Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!ames!pasteur!eden!mao
From: mao@eden.uucp (Mike Olson </>)
Newsgroups: comp.databases
Subject: Re: Oracle: Previous Record
Summary: non-relational extensions to sql are all the rage
Message-ID: <15524@pasteur.Berkeley.EDU>
Date: 17 Jul 89 18:22:57 GMT
References: <3900001@tdpvax> <20532@cup.portal.com>
Sender: news@pasteur.Berkeley.EDU
Reply-To: mao@postgres.Berkeley.EDU (Mike Olson)
Organization: Postgres Research Group, UC Berkeley
Lines: 30

in article <20532@cup.portal.com>, Bron@cup.portal.com (G.B. Faison) writes
>In an article written 7/7/89 17:48 miker@tdpvax.UUCP writes:
>
>> I am using Oracle on SCO386 Xenix and I wish to be able to be
>> able to retrieve a previous record.
>
> ... you should think *sets* not record-at-a-time!
> Previous record has no valid meaning in a truly relational system
> where by definition physical ordering is irrelevant.

a valid point, but virtually all commercial implementations of SQL provide
pretty hefty non-relational extensions.  in a real relational system, for
example, duplicate tuples are forbidden.  it turns out, though, that the
pentagon and the phone company care more about solving their own problems
than preserving the semantics of relational calculus and set theory.  the
database systems with which i'm familiar at least offer the option of
having duplicate tuples in a single relation (or rows in tables, if you
swing that way).

the ansi sql committee clearly understands the demands of the marketplace.
the appearance of cursors in the new standard is evidence of that.  it
may not be relational, and you may not even think it's pretty, but it's
what people want.

in general, when a big customer asks for a procedural or non-relational
extension to a vendor's SQL, it'll get put in.  marketing departments are
notoriously weak on theory.
					mike olson
					postgres group
					uc berkeley