Hi Simon, excellent.
I've been feeling and expressing this same idea. Feature of Interest
should be an earth realm or a name place from a gazetteer, where we
could infer the earth realm. Should not be a geometry, as Ron said.
But, my feeling is that when a domain scientists refer to a type of
data ( trajectory, station, profile .. ) they are really referring to
characteristics of the observing procedure ( in this case .. the
constraint behavior of a sensor or platform ) which is confused
sometimes with the feature of interest.
-Luis
On Mar 13, 2008, at 9:53 PM, <Simon.Cox@xxxxxxxx> <Simon.Cox@xxxxxxxx>
wrote:
Actually I think the issue is that the domain scientists, quite
reasonably, are used to working with some domain-assumptions. For
example, when you are an atmospheric scientist talking to other
atmospheric scientists, the subject of your studies and observations
is
the atmosphere. Doesn't need to be stated explicitly, or maybe is just
inferred from the fact that elevation component of the location is >0.
I'm comfortable with this. That's really why the SamplingFeature is
useful. It allows you to work with a feature that is primarily defined
by its shape for everyday purposes, without totally abandoning a
rigorous model that recognises the fact that there is a domain feature
underneath, and it is really the domain feature that has the
properties.
To help with this, here's some "convenience" features, that you can
use
as the "feature of interest" of an observation, of the "sampled
feature"
of a sampling-feature ;-)
<https://www.seegrid.csiro.au/twiki/bin/view/CGIModel/CGIFeatureRegister
#Register_content>
