Path: utzoo!mnetor!uunet!husc6!cmcl2!nrl-cmf!ames!pasteur!ucbvax!ucsfcgl!cca.ucsf.edu!daedalus!brianc
From: brianc@daedalus (Brian Colfer)
Newsgroups: comp.databases
Subject: Re: Informix 4GL Question?
Message-ID: <1180@ucsfcca.ucsf.edu>
Date: 9 Mar 88 20:39:55 GMT
References: <714@uel.uel.co.uk> <2314@geac.UUCP> <1169@ucsfcca.ucsf.edu> <2390@geac.UUCP>
Sender: root@cca.ucsf.edu
Reply-To: brianc@daedalus.UUCP (Brian Colfer)
Organization: UCSF Dept. of Lab Med
Lines: 58
Keywords: Informix 4GL, SQL, Relational theory

In article <2390@geac.UUCP> daveb@geac.UUCP (David Collier-Brown) writes:
>
>  That's what I was hoping to discover.  Relational theory excludes
>significant consideration of certain uses of the data which it
>describes.  And that is a prefectly reasonable thing for it to do:
>it is concerned with the nature and organization of data, not ad-hoc
>manipulations such as the ones I raised.

  See my followup with CURSORS.

>
>  This raises a semantic-gap problem, though, in that selected
>applications need operations which are not well-defined within the
>constraints of the theory.  

   I see a distinction between the issues of a good user interface and 
   a good database model.  I think that the relational model is clearly
   a good model.  I don't see this as advocating kludges rather it is 
   building on sound theory and research with a modular approach.

>>                                                             Plus from a
>>application perspective to arbitrarialy select a set of row(s) because of 
>>its position is illogical. 
>
>  Quite the contrary!
>
>  In this case, one is doing bibliographic query on a very large
   ...
>  Human inspection of an arbitrary subset, *traditionally*
>the first N encountered, is required to discover if the data
>requested is the data desired.  Response is required to take less
>than a minute.

CURSORS would accomadate this. But even here I think its better for the
user to be arbitrary rather than the relational model.  For example 
get a count and then select the set with the min. date  and allow for 
an interupt if its still too long. 

>
>  Without a theory which accepts such illogical (:-) constraints as
>human, processing, communication and display limitations, one has
>two choices:
>	1) kludge it, or

CURSORS should do it.

>	2) improve the theory.

WHY? I know that I all I can come up with are impressions of possible
extensions.  The foundation of the relational model is sound.   It is
based on predicate calculus which has a long and established history.


==========================================================================
             : UC San Francisco            : brianc@daedalus.ucsf.edu 
Brian Colfer : Dept. of Lab. Medicine      : ...!ucbvax!cgl!brianc 
             :  PH. 415-476-2325           : brianc@ucsfcca.bitnet
==========================================================================