[Data-modeling] Events

[Jonathan W. Lowe]

On Thu, 2008-02-14 at 00:05 -0500, Tom Morris wrote:
> > As I see it (now), the fundamental building block is just a time period
> > which is simply two date/times, a start and an end.
> 
> I think there are classes of events which are basically considered
> instantaneous, so a duration shouldn't be mandatory.  For example,
> childbirth can be a long, involved process, but a birth certificate
> will typically have just a single date and time which is considered
> the official time of birth.
> 
> > Currently the /time/event includes additional properties such as "Location"
> > and "People involved" which I don't think should necessarily be on the event type.
> 
> I agree that events don't necessarily involve people, but I'm hard
> pressed to think of an event which doesn't happen at a location,
> however ill defined that location might be (e.g. Earth, our solar
> system, etc).

The reverse is certainly true -- a place is also a time.  However, when
modeling locations, people associate sets of coordinates with properties
(such as a name) or with types (such as "Company" or "Body of Water")
but often omit any timestamps. The implication is that the association
has a timestamp of valid-from = "now" and valid-to = infinity even
though the modeled object may not always have existed or been located as
modeled "now" and may not continue to exist or retain the same set of
location coordinates forever.

I suspect that people omit timestamps when modeling locations because
the majority of the world's use cases involve decision-making in the
present or a near-term future rather than the past (e.g. legal actions
such as described in the 9/11 article) or longer-term future (e.g.
climate change).  Even after establishing a base set of locations, most
GIS customers I've ever encountered want to replace rather than archive
older data as soon as newer data becomes available.  I mention use cases
and "the majority" because they seem (understandably) to be key litmus
tests for Freebase schema decisions in this alpha stage of development.

That said, modeling spatio-temporally is a lot more interesting than
modeling purely spatially.  :-)
> 
> > For added thought, I'd like to point you at an article that Robert sent to me
> > about Events, and about 9/11 and how many events actually occurred when
> > the planes hit the towers.
> 
> A lot depends on your perspective and the level of detail at which you
> are modeling things.  If you were considering things from the
> perspective of the thousands who died, you might find that the
> "correct" answer was not one event or two events, but thousands of
> events, each occurring at different time and place (presuming you were
> recording at sufficient precision).
> 
> Tom
> _______________________________________________
> Data-modeling mailing list
> Data-modeling at freebase.com
> http://lists.freebase.com/mailman/listinfo/data-modeling
>