Re: Database design, Keys and some other things

JOG wrote:
> I think it comes down to the fact, it is easy to confuse a predicate
> with the situation it describes, but there is a real difference.
> Consider the predicate:
>
> "The sky is blue in the daytime"
>
> This has may be represented a a set of three items {sky, blue,
> daytime}. But we can say some extra things about this set. First who
> stated it - me. Second when it was stated. Third that it is the first
> thing I have said on the matter. Fourth, we can comment on the truth of
> the statement - that it, in england, is only true.. say 25% of the
> time. tops.
>
> Now these are all pieces of information about the PREDICATE itself, not
> the situation it describes. They are metadata, and encoding them with
> the information about the situation is entirely incorrect - they are
> attributes of the container, not of the content.
>
> Mathematically we might write:
>
> P = {"sky", "blue", "daytime"}
> Author = {(P, "James")}
> Created = {(P, "27th Sept 2005")}
> etc...
>
> Saying { sky, blue, daytime, James, "27th Sept 2005", 25% } is wholly
> wrong. It is a confusion of data and meta data.
>
> A surrogate key, created specifically to represent a statement, is no
> different. It is an artificial way of referencing the predicate. This
> is meta_data about the predicate and has no place existing in the same
> area as the elements of the predicate - the real information, that
> exists in the real world.
>

philosophically, i'm likely in over my head but this issue has interested me for some years as i think there are important practical consequences to it. to me, as far as db or maybe i should say the RM rather than db as that is my interest, whether some value is metadata seems relative to the way it is used. when all is said and done, whether it is or isn't is in the hands of the beholder.

not to be flip, "{ sky, blue, daytime, James, "27th Sept 2005", 25% }" *does* exist in the real world, because i saw it in JOG's post! i see no problem if some user understands the two rightmost values as metadata, but the db needn't be so precise and can still be useful and precise in its own realm.

talking only as far as the RM is concerned, i believe that all metaphysical "meaning" is intended to be external to the db's operation. (although i have heard there was some controversy when Dr. Codd seemed to abandon this position in his 1979 paper). i think this would mean that so as far as as the RM is concerned, the quote could equally mean "the daytime sky above James was 25% blue on Sept 27th" as well as lots of other things such as "James gets it right 25% of the time", or even "the sky *wasn't* blue, etc.". from the RM's perspective, or at least the earlier descriptions of the RM, none of that matters, only the RM's internal logic in manipulating the values involved.

you may well be speaking of things which go beyond the RM and i mean no criticism of that. just thought i'd give my practical db interpretation of what you say.

maybe i'm saying the same thing as x from not-exists.org when he or she said "Then, after the translation of this model to the relational model, the date become the attribute of some relation (not of some entity) and the problem "which is the entity of this attribute ?" is solved.". to me, this is part of the potential for db - to be able to transfigure such a predicate and combine it with others in some internally consistent way *without* getting tied up knots about the meaning of reality!

cheers,
paul c. Received on Tue Sep 27 2005 - 20:41:48 CEST