- From: Austin Tate <a.tate@ed.ac.uk>
- Date: Sun, 09 Nov 2003 11:32:33 +0000
- To: public-sws-ig@w3.org
At 10:24 PM 08/11/2003 -0500, Drew McDermott wrote: >All we need is the hasPrecondition property that relates a >step to a Condition. I agree with this... but here may a way to do this that generalises nicely... We have statements (about the "world" or domain) in some language that can be evaluated as holding true or false. Some of these are temporarily invariant. But in many cases these can be temporally specified as holding at a time point, or between time points, or in some other temporally defined scope. In my work we call this a "world state constraints" so we can relate it to the more general notion of constraints of many kinds (temporal, spatial, resource, world state, etc.) and have a simple extendable way to relate all constraints in a process model. We then have specialised uses of these in things like "conditions" at time points (often at the begin time point of ac activity, but also on other temporal specifications), and effects. I would argue that such conditions should be on the time specifications such as time points (normally but not always associated with an activity in a process model) and not on the activity itself. This allows much more flexibility - as has been found by those working on PIF, NIST PSL and other flexible process models. So we then have these specialised to give us convenient specialisation for what we want to use all the time ... a) precondition - defined by a statement that holds at begin end of an activity. b) effect - defined by a statement holds at end of an activity Another that we will need I am sure (we already use these in our process models) c) condition that holds from begin to end time points of an activity (preservation conditions). Austin
Received on Sunday, 9 November 2003 06:33:40 UTC