SUO: SUO Votes on SUMO and IFF
Jim,
A vacation and other prior commitments take me offline from this list for
much of the period through October 6. I would like to register my votes on
both matters before us and my basis for votes in the future.
I vote NO on SUMO, YES on IFF.
I gather there may also be as many as four more candidate documents. The
make-or-break issue for me for all submissions, as flagged in the two
previous votes, is whether the main author(s) of an SUO candidate document
understand that the goal can only be the production of a reference
ontology. If so, they should be able to provide preliminary specs on the
likely structure and limits of their submission, at least to the extent of
specifying what features are so basic as to be non-negotiable. I don't
think it is necessary to "agree" with any artifact at this stage in order
to accept it as a candidate document. IFF describes itself as a reference
document, provides preliminary specs, and lists which of these are
non-negotiable. SUMO provides some preliminary specs but has for over a
year consistently resisted the methodological burdens of presenting itself
as a reference object. I put two direct questions to the proposers retest
these matters. IFF had answers, SUMO did not. Hence my votes.
A few further remarks:
1. On SUMO: SUMO is funded, which is a positive sign. This gives it the
resilience to reconfigure and resubmit. Several of us have gone down the
merger path before. It can be a useful winnowing and integration device,
but not self-sufficient. With the SUMO group, I agree that there is no
escape at the "semantic" level from the exercise of human judgment in
resolving the incompatibilities that inevitably crop up in any merger
approach ("bow ties" in which an inheritance relation in one ontology is
inverted in another to be merged with it). But if the result is to gain
acceptance as a de facto standard rather than being issued as a "black
box" product, the grounds of judgment need to be well articulated. If this
is done other products can be aligned (not merged) with it, tools can be
created or reconfigured for it, adaptations can be made by distributed
users.
2. On IFF: My positive vote on IFF is based on issues generic to any
submission and not specific to its own distinctive structure. I have
quite a few questions specific to IFF, which presents a content-neutral,
meta-level apparatus designed to permit transitivity of inference from an
SUO to domain ontologies. Grounded in category theory it appears to permit
a looser kind of inference (derived from embedding) than first order
logic. I would like to explore the practical sphere of its application,
the scalability and interoperability of its approach, and of course also,
whether my understanding of it is correct. It is grounded in three person
years of prior methodological work.
3. On procedure: As a member of the P&P Committee I should like to
register my regret that we did not reach consensus on basic procedures
before the chair took the step of submitting resolutions to the larger
group.
4. On sustainability: As in the previous straw vote, I think it important
to look behind the numerical voting and to listen to what is being said
and what human resources are being committed. If we are indeed to have an
array of candidate documents rather than a single starter, a reasonable
test of sustainability for each is whether as many as four members not
financially coordinated with the submitter(s) are prepared to participate
in reviewing and discussing it and, in the mind's eye, maintaining an
artifact that might emerge from it. Five unrelated members is the number
that is often required at the end of a standards' process. As anyone can
participate in work on more than one document, four interested parties
plus the submitter(s) does not seem to be too high a bar, and a more
accurate gauge of sustainability than a numerical majority.
5. On a White Paper: It seems a waste of the talent on this list not to
prepare a more general review of the field, with specific attention to the
specifications required for an upper level ontology. For example, in the
medical/health field alone there are some twenty relevant artifacts, some
public domain, some proprietary, some a combination of the two. There is a
growing service industry in custom-making upper level artifacts for large
organizations with heterogeneous legacy systems. Several people on this
list have called our attention to relevant ISO and government-supported
projects. In other words, there is a growing array of products and
existing practices from which to draw specifications essential to any
standardization process. In addition, we have been informed of a
successful example of a reference object-cum-white paper approach. Other
commitments preclude me from being the *prime* mover or funder of a white
paper, but I would be pleased to help if there is a live body who wants to
undertake the work.
Lee