DARPA AFRL Planning Initiative
SPAR Core Group
Report for Meeting 4
Canterbury Hotel, Sutter Street, Downtown San Francisco - 4/6-Nov-97

Attendees | Take Home Messages


Briefing, discussion and break out session during ARPI Workshop. This meeting involved a number of Specialism Experts Panel and User Panel members who were at the ARPI Workshop as well as other commentators.

Take Home Messages

Here is a summary by AT of the main points that he came away with from the meeting.
  1. Many people could not see how the top level abstract model could realistically support all the nitty gritty detail they know they need in military plans. We have to do a better PR job as several people put it.
  2. Several people (e.g. DM) thought we should just accept the top level model as is, and not spend all our time refining it. I take this to mean we should REDUCE effort on the top down approach, but not necessarily stop our discussions where we are aware of issues to resolve (e.g. agreeing the 20 sentences or their replacement will leave us in a better position to keep a firm grip on the conceptual model as issues arise during the implementations discussion for JFACC and TIE 97-1).
  3. We need to ground the work in real military plan representations without losing the top level conceptual model. This is where effort should focus next. We plan to develop detailed specialisations for SPAR for: Several others will be seeking to ensure specialisations are possible for other programs, but these 2 will focus the core grounp work over the next few weeks. We expect to explain the work we are doing by having (along with those in the core team we can consider as primarilt responsible for making their level happen): a top level SPAR-generic (AT/AP); a middle level SPAR-military (aka AITS Warplan - AP/TC); and specific instantiations - in particulat JFACC Plan Rep (BS/KM/AP) and TIE 97-1 (BS).
  4. We should not rush out the next version until input from 3 can be rolled into our thinking. SPAR 0.2 was meant to be just an increment of the outstanding issues list added to the current documen t after we had quick feedback. However another version so soon after the meeting that does not alter in terms of the core model could be misinterpreted. My suggestion is that we replan the releases as follows: This would still allow us to give sensible input to AITS Warplan efforts around February/March as is required. The main RFC stage between 0.2 and 0.3 is then 2 months each rather than the originally planned 3 months, but that is not too bad.
  5. Uncertainty needs to be built in more widely than in numeric values. But it may be possible to make progress on sharing SOME information by employing a basic mechanism of allowing specs. of a wider range of entity values representing things evaluable in the real world using uncertain values combined with definitive value bounds where computable. I suggested a new design guideline to addres sthis in a previous message.
  6. There was some comment (e.g. DE) that SPAR should not attempt to be a basis for implementation structures in some systems, and that it should concentrate on being an interlingua only (like PIF and NIST PSL). Quite crucially, I understand that Tom Garvey confirmed during the Wednesday morning breakout session that he agreed that SPAR should just be an interlingua and not attempt to be the basis for implementation structures. This is contrary to the way I understaood his guidance befopre, and conflicts with some of the issues we were asked to address such as efficiency and so on. As I was out of the meeting for the half hour that this "change" in guidance took place, I would like Tom to confirm this new guidance, or to clarify what he actually said and wants from SPAR.

ARPI Home Page | SPAR Home Page
AIAI Page maintained by Austin Tate (a.tate@ed.ac.uk), Last updated: Mon Nov 24 13:08:37 1997
Please make contact if you have any comments on these pages or the SPAR Project.