Re: Dreaming About Redesigning SQL

Jan Hidders <jan.hidders_at_pandora.be.REMOVE.THIS> wrote in message news:<3f851baa.0_at_news.ruca.ua.ac.be>...
> Paul G. Brown wrote:
> >
> > I think I grok the principles supporting 'abstract' types. I've not
> > read the Tarski but in my earnest youth I railed against REF/DEREF for
> > all of the 'right reasons' until it was pointed out to me that databases
> > routinely store filenames and then reason about the filenames perfectly
> > happily thank you very much and filenames are just references so would
> > I please just shut up, OK?
>
> Er, well, file names are actually not abstract because they have a
> concrete representation that you can read.

  We could go round and round on this. I'd observe that there is a whole   class of things which can potentially go into tuple attributes that aren't   'values': for example, you might put a query in there, or a reference, or a   filename, or an executable script to generate a value. The information   content of such values are always 'one step removed' from their   representation, and may in fact be ambiguous.

  The Information Principle says that data is stored as values in tuple   attributes. All of the above examples violate this principle. Filenames   then, are 'bad' for the same reasons REF/DEREF is 'bad'. But few people   would say that you shouldn't plonk filenames in attributes.

    KR

   P "still don't like REF/DEREF, but can live with queries in attributes" b Received on Thu Oct 09 2003 - 23:47:21 CEST