Hi Jon, all,
My understanding of this issue is that the CRS should be identified by a URN
but not necessarily a 'code' so your suggestion of resolving to a machine
readable document is perfectly valid as long as that document has an
identifier.
But one of the big problems with referencing other things in this way is that
a URN can effectively refer to anything and there may not even be a resolver
service at the other end (there might be a hard copy document!) - so
ultimately I think your application needs to agree upon and understand
the 'type' of URNs it is going to handle and deal with those. The benefits of
using URNs of course are that somebody else does the work of
defining 'things'.
OGC are investing time in URN development:
http://www.opengeospatial.org/ogcna
But that's not to say the CF community shouldn't also invest time in defining
resolvable coordinate reference systems specific for our needs.
This does bring up the whole interoperability argument again though... :)
I agree it is one of the main challenges for CF when it comes to OGC.
Cheers,
Dom
On Thursday 09 October 2008 10:46:43 Jon Blower wrote:
Hi John,
This is a serious general problem I think. In the case of
parameterized projections you could express the projection in
something GIS-friendly like WKT, but not all of these projections will
have an EPSG code of course. (CRS identifiers don't need to be EPSG
codes, one could come up with another authority, but that's not the
real issue.)
It is a real problem that there are an infinite number of CRSs but WxS
expects the CRS to be identified by a code. Perhaps a WxS extension
could use a resolvable URI as the code, which points to a document
that describes the CRS in machine-readable form?
We also have the problem of the "arbitrary" coordinate systems that
can only be expressed through an exhaustive listing of the lat-lon
points that comprise the grid.
I think this is one of the main challenges for mapping CF-NetCDF to OGC.
Cheers, Jon
