Re: Declare before use constraint in ontology editor.
macgregor@ISI.EDU (Robert MacGregor)
X-Sender: macgreg@quark.isi.edu
Message-id: <ab8d23c11a0210049b4e@[128.9.208.151]>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Wed, 15 Mar 1995 15:42:09 -0800
To: rice@HPP.Stanford.EDU, m.uschold@ed.ac.uk
From: macgregor@ISI.EDU (Robert MacGregor)
Subject: Re: Declare before use constraint in ontology editor.
Cc: ontolingua@HPP.Stanford.EDU
><m.uschold@ed.ac.uk> writes:
>>>I disagree with the design decision to force classes to be created
>>>before referring to them.
I can't resist responding to this one.
I agree with possibly everything Jim Rice said in his response,
modulo, I'm not sure of what kinds of forward references are
allowed by Ontolingua, and which are not.
This reminds me of the flak we as Loom designers used to take
from Tom Gruber. In fact, many of Gruber's complaints about Loom
were reasonable, and triggered an attempt in the follow-on to
Loom to remove as many precedence restrictions as possible. However,
the restriction on defining parents before children in a class
hierarchy appears to be essentially necessary -- once you know
the parents (recursively) of a class or instance FOO, you can begin to do
useful inferences with FOO. Before that, you can't. One could
certainly imagine placing a whole layer above the KR system
that does no inference, but just keeps track of definitions and
assertions (i.e., its purely syntactic). This would add lots of
space overhead for what I perceive as relatively little gain.
I suggest that if M. Uschold wishes to have this capability, that
he add it himself. I doubt if he appreciates how hard it is to
build a logic-based KR system.
- Bob
Robert M. MacGregor macgregor@isi.edu
USC/ISI, 4676 Admiralty Way, Marina del Rey, CA 90292 (310) 822-1511