Forslund's message re specifying compression came to me while I was typing
mine re subscript ranges.
On the topic of reasonable extensions, the Livermore library PDB and the SILO
extension to netCDF add an additional type, "object", related to C structures.
(also directories, whose utility has been noted elsewhere).
Maybe the OOP folks can help us *minimally* extend CDF while adding some
ability for the people in the field to add to the netCDF file new types, not
anticipated by Unidata, together with pointing to the coding to deal with
those types. Maybe a compression scheme would be such a thing, at least
in its prototype stage.
Let's not make the distribution netCDF too complicated.
