Re: Columns without names

Marshall wrote:
> paul c wrote:
> > JOG wrote:
> > >
> > > I find it apt to view a relation predicate for them as:
> > > "There is a people_relationship where name is X and age is Y"
> >
> > Wouldn't a more
> > likely predicate be "person X has age Y"?
>
> Although I distinctly prefer Paul's formulation, I'm not sure
> the difference is anything more than better phrasing.
>
> Well, on second thought, I've changed my mind. JOG seems
> to be including the relation name in the predicate, which
> I would consider a mistake.

Maybe, maybe not - there is some method behind the madness of doing so (but I am of course not naive enough to think I can't be persuaded otherwise), based on something that's bugged me for a while.

Given the intension of a set of tuples is defined by a structural predicate (in combination with any constraints) there may be a set-theoretical issue with exactly /how/ this intension manages to collect statements from different contexts, but which use identical attributes.

Imagine, f'rinstance, I have a data collection in which I have propositions of dogs' names & ages and owners' names & ages. Both sets of propositions would traditionally be represented as (name:x, age:y), but yet are housed in separate relations to form my manipulable datatbase. How exactly has this been determined by the set intension? It feels unsatisfactory mathematically to say 'I just plonked it in there' and mix extensional and intensional definitions. It would be theoretically pleasing to have a formal method based merely on the set's intension.

I have a hunch that there is extra situated information concerning a proposition that /we know/ intuitively because of our domain knowledge - how the data was collected, recognition of name conventions, whatever - we know what relation it should be in, but are not explictly stating it, nor encoding it in the intension of a relation set.

>
>
> Marshall
Received on Tue Sep 19 2006 - 01:57:11 CEST