Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!cs.utexas.edu!uunet!daitc!daitc.daitc.mil
From: jkrueger@daitc.daitc.mil (Jonathan Krueger)
Newsgroups: comp.databases
Subject: Re: Ingres and referential integrities
Keywords: Ingres referential integrities
Message-ID: <607@daitc.daitc.mil>
Date: 22 Jul 89 19:07:55 GMT
References: <3081@rti.UUCP> <9653@alice.UUCP>
Sender: jkrueger@daitc.daitc.mil
Reply-To: jkrueger@daitc.daitc.mil (Jonathan Krueger)
Distribution: comp
Organization: DTIC Special Projects Office (DTIC-SPO), Alexandria VA
Lines: 22
In-reply-to: debra@alice.UUCP (Paul De Bra)

>in the hierarchical databases referential integrity comes naturally, it
>is almost built into the model. With a relational database checking
>for referential integrity involves real (computational) work.

The reasoning is entirely valid but skips two relevant points: HOW
MUCH is the performance difference?  and how much is the performance
ADVANTAGE when you need to perform queries not expressible in
hierarchical terms?

So my hunch on INGRES's lack of referential integrity isn't
performance, but development complexity, and RTI's desire to implement
it in a more general way that includes derived columns, user-defined
rules, arbitrary grouping, and the like.  How do current commercial
relational DBMS define the integrity "no person can be his own
ancestor or descendent"?  Sorry, folks, that's not referential.  How
do current commercial hierarchical DBMS define it?  I'd be interested,
particularly in how concise and general the form is, and how data
independent and and separable from user interface.  Could somone post
how one states this in ADABASE, IMS, other?

-- Jon
--