RE: SUO: Conformance (actually, "standards profiles")
Adam, Chris, et al.,
The recent discussions of conformance prove again one very
simple point, which Pat, I, and others have given up trying
to get across to some people who have buried their heads
in the sand:
A single monolithic ontology is totally unworkable for
many different reasons. Among others, it is impossible
to define any meaningful notion of conformance for such
a monstrosity.
Chris P. wrote:
> Do you literally mean 'define' - so that we could not add
> a term by fitting it within the hierarchy and then adding new
>'axioms'. This seems too strong. It would make life easy as nothing
>interesting would conform.
Yes indeed. But life would not only be easy, but conformance
would be meaningful if we had a well-structured ontology,
organized as a lattice of theories. Then we could specify
exactly which combination of modules we were using without
change, which were not used, and which were being modified
by some kind of additions, deletions, or revisions.
That is exactly what programmers do when they use a library
of subroutines, classes, or whatever: declare exactly what
they are using, not using, and modifying. It is simple,
clean, elegant, and it works.
>However, as I think has been said before, only requiring some kind of
>fitting under would allow odd terms to be smuggled in by subsuming them
>directly under Entity.
Fitting under is exactly what is necessary for any lower
level terms, but no one has ever specified where the upper
level ends and the lower levels begin. In other words,
the SUO should be organized in such a way that things get
progressively more specialized as you go down, but there
is no sharp line between the "upper" and the "lower".
Therefore, it makes infinitely more sense to group the SUO
into well-structured modules, which fit together in a well
structured lattice. Then those modules that depend on others
would "fit under" those modules above them. Everything would
be very clean, simple, elegant, and more importantly, USABLE,
MAINTAINABLE, and CONFORMABLE.
>It seems to me that there is a sensible halfway house. Where we identify a
>level within the top module under which all new (i.e. not completely
>defined) terms have to be hung (and maybe allow some special exceptions, in
>case we are not as exhaustive as we thought). These could correspond to some
>basic ontological categories.
I agree, but with one minor change: replace the term
"top module" with "lattice of modules" and define conformance
to some specified collection of modules that have been chosen
for the particular application.
> This is a commonplace in philosophical
>ontology. So an endurantist top ontology would be incompatible with a
>perdurantist one - e.g. a perdurantist object would have temporal parts
>where an endurantist continuant would not etc.
Yes, and there are many other such areas where incompatible
axiomatizations must be accommodated -- as Pat Hayes has
pointed out in many different postings. The only difference
between Pat and me on this issue is that I have not canceled
my SUO subscription in disgust.
>In fact one could 'define' a top ontology as one that contains an exhaustive
>list of the basic ontological categories - with sufficient 'axioms' to
>characterize these.
Here is where I disagree: There is no such thing as an
"exhaustive list of basic ontological categories". Every time
any scientist makes a new discovery, that so-called exhaustive
list is likely need some kind of revision. In many cases,
it is just a simple addition to some lower-level module,
but occasionally, you get somebody like Einstein who turns
the whole thing upside-down.
Again, if you had a modular ontology, you could accommodate
such issues. But with a monolithic ontology, the only thing
you can do is to replace the entire edifice with a totally new
one. In computer terms, a modular ontology is like Unix, which
can grow indefinitely without making old systems obsolete.
A monolithic ontology is like Microsoft, which creates havoc
every two years in order to maintain its profitability.
>It would be easier to identify the top ontology (and its lower boundary) if
>we had a top-bottom segmentation as well as a side-to-side one. An argument
>for some kind of modularization?
You betcha!
John Sowa