Re: Storing data and code in a Db with LISP-like interface

Frank Hamersley wrote:
> Alvin Ryder wrote:
> > Marshall Spight wrote:
> >> Alvin Ryder wrote:
> >>> Marshall Spight wrote:
>
> [..]
>
> >> The RM is a practical application of set theory. Is set theory
> >> good for some kinds of data but not others? Set theory
> >> is foundational.
>
> [..]
>
> >> What kinds of data can't you put in sets?
> >
> > Sure you can put any kind of data into a set but if you know that that
> > data is temporal or spatial you can provide further intelligence and
> > facilitate that data better.
> >
> > Just because the RM is based on set theory (amongst other things) it
> > doesn't mean it has to stop there. It doesn't have to mean "RM" equals
> > "set theory".
>
> Some (bigots) in CDT hold if the requirement can't be represented in set
> theory then it isn't or shouldn't be considered part of the RM. I don't
> subscribe to this view although I admit the concept has some appeal in
> terms of keeping the RM simple and compelling.
>

The relational model can represent any data but what seems to be shocking to some is the notion of building on top of it. CJ Date says it best "a kind of building on top of the RM; it doesn't change the model, but it does make it more directly useful for certain tasks".

> However considering the temporal issue you mentioned I am wondering
> where time fits in set theory. I admit I am not well read on the
> mathematical stuff and would be interested to see what better informed
> types have to say on the subject.
>

Sets can contain time values but the question is do we extend the relational algebra to support temporal issues better?

For further reading you might like to check out the book "Temporal Data & the Relational Model" by CJ Date, H Darwen, H Lorentzos.

To be honest I haven't read that particular book but I've worked with Lorentzos' original papers and now Date is involved so I'm sure it'll be ok.

> > Its not only a question of can you merely put any data into a set, its
> > a question of is the base RM *good* at handling temporal, spatial,
> > dedictive, oo, multimedia, unstructured and document library type data.
>
> Regardless of the prior points IMO the measure of *good* is a merely the
> capacity of the implementor(s) to handle that data in a way that the
> user(s) find responsive to their requirements. I don't think the RM
> would need to be changed to accommodate any of these other aspects.
>

Various queries in the spatiotemporal realms become much cleaner and nicer after you add certain operators to the relational algebra. You can then end up with a spatial-join for example.

> > For certain advancements changes need to occur at the RM level,
> > relational algebra and if necessary even at the set theory level (I
> > doubt that it is static).
>
> FWICT Set theory is pretty solid where it is right now.
>

I guess you're right but when my office was near the maths department I used to over hear lots of strange muttering, at any rate you can derive additional operators from more primitive ones so you can easily extend it in that sense.

> > Codd and others have not been idle in this pursuit, though of course
> > it'll be decades before it all becomes common knowledge.
>
> Do you mean the history or stuff under development or both?
>
> Cheers, Frank.

No worres Frank, thanks for the discussion.

Cheers. Received on Wed May 03 2006 - 12:10:12 CEST