SUO: Unanswered Questions About SUMO Set Theory
At 23:42 2002-05-22 -0700, John F. Sowa wrote:
>
> Todd, Adam, Bill, et al.,
>
> As this discussion indicates, there are many points of view
> about the goals of the SUO and even about the interpretation
> of the statements of the goals embodied in the charter.
>
> Several of the founding members have a narrow set of goals,
> which Adam has interpreted as implying a single, monolithic
> ontology:
>
> > There are different opinions on this list. When we wrote the PAR
> > for this group the intent was to have a standards effort aimed at
> > ontology content - creating a set of terms and formal definitions.
>
>
> I agree that content was and still is a central focus of the SUO
> project. However, there is a very serious question about the form
> in which that content is presented and made available. Another
> very serious issue is the recognition of multiple, often conflicting
> selections of content, which are required for various purposes.
>
> > A large number of people who have joined this group are interested
> > in a different direction however.
>
> I am probably one of the people whom Adam includes in the "different"
> direction department, even though I agree with Adam that content is
> essential. The fundamental difference is that I would add the
> following qualifications:
John-
My sense of why Adam includes you in the "different direction department" is that he is being polite. Simply put (or "bottom line", as you like to say): you've not contributed any standards wording to this project ... on any of the approaches. You've not proposed any improvements ... you've only had complaints, platitudes, and digressions. Sure, not everyone will agree with what has been proposed, but there is little "meat" in your statements. You've complained many times about defects in SUMO, yet provide no alternative wording (inside or outside the context of SUMO). "Where's the beef?"
> 1. No one today (and perhaps no one ever) is smart enough to select
> the best or even a good set of theories that would qualify as
> an obviously great set of axioms and definitions that should
> be given a privileged status over any and all others by being
> certified as an IEEE standard.
I believe you have a misunderstanding about standards. Standards are supposed to be "good enough" for the stakeholders involved. While perfection is a desirable goal, sometimes it's not possible. But the real question is: can we get to something that is good enough? The reason why good enough might be (well) good enough is that we have a corrigendum (bug fix) process, an amendment (minor improvement) process, and a revision process (major improvement, reality check on current technology).
So one shouldn't think about the standard "for now and forver", but a "good enough standard that can be built upon and improved over time".
An example, the ISO folks were working on units of measure in ISO 31-* (I mentioned this in a previous E-mail). However, a handful of years after the third edition of the standard was released, they need to make changes (fixes/improvements). One of the changes concerned the status of whether or not a "radian" (as in geometry) was a unit of measure (I've simplified the issue). They decided to change their mind (based on input from CIPM). I'm sure many people on this list could make arguments for either side, however, the ISO folks made a decision (and making decisions is an Important Thing). The users of the standard adapted to the new definitions. And there were ways to refer to both definitions: the 1992 edition, or the 1992 edition with Amendment 1 (1998).
So this all worked because: (1) they were willing to put words on "paper", (2) they were willing to get review and external input, (3) they *desired* consensus (I'm not convinced all SUO WG members want consensus), (4) they were willing to craft wording that balances needs (not too overspecified, not too underspecified, etc.), and (5) they got to a document that (largely) reflected reality and got industry adoption.
And this worked with their users because they understood the nature of the standard, including areas of potential change/error/weakness.
> 2. On the contrary, the history of science and technology shows
> that all theories are at best adequate approximations for some
> particular purpose that has been felt at some particular point
> in time. Even the best of them are hopelessly inadequate for
> other purposes at other times.
True. However, I'd say that for every great scientist/technologist, he/she did not think "Something much much better will come along when I'm dead and gone, so there is no point in writing down my ideas, hardening/refining them with feedback, and revising them when new things come along". OK, so maybe there were some scientsts/technologists that thought like this, but they didn't share their ideas with the public. :-)
On one hand you say "the world is forever changing", and on the other hand you say "we can't agree upon anything, until things stop changing". While you're figuring that one out, it would seem reasonable that you let the rest of us make some progress on getting towards "good enough".
> 3. Therefore, the SUO should adopt a framework that can accommodate
> all possible theories, show precisely how they are related to one
> another (as generalizations, specializations, or alternatives),
> state clearly what purpose(s) each one is intended to satisfy,
> and include a history of endorsements, certifications, and
> cautionary notes that various users have experienced.
More platitudes. Got some standards wording to propose?
> SUMO is one collection of theories that many people believe are
> valuable and some people would claim as sufficient for a wide
> range of applications. OpenCyc is another collection of theories
> with similar claims. I believe that both of them are valuable,
> but I would not want either one to be privileged over the other
> or over many other equally worthy theories.
I'm interested in "good enough". If there are several that are "good enough", then they can be related.
> My recommendation is to accept the theories (or microtheories) of
> SUMO and OpenCyc as content modules within a framework that could
> also accommodate other contributions as well.
So how does that statement translate into standards words?
> The IFF working group is developing such a framework, and I would
> be happy to support a collaboration of the SUMO and OpenCyc
> proponents to cast their collections within the IFF framework
> (perhaps with further developments and modifications of IFF
> that might be needed to accommodate them).
>
> But I would never support either SUMO or OpenCyc by themselves
> as IEEE standards. I also have some reservations about CycL
> as a proposed standard. I would endorse CycL if and only if
> it were included in the CL (Common Logic) framework that also
> supports KIF, CGs, and the infix notation for predicate calculus.
More platitudes. Got some standards wording to propose?
-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