SUO: RE: Conformance
Dear Frank,
Some comments below.
Matthew West
Principal Consultant
Shell Information Technology International Limited
Shell Centre, London SE1 7NA, United Kingdom
Tel: +44 20 7934 4490 Other Tel: +44 7796 336538
Email: matthew.r.west@is.shell.com
Internet: http://www.shell.com
> -----Original Message-----
> From: Frank Farance [mailto:frank@farance.com]
> Sent: 17 October 2001 20:22
> To: standard-upper-ontology@ieee.org
> Subject: SUO: Conformance
>
>
>
> At 10:28 2001-10-17 +0200, West, Matthew R SITI-ITPSIE wrote:
> >
> > Dear John,
> >
> > Well I think I am with you. In this discussion I have
> > been taking the position that the SUO is essentially a
> > posh dictionary, and that the only sensible conformance
> > statement you can make is of using the terms in it with
> > the meaning given in the "dictionary", rather than some
> > other meaning.
>
> Matthew West and John Velman-
>
> I agree with a good number of your points, including your
> impression that the SUO work is mostly a "posh dictionary"
> (in standards terms: a registry of axioms).
MW: If others agree with this too, that would be significant.
> This is why I've
> suggested two main discussion points: (1) what does
> conformance mean?, and
MW: I have answered this. If the SUO is a posh dictionary,
then conformance is using the terms with the meaning defined
by the dictionary (i.e. the axioms etc)
> (2) what criteria should we use for
> including items into SUO?
MW: a) It has a formal definition (KIF axioms) that provide an
adequate (weasel word) definition only in terms of other
SUO terms.
Note: Some terms will be really primitive, e.g. they are
defined by pointing (ostension).
b) Inclusion of the term has utility.
c) The same concept does not already exist with another name.
>
> Rather than initially suggest a registry, I just provided the
> background information on the "conformance" topic. In my
> experience, the discussion of conformance is always a
> contentious topic (and we are having a lively discussion in
> SUO). I can "live with" SUO being a registry. This leads to
> my next point ...
MW: Actually I think a registry has some advantages. We have
experience in EPISTLE of developing a registry for reference
data (mostly the same sort of concepts as will be in the SUO,
but possibly lacking the same formality of definition - we
have about 5000 terms at present).
>
> Since my hunch was that SUO might be a registry, I had
> suggested several times: we need to develop a criteria for
> what is included/excluded. From a technical point of view
> (not a P&P perspective), there are two main components to a
> registry: (1) the table of items, and (2) the criteria for
> additions, deletions, and changes. We should still develop criteria.
MW: We have an administrative procedure and a technical
procedure. They can be found as parts 4 & 5 of ISO15926, and are
probably available at:
http://www.iso15926.org.uk
>
> This would be a good experiment: if we develop criteria and
> view the SUO work as a registry, is this good enough?
MW: It would be a start.
>
> > ...
> > > -----Original Message-----
> > > From: John.Velman@HSC.com [mailto:John.Velman@HSC.com]
> > > Sent: 16 October 2001 20:27
> > > To: standard-upper-ontology@ieee.org
> > > Subject: RE: SUO: Re: Conformance
> > >
> > >
> > >
> > >
> > > All,
> > >
> > > I haven't been able to keep up with all the correspondence on
> > > conformance,
> > > so if these remarks parallel someone else's please accept
> my apology.
> > > Except for a note from Frank (IIRC) mentioning that the
> > > conformance had to
> > > do with an implementation of the standard, not with the
> > > standard itself,
> > > I've seen nothing that pins down just what is being discussed
> > > (although
> > > much of the discussion has been interesting, nevertheless).
> > >
> > > <rant>
> > >
> > > Referring to the PAR, the scope states that this standard
> > > will "specify an
> > > upper ontology..." It goes on to say that an "ontology is
> > > similar to a
> > > dictionary or glossary, but with greater detail and
> > > structure..." The scope
> > > does not delineate any other application or
> implementation that the
> > > standard applies to.
>
> Agree. And the PAR is intentionally minimalist because we
> must do everything that is mentioned in the Scope of the PAR.
> A common mistake in IEEE PARs (but not other standards'
> project statements) is to overcommit on the project: in the
> early stages, everyone is hopeful and their perspectives are
> broad. However, later on as the standard is developed and
> refined, the WG has a much clearer and precise perspecitve of
> what is possible (to standardize) and what they want. In
> IEEE, they are particularly firm about the PAR Scope matching
> the final product, so the best advice is to be minimalist
> (rather than ambitious) in PAR Scope wording.
>
> The current PAR Scope wording wouldn't prevent us from
> understanding (and standardizing) SUO in related usages
> (e.g., as described in the Purpose wording) <-- it's all
> dependent upon what the SUO WG wants.
>
> > > The scope of the PAR sounds much like the scope for a
> > > terminology standard.
> > > What is the implementation of a dictionary---the
> dictionary itself?
> > >
> > > While the parallel isn't exact, it seems to me that it is
> > > pretty close. As
> > > an example, ISO1087 Terminology---Vocabulary, is a standard that
> > > "establishes a basic vocabulary for terminology science and
> > > its practical
> > > applications." ISO1087 has no conformance clause. It has
> a scoping
> > > statement that it is applicable to terminological activities.
> > >
> > > The Purpose clause of the PAR does state that
> "Applications based on
> > > domain-specific ontologies that are compliant with this
> > > standard will be
> > > able to interoperate (to some degree) by virtue of the shared
> > > common terms
> > > and definitions." Compliance and conformance are often used
> > > as synonyms,
> > > more or less, so this seems to imply that we are thinking
> > > about a standard
> > > that will specify requirements on domain-specific ontologies.
> > >
> > > Seems to me that this is not well stated, and I'm sorry I
> > > missed it at the
> > > time. It seems to me that the meaning of this statement
> > > could have better
> > > been expressed by: "Applications based on domain-specific
> > > ontologies that
> > > use terms and definitions from this standard will be able to
> > > achieve some
> > > degree of interoperability."
> > >
> > > Perhaps I'm befuddled by this partly because I have been
> > > dwelling in a part
> > > of the standards world where conformity means strictly (ISO
> > > IEC Guide 2,
> > > Clause 12.1) "fulfillment by a product, process or service of
> > > specified
> > > requirements."
>
> John-
>
> Another excerpt in ISO/IEC Guide 2 might be helpful, too:
>
> 5 Types of standards
>
> NOTE - The following terms and definitions are not in
> tended to provide a systematic classification or
> comprehensive list of possible types of standards. They
> indicate some common types only. These are not mutually
> exclusive; for instance, a particular product standard may
> also be regarded as a testing standard if it provides test
> methods for characteristics of the product in question.
>
> 5.1 basic standard
>
> standard that has a wide-ranging coverage or contains
> general provisions for one particular field
>
> NOTE - A basic standard may function as a standard
> for direct application or as a basis for other standards.
>
> 5.2 terminology standard
>
> standard that is concerned with terms, usually
> accompanied by their definitions, and sometimes by
> explanatory notes, illustrations, examples, etc.
>
> So the SUO work might be some combination of a "basic
> standard" and a "terminology standard".
>
> Regarding the "implementation" (use) of SUO, the following
> ISO/IEC Guide 2 excerpts are relevant:
>
> 10 Implementation of normative documents
>
> NOTE - A normative document can be said to be
> "implemented" in two different ways. It may be applied in
> production, trade, etc., and it may be taken over, wholly or
> in part, in another normative document. Through the medium of
> this second document, it may then be applied, or it may again
> be taken over in yet another normative document.
>
> 10.2 application of a normative document
>
> use of a normative document in production, trade, etc.
>
> 10.2.1 direct application of an international standard
>
> application of an international standard irrespective
> of the taking over of that international standard in any
> other normative document
>
> 10.2.2 indirect application of an international standard
>
> application of an international standard through the
> medium of another normative document in which it has been taken over
>
> So "implementation" can mean direct use (someone/something
> uses SUO directly) or indirect use (someone/something uses
> another standard that includes SUO as a normative reference,
> e.g., a domain-specific ontology that uses SUO). A
> terminology standard may be "implemented" via incorporation
> by normative reference.
>
> (See my comments below marked with "[*]" for more discussion
> on "conformance".)
>
> > > Putting the horse on the other end of the cart, if we really
> > > want to have
> > > conformance, we need a provision in the scope of the standard
> > > specifying
> > > the particular products, services, or processes that are
> > > intended to be
> > > covered ---secondary ontologies, applications using
> > > ontologies?---and a
> > > clause specifying the requirements on them.
>
> As I said above, the PAR is minimalist ... if there were
> agreement on particular types of usage (see my dicussion of
> "conformance paradigms" in the 2001-10-03 E-mail titled "Some
> thoughts on conformance"), then we might be including them in
> our standards wording, but we might have not had the
> foresight to understand the right conformance paradigms when
> the PAR was written <-- this is normal in the standards process.
>
> My purposes for asking the questions on "conformance" and
> "criteria" were to pin down (as you say) some of the
> important aspects of the standard that we are developing.
>
> > > The discussion has been about conformance of other ontologies
> > > to the SUO.
> > > This still strikes me as a strange concept. However, to
> > > accomplish this,
> > > perhaps the title of the standard should be: "Standard Upper
> > > Ontology and
> > > Requirements for Lower Ontologies." Assuming we eventually
> > > use a modular
> > > form (which John Sowa and others recommend, and I endorse),
> > > this framework
> > > could become very interesting.
>
> From a standards perspective, your suggestion might be
> accomplished as a multi-part standard. For example, 1600.1
> is the registry, 1600.2 is the registration authority process
> (i.e., criteria and change control P&P), 1600.3 is the
> "{Standard | Recommended Practice | Guideline} for Lower
> Ontologies". A good reason for keeping them separate is that
> users might what to refer to them separately, e.g., I might
> only need 1600.1, but not need 1600.3 because I'm not
> developing a lower ontology.
>
> > > Stylistically the right way to do this will be for the Scope
> > > to delineate
> > > the products, services, or processes covered, followed by a
> > > clause stating
> > > the requirements on them. The conformance statement might
> > > then be like
> > > that in the scope of IEEE1320.1 (IDEF0) which says "A
> model or diagram
> > > conforms to this standard if it adheres to all mandatory
> > > provisions (marked
> > > by the verbs 'shall' or 'is') of this standard."
>
> A Side Note: While the conformance statement of IDEF0 might
> seem simple, I'd guess that many smart people worked many
> many hours to "get it right".
>
> > > This strikes me as a more complicated and somewhat different
> > > job than just
> > > creating the upper ontology itself, which seems hard enough.
> > >
> > > On the other hand, it would be useful to have a
> clause---possibly an
> > > informative annex, rather than a normative
> clause---describing how the
> > > Upper Ontology can be used in support of a domain
> specific ontology.
>
> [*]
> If SUO is only a "terminology standard" (as per ISO/IEC Guide
> 2), then I agree. If SUO is a "basic standard" (or "basic
> standard" combined with other types), then conformity
> (conformance clause) should be developed.
>
> I don't think it is fruitful (yet) to spend time trying to
> categorize the SUO work in any of the ISO/IEC Guide 2
> standards categories ... the discussion will go on endlessly
> without conclusion. However, I do believe it is productive
> to talk about "what conformance means" (and to develop a
> "criteria" document). The "conformance" discussion is useful
> because we are discussing how SUO will be used in the context
> of the standard. Once we have a better understanding of what
> we mean by conformance, then the category of the standard
> (basic, terminology, etc., combined) is likely to be obvious.
>
> Yes, this might appear to be "putting the cart before the
> horse", but this problem solving technique gets us to
> solutions quicker (i.e., investigating the concept of
> "conformance" early on speeds the standards development process).
>
> -FF
> --------------------------------------------------------------
> ---------
> Frank Farance, Farance Inc. T: +1 212 486 4700 F: +1
> 212 759 1605
> mailto:frank@farance.com http://farance.com
> Standards, products, services for the Global Information
> Infrastructure
>