I wrote:
> ... plans are for version 9.2 of ESRI's GIS software, available
> about January 2006, to be able to read and write CF-compliant netCDF
> data.
and Rich Signell asked:
> Does this mean that CF has agreed on how to specify full geospatial referencing
> information?
No, I should have been clearer. It means ESRI asked us (John Caron,
Don Murray, and I) for some advice about what to do to fill in the
gaps, and we discussed an approach that involved
- using CF conventions where they apply (coordinate types, coordinate
systems, grid mappings, and projections);
- assuming reasonable defaults when the CF conventions are not
explicit (e.g. WGS84 spheroid);
- committing to support the CF conventions as they evolve;
- advocating additions to CF that complete its ability to represent
needed geospatial referencing information; and
- in the meantime, when writing netCDF files that require geospatial
referencing not addressed in the current CF conventions, adding
extra information in the form of extra variables or attributes in a
well-documented form.
This last bullet relies on a loop hole: adding extra attributes not
mentioned in the CF conventions to otherwise CF-compliant data
preserves CF-compliance. So in that sense the data is CF-compliant,
but since these extra attributes aren't in the CF conventions,
software that reads CF-compliant data won't necessarily know how to
interpret them.
For example, if a dataset contained a text attribute named
"geospatial_reference" with a WKT (Well Known Text) or GML
representation for geospatial referencing, it might still be
CF-compliant. Later, if the CF conventions evolve to specify a
different set of attributes to represent the same information, then
including both representations will make the data both CF-compliant
and backward compatible.
I think this is a reasonable approach to recommend to developers who
have to deliver software that must be feature-complete fairly soon,
but it can obviously result in diverging conventions and loss of
interoperability if we don't agree on the needed additions.
--Russ
_____________________________________________________________________
Russ Rew UCAR Unidata Program
russ at unidata.ucar.edu
http://www.unidata.ucar.edu
Received on Thu Apr 21 2005 - 16:25:05 BST
