Re: the precondition property in OWL-S 1.0

   [Austin Tate]
   Note my comment below is for the SWSL process model really and not just for 
   tidying up OWL-S as its too radical a step. But it might be something that 
   could guide any changes in OWL-S to...

   I am not keen on adding lots of separate properties to activity... where do 
   you stop... do you add resource, spatial, authority, temporal constraints 
   (some people even try to maintain successor and predecessors lists in each 
   process step!).

   I would favour a much simpler underlying idea that an activity always has

   [three perfectly reasonable suggestions elided]

   Then we can avoid the syndrome of 25 properties on activity - growing to 75 
   properties as more requirements emerge.

It's not clear what your argument is.  Your suggestions for properties
make sense, and may be the best set for interfacing to a
constraint-based planner.  But if someone is concerned about some
aspect of an activity that relates to the other 72 properties, what
harm can it do to let them use more properties?  
Different applications could look at different subsets of properties
and be pretty much oblivious to each other's concerns.

If you're committed to an implementation in which an activity is
represented by a vector of properties, it might make sense to worry
that things are getting out of hand.  It seems a little premature at
this point.

                                             -- Drew


-- 
                                   -- Drew McDermott
                                      Yale Computer Science Department

Received on Monday, 10 November 2003 17:55:49 UTC