Re: A couple more comments on the netcdf requirements list.


> Here's another brief comment on the 'netCDF4 requirements' list,
> The list includes chunking, but does not include compression.
> I don't know if compression support is 'required' for netCDF4,
> but it is supported by HDF5 and is a huge performance plus.  We
> know that NASA is extremely interested in compression.
> I would include 'Compression and other filters' as an item, or
> else add it to 'Chunking and compression'.  

Thanks, this was an inadvertent oversight, since compression is
clearly mentioned in the original proposal:

 * Determine and create interfaces to specify chunking and compression
   parameters on file creation. Implement unit tests for chunking and
   compression and test the implementation.

I think it should be a separate item from chunking, so "compression
and other filters" sounds right.

> Another feature that should be considered for netCDF4 is
> HDF5 'file drivers' other than MPI/IO.  These are simple
> to add to the API, require nothing of the netCDF profile,
> and have significant benefits for some users.

OK, thanks for the suggestion.  If we can get this flexibility for
free by using the virtual file driver stuff in HDF5, it sounds like we
should do it.



Russ Rew                                         UCAR Unidata Program