Re: SUO: Modularity
Let me give you one real instance that I have been involved in that makes
me worried on this score...
About 10 years ago under the DARPA Planning Initiative when it started it
was recognised that a shared representation of plans would be a "good
thing". There was a lot of willingness to arrive at such a shared
understanding as many of the researcher (in universities, government and
industry) were to work on things together and a lot of the researchers had
already developed quite rich and partially overlapping representations of
processes, plans, activity, world state conditions, temporal constraints, etc.
About 14 of the groups involved worked with professional support and some
great seed ideas from James Allen to create KRSL. It was a sort of
pre-standards activity in some ways. It was going very well indeed and
many of us had started to write importers and exporters to our own plan
representations for the agreement as reached by the whole 14 groups in a
document create by the group. But it really was an UPPER ontology of
processes and plans, and people had gone just as far as they could whole
retaining flexibility and openness to differences with refinements of the
agreed core.
What happened then ruined it. There was a big demo coming along, and of
course they needed the detail for the actual systems that would use the
upper ontology in that demo. The people involved simply kept extending the
single document. There was no attempt to delineate the boundary between
the core that had broad consensus and the extensions needed to fit to
specific systems and tools. the ontological commitments made at this stage
meant that almost all other people's systems were incompatible and many
more flexible approached wee excluded if you want to use the KRSL as it was
finally documented.
The result - no one used it beyond the specific demo. If we had just had
the sense to keep the agreed core in one document and add an appendix to
show a SPECIFIC refinement to meet certain needs I still think that core
would have gone on to become the basis for process and plan representations
in many military and other systems today.
Instead, it took a 5 year gap, and a reconstituted sub-set of the original
group working working on KRSL2 starting from the earlier real upper
ontology drafts and working with new partners (Adam Pease associated with
the IEEE SUO effort being one in fact) to get this going again. A wasted
opportunity. This work fed into PIF Core and NIST PSL Core by the way.
I don't want the IEEE SUO effort to go the same was as the original KRSL.
Austin