default CRS for meteorological data

There is an argument (again) that all coordinates in CF should have CRS specification.
http://mailman.cgd.ucar.edu/pipermail/cf-metadata/2014/057478.html

My thought:

- I agree Jim says default CRS is needed for backward compatibility.

- The default should be WGS84 http://epsg.io/4326

- For NWP model data, even if the modeller says the coordinates are spherical, that means only the equations are spherical.  I've never seen surface elevation and observations are injected *after* coordinate conversion.

- For observational data, I hope the data source is migrated well to WGS84.  Archived data is an issue, but that can't be an excuse to continue to create the issue.

Similar argument should exist for vertical CRS.  I think it is convenient to use "MSL height" http://epsg.io/5714 which has no clear definition as an default.