SUO: Re: Formal SUMO Draft -- *Date 02 Feb 2002
¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤
AP = Adam Pease
JA = Jon Awbrey
MW = Merriam-Webster
JA: In my opinion it is a fundamental mistake to specify
a particular logical language, for example, KIF or
any other, as a part of the compliance conditions.
This would be as bad a practice as stipulating
that a compliant ontology has to be written in
English as opposed to French or German, or to
use an even more notorious analogy, that
a program has to be written in ADA.
JA: Requirements should be specified at a higher level
of abstraction and generality than any particular
ontology oriented logical formalism.
JA: Among the more deleterious side-effects of
thinking in only one language is a constant
tendency to confuse that language with reality.
AP: The compliance conditions are not intended to stipulate any particular
language as being required for a conforming ontology. I can see that the
wording might lead one to believe that so I'll suggest an alternate wording.
AP: Every term T in O appears in the SUO or it has a definition in O and the
definition is a well-formed formula of SUO-KIF (or when translated from
the language used for O into SUO-KIF) containing only terms which are
definable in the SUO.
SUO WG Members,
I can see that the plainest words that I know are bound
to meet with seeming incomprehension in certain quarters.
Adam's dissemblage of responses and trivial revisions do
not, of course, address the critical point that I raised.
On the off-chance that some of this might be remedied by
a point of information, I will instance the relevant gloss
from the brand of dictionary that we have here in steerage:
MW: specify. (vt)
1. to name or state explicitly or in detail.
2. to include as an item in a specification.
synonyms: see 'mention'.
MW: mention. (n)
1. the act or an instance of citing or calling
attention to someone or something especially
in a casual or incidental manner.
MW: mention. (vt)
1. to make mention of : refer to.
synonyms: mention, name, instance, specify.
shared meaning element: to refer to someone
or something in a clear unmistakable manner.
In short, those of us who were not born yesterday will recognize that citing
the name "SUO-KIF" as a constant parameter in the conformance specification,
instead of relegating it to a variable implementation parameter, will make
SUO-KIF a 'de jure' standard language, according it pride, prejudice, and
privilege of place above all other 'de facto' standards that may be found
to prevail from time to time, and often very transiently in the customs
and practices of a given application domain or theoretical community.
If we look back over the history of software engineering, I think that we can
all count very few cases where such a forced fixture in a spec did not become
an albatross, a millstone, a mortemain, and several other bad things that we
have probably called them as time swore on.
But, still, given the opportunity for reflection in which I now
luxuriate due to Jim's dedication to "openness and due process",
maybe I can almost see that another factor might be afoot here.
And so I will spend some time making a study of conformance clauses that
I have known in actual practice. I feel that I need to look at a number
of concrete examples, in order to avoid being swept away by a whirlwind
of abstractions. And I sense that I need to stick with cases whose real
practical consequences are familiar to me, because the gap between words
and deeds can often appear so deceptively trifling, at least, at first,
and until much later, when it's already too late. Also, I'm deliberately
choosing examples of conformance specifications that are far afield from
ontology standards, so as not to confuse the earthier issue with matters
of revelation, all the better to draw generalizations with a broad base
of experiential support, and in hopes that these comparisons will shed
some light on the question of specs before our eyes.
Example 1. Instruction to Authors. I once submitted a paper to a journal,
the toughest part of whose whole instruction to authors was this paragraph:
| Given the diverse nature of both the field and our readership, we ask that
| authors present early in their manuscripts the manner in which [Topic X]
| is to be understood. Manuscripts should be accompanied by 3 1/2 inch
| computer disk. Macintosh Microsoft Word is preferred, but other
| popular word processing applications are also acceptable,
| as are PC disks.
I can tell you, too, that the journal in question "specified" no particular
style sheet at this time, accepting any reasonable convention for references
and so on, and all manner of styles from APA, to MLA, to personal inventions
could be found in its pages.
Example 2. Thesis Requirements. Somewhere in every graduate handbook
that I have ever read there is a line about a thesis or a dissertation
being a "significant, original contribution to research in the field",
and then there will be a little red or green booklet that tells you
right on down to the page margins and the placement of the captions
what the format of the acceptable document will be.
Let me tell you what is interesting to me in all of this rigor-ma-role.
A formal specification is a lot like a program. I have known authoritative
sources who say that it is exactly like a program, but I would personally
like to leave a little room for casual neatness in my formalities of specs.
So it seems to me like we ought to be able to draw on our insights and
our training -- heaven help us if they ever meet! -- with programs, that
is to say, as some authorities do indeed say, with "effective descriptions" --
and here Bob Dole has forever ruined the acronym for us -- to think about
conformance specifications in the light of what we may have learned along
the way about formal spec methods, program correctness, provable apps/imps,
and "effective verifiability" (EV) in general. It could happen ...
The most generic but still informative descriptor that I can think of as
applying to an ontology is "theory". Theories and declarative programs
are a kindred lot. Therefore, let us contemplate a notion EV(S, T) of
an "effectively verifiable" theory T relative to a standard S, where
EV means "computable" or checkable by a computational program.
Is such a notion likely to be well-defined? If so, workable?
Exercise for the reader ...
Jon Awbrey
¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤