterprise of interest",
> why does such a distinction matter?

ISTM it relates to whether we informally consider a UoD and various external predicates to exist /a priori/ (ie independently of the DB) - even though we regard the UoD and the external predicates as outside our mathematical formalism.

If we need to name some things in order to state facts about them then I don't think it's particularly useful to this theory group to try to distinguish between "natural" and "artificial" names. I believe this is even true if we happen to use the DB to help allocate names for informal things that we actually consider to be in the UoD.

I think an /abstract identifier/ should be defined as an identifier that can be regarded as a name of a variable (which holds an abstract value) within some context within the DB (and not in the UoD). When I say context I mean that there is some defined scope (hopefully as small as possible) in which the name is meaningful. For example within the context of representing a tri-surface (ie triangulated irregular network) /value/ it may be useful to introduce a scope in which abstract identifiers are names of vertex values. Note that whenever we have a binding from an identifier to a value within some scope I would say by definition we have a /variable/.

A tuple that contains an attribute value than is an identifier outside the UoD cannot represent a self-contained (ie independently verifiable) fact on the UoD. This is why I think one should introduce as few abstract identifiers as possible. The idea that a domain expert can regard each tuple of a relation as a self contained fact is extremely valuable. Received on Thu Nov 13 2008 - 03:49:54 CET