Frames are (almost) enough for EDIfritz@rodin.wustl.edu (Fritz Lehmann)
Date: Tue, 20 Sep 94 06:44:43 CDT
From: email@example.com (Fritz Lehmann)
To: firstname.lastname@example.org, email@example.com, firstname.lastname@example.org
Subject: Frames are (almost) enough for EDI
Cc: email@example.com, firstname.lastname@example.org
Michael Laing wrote on the list email@example.com:
I think an interesting area to explore is the construction of protocols whereby
'frames' can be interchanged. A related question is: are 'frames' sufficiently
general - this may perhaps be answered (approximately!) in a practical sense by
looking at well-defined & useful information objects like those in legacy EDI,
OpenDoc, RFC822 mail (easy), POP, SMTP, etc., and modelling them in 'frames'.
Furthermore, can such protocols: preserve the generality of 'frames' and be
syntactically expressed in ways that are: easy to process and 'extensions'
(loosely) of existing practice (to promote acceptance).
And would they be useful?
I don't know all of those formats but I strongly
suspect that they all can be accommodated in a frame-based
semantic network system. The main thing lacking in basic frames
is the proper handling of negation and disjunction; these are
handled correctly in John F. Sowa's "Conceptual Graphs",
an order-sorted (i.e. with a type hierarchy) frame-based semantic
network system with negated "contexts". See his book "Conceptual
Structures". The standard form of Conceptual Graphs has (at least) the
full expressive power of typed First-Order-Logic. So you could define
for EDIFACT "A Purchase Order of type-X is one without terms of delivery
(segment group 11) if it has either fixed transport means
(segment group 9) or if the delivery schedules are specified
(segment group 15)." A formal conceptual dictionary-definition
of an element, segment, or transaction-type would specify a substitution
of a large, descriptive conceptual graph for a small one.
The ISO/IEC JTC1/SC21/WG3/CSMF Interim Rapporteur Group just met in
Seattle to complete the basic "CSMF" or Conceptual Schema Modelling Facility
standards proposal (essentially the old IRDS normative language). This
includes Conceptual Graphs (and a logic language called "KIF") as a concept-
definition language. This CSMF would be an appropriate language in which to
commence a programme of formal conceptual definitions of EDI objects.
The main obstacle is not so much finding a suitable logical language;
rather it is getting the right underlying "real-world ontology" needed to
properly define the business practices involved in EDI. This requires not
only sophisticated models of trade and enterprises, but also some of the
applicable commercial law, as well as real-world basics such as time,
space, events, measurement, people, obligation, etc. I'm working on the
"CCAT" ontology repository of the "PEIRCE Project", a voluntary
collaborative worldwide effort to build a testbed for Conceptual Graphs.
CCAT refers to the "Conceptual Catalogue/Ontologies" interest group. We
hope to put together a basic "core" of useful ontologies that can be used
for free "off-the-shelf". The concepts underlying ANSI X12 and EDIFACT
EDI would be good candidate target-concepts to test the usefulness of
these basic ontologies in formal definitions of real-world activity.
Another good candidate for target-concepts is the PDES/STEP standards for
product description. (STEP and EDI are quite complimentary I think, or at
least they should be.)
EDI would consist of two phases: first a negotiated exchange of
ontologies and data-type-definitions to get both partners in synch, then
interchange of transaction "frames" as Laing described.
If there is any interest I can send an email summary of CCAT to the
edi-new list. If you think you might want to use, or contribute to, CCAT
ontologies, let me know.
Yours truly, Fritz Lehmann
GRANDAI Software, 4282 Sandburg Way, Irvine, CA 92715, U.S.A.
Tel:(714)-733-0566 Fax:(714)-733-0506 firstname.lastname@example.org