- From: Austin Tate <a.tate@ed.ac.uk>
- Date: Sat, 17 Jan 2004 11:12:19 +0000
- To: pat hayes <phayes@ihmc.us>
- Cc: public-sws-ig@w3.org
At 05:16 PM 16/01/2004 -0600, pat hayes wrote: >>I have argued that SWSL should simply allow a process description to be >>made up of: >> >>a) a set of activities to be performed, each of which are considered to >>have a begin and end time point > >Suggestion: each process occupies (? lasts for, endures during, takes) a >time-interval, and intervals have begin and end timepoints (and indeed are >defined by them, uniquely.) Intervals are handy things to have in the >ontology anyway. Agreed... and that is what I meant. An activity always is associated with a time interval defined as the begin time point of the interval and the end time point of the interval. *** As you know we formulated this simple description (used in NIST PSL too) long ago along with James Allen - Mr. Interval:-) - and he has no problem with seeing intervals as having two time point ends (begin and end) - transformation between the two is always possible - contrary to some early arguments that this might not be so. >>b) a set of constraints (of predefined and extendible types) on and >>between those activities and the objects/entities in the world associated >>with them >> >>c) a set of annotations (of predefined and extendible types) on the above >> >>We would then re-engineer back those predefined things we want from OWL-S >>to make sure we cover the specific instance everyone knows and loves... >>but in a more coherent and very extendible framework. > >I entirely agree. Lets keep it simple... one of the criticisms of much of what is now happening is that these are exotic and far too difficult to use... and will have short life as technology and logics change. We need an underlying abstraction that works VERY long term. That is what I suggest we use the above for - and its pretty much what the SWSL WG agree we need as a sort of introduction. But we then have to follow up by showing the more detailed models and logics for those models we use can relate to this abstraction in all cases. then these can come and go and be replaced as better representations come along - without losing the relationship through the abstraction. Austin *** by the way in my work we call something an "event" if it is associated with a time point. This sis what some of you may be calling an instantaneous activity. But even a time point has sort of two sides so you can talk about the side before the time point and the side after the time point. Its sort of an infinitesimal small interval where you cannot (in your model) divide it into sub-intervals - as Pat said.
Received on Saturday, 17 January 2004 06:23:26 UTC