A couple of comments regarding the proposal for semantics
for DC.Coverage [http://www.sdc.ucsb.edu/~mary/coverage.htm] :
1. In Helsinki it was agreed that the syntax for DC.Time
should be revised to include the possiblity of
(i) various forms of time-spans, including disjoint spans, and
(ii) fuzzy times.
It was commented that this syntax (when complete) should also be
transfered to DC.Coverage - logically being used also for spatial
extent when a "bounding box" style of extent is being specified.
2. The .Polygon and .3d sub-elements need a little cleaning up.
In the current draft:
.Polygon refers to an ordered list of vertices (chain) in 2D
defining a (possibly) non-convex hull, and
.3d refers to an unordered list of vertices which is specified
to be interpreted using the convex-hull enclosing them (note
that using the convex hull will in some cases result in some
vertices being ignored).
I would suggest that for completeness, at least, additional
sub-elements .Polyhedron and .2d should also be defined, with
complementary definitions.
( ------- From here on we get into computational geometry
issues so meta2 listeners might want to tune out ------- )
The use of .2d (convex hull in 2D) is uncontroversial.
For .Polyhedron there is the issue of the "ordering" of
vertices in 3D - what kind of connectivity is implied?
I think we can offer a bit of help here: it has generally
been thought that only the convex-hull of a set of points can
be uniquely defined. However, my colleague Dave Watson has
now developed a robust algorithm for finding the non-convex
hull through an arbitrary set of vertices in n-dimensional
space. Note that for some datasets this non-convex hull
will contain holes, and may consist of multiple disconnected
pieces, but it will always use all of the data.
We intend to make software to compute this available in due
course, but for the present purposes it is merely sufficient
to note that this problem of computational geometry is now solved,
so the restrictions implicit in the draft definition of DC.Coverage
sub-elements should now be understood to be "arbitrary" rather than
"necessary". Maybe .3d should be replaced with .hull.3d.convex,
.hull.2d.convex, .hull.3d.nonconvex, .hull.2d.nonconvex, etc??
Danger of too many levels of qualification, however.
--
__________________________________________________
Dr Simon Cox - Australian Geodynamics Cooperative Research Centre
CSIRO Exploration & Mining, PO Box 437, Nedlands, WA 6009 Australia
T: +61 8 9389 8421 F: +61 8 9389 1906 [log in to unmask]
http://www.ned.dem.csiro.au/SimonCox
|