Jonathan,
> Should we put a general convention in CF that text attributes can only
> contain US ASCII characters? I think that would avoid some problems.
However it would also appear to place a new restriction on variable (and
dimension) names that is not currently in CF, namely that only US ASCII
characters could be used in such names. I don't believe the current CF
has any such restriction on the characters used in variable names.
This new restriction follows, because variable and dimension names are
used in text attribute values for attributes such as coordinates,
bounds, cell_measures, cell_methods, and formula_terms.
Unidata added support for UTF-8 names some time ago to permit
non-English variable, attribute, and dimension names in netCDF files.
This was done is a way that did not break backward compatibility:
http://www.unidata.ucar.edu/netcdf/docs/faq.html#fv22
That means there may already be CF-compliant data that makes use of
non-ASCII characters in variable and dimension names and hence in text
attributes such as coordinates.
--Russ
Received on Mon Mar 29 2010 - 10:43:25 BST
