- From: pat hayes <phayes@ihmc.us>
- Date: Fri, 16 Jan 2004 11:42:10 -0600
- To: Austin Tate <a.tate@ed.ac.uk>, Bijan Parsia <bparsia@isr.umd.edu>
- Cc: public-sws-ig@w3.org
>At 10:19 16/01/2004 -0500, you wrote: >>I'm unclear how the "instantaneous" nature of APs affects APs >>executing in parallel. > > >I don't know what NIST PSL does here... and the tight assumption its >probably got something to do with the underlying theory. We always >assume any activity has finite even if very small time so other >things can interfere. That sounds like a density assumption, if the interference is supposed to happen between the start and end times. That is incompatible with most temporal DB models, be warned. >In fact we even assume that our models are also imperfect and >non-modelled interactions are always possible! > >Of course you can do nicer theoretical tings if you make the >together assumptions... but the real world is just that - and messy. No need to have mess where you can be tidy, though. This issue has been gone into in unbelievable depth, and you can take your preferred option off the shelf. Its really about time-intervals. First, distinguish between atomic in the sense of having no proper subintervals, ie not being divisible, and instantaneous in the sense of taking up no time at all. Not the same idea. Instant(aneous interval)s are 'intervals' that are really pointlike, in the sense that their beginning and ending points are the same point. It is still possible however to distinguish an instant from a bare point, if you want to do so: eg you might want to say that an instant, but not a point, can be the duration of a process. Or, you can say that they are the same thing, which gives you a simpler time-interval theory (eg the Allen relations hold: they don't for instants because if an instant starts an interval then it also meets it) but then you have to either allow points to be durations (of zero length) or else not have zero-length durations. If you allow atoms that are not instants, then time can't be dense or continuous: this is the 'clock-tick' view of time used by almost all temporal DBs. And you need to distinguish atoms from instants or points. All these various options are internally coherent, but you get into muddles if you try to mix them together. They are basically views of the structure of time, so they apply to parallel processes. I would guess that trying to have parallel processes where the components are described using different temporal models will cause something to break, so would suggest deciding on a single temporal ontology first and sticking to it. Pat -- --------------------------------------------------------------------- IHMC (850)434 8903 or (650)494 3973 home 40 South Alcaniz St. (850)202 4416 office Pensacola (850)202 4440 fax FL 32501 (850)291 0667 cell phayes@ihmc.us http://www.ihmc.us/users/phayes
Received on Friday, 16 January 2004 12:42:16 UTC