Re: RM formalism supporting partial information
Date: Fri, 16 Nov 2007 18:17:31 GMT
Message-ID: <%8l%i.1667$cD.831_at_pd7urf2no>
Bob Badour wrote:
> Alfredo Novoa wrote:
>
>> On 16 nov, 18:06, Bob Badour <bbad..._at_pei.sympatico.ca> wrote: >> >>>> I agree. I propose to ditch the term "attribute" and to use "heading >>>> element". >>> >>> I wouldn't even bother with that. In a predicate or proposition, the >>> name is a variable name--either free or bound. In a tuple, the name is >>> also a variable name or the name of a dimension or um a domain name. >> >> In a tuple I would call "tuple elements" to the pairs formed by name >> and value.
>
> I am not sure the name is necessarily part of the element other than to
> identify it. I am not sure if that is clear enough. In mathematics, we
> sometimes use names and we sometimes use position to identify the
> elements of tuples.
>
> I should also say, while I was originally drawn to the algebra, I have
> come to realise the calculus is much more illuminating. In the calculus,
> the name is a variable name -- either free or bound depending on the
> context.
No argument and it would be nice to be able to talk only of variables. But I think people who are interested in the construction of a dbms (as opposed to the construction of an application) are likely to find an algebra to be a more opportune way of implementing that, relying on some equivalence of a calculus and algebra (even if most dbms designers probably couldn't prove an equivalence). If you accept that, then a tuple seems an obvious intermediate device for making a dbms, albeit an "under-the-covers" device, never to be exposed to the app designer.
If Codd or Ullman or whoever hadn't shown that equivalence, then I wonder where we'd be today (although some wag might say "exactly the same place" because it's not clear that SQL follows any coherent algebra!). Received on Fri Nov 16 2007 - 19:17:31 CET