status of nj22 WRT netCDF-4

John Caron caron at unidata.ucar.edu
Fri Sep 22 07:42:17 MDT 2006



Simon Wood wrote:
> Thanks John, though I'm now a bit confused I think.. ;-(
> 
>> True:
>>  nj22 implements the CDM data access layer
>>  the CDM data access layer is the same as the NetCDF-4 data model
>>
>> Not true (yet):
>> nj22 implements NetCDF-4  Details:
>> When the NetCDF-4 file format is finalized (waiting on HDF5-1.8), then 
>> nj22 will be able to read NetCDF-4. It will not immediately be able to 
>> write, and depending on how complicated things are (HDF5 is pretty 
>> complicated, esp compared to netcdf-3), it may end up only only 
>> writing a subset of netcdf-4. I would guess eventually it will write 
>> the whole thing, but that will depend on funding. There has never been 
>> an independent  library that writes (or reads) HDF5, so we are wary of 
>> promising what hasnt been done before. We have prototyped HDF5 reading 
>> enough to know we can do that.
> 
> 
> Hang-on, I thought that the HDF5 stuff was required to implement the new 
> features in the netCDF-4 data model?  Certainly that is how I understand 
> things to be in the netCDF-4.0-alpha libraries -- to use 'netCDF-4 
> format' requires HDF5-1.8 (or 1.7.52), though of course things may be 
> different in the nj22 world.  What storage model does nj22 use to 
> implement the CDM data access layer if not HDF5?  Does nj22 extend the 
> netCDF-3 file format to support eg multiple unlimited dimensions etc? or 
> does it not support all of the netCDF-4 data model? (is the CDM data 
> access layer a subset of the netCDF-4 data model?)
> 
> What am I missing?

Hi Simon:

The nj22 and netCDF-4 C library are completely seperate implementations of the CDM, and the nj22 does not use HDF5 (for one thing, there is no java HDF5).

What you are missing is the difference between the data model and the persistence format, ie file format. nj22 reads many file formats, but writes only to netcdf-3 files, and so cant write out the complete CDM to disk. For many uses, theres no need to, since the data is already on disk in some other format (GRIB, NEXRAD, etc). when netCDF-4 file format is finalized, it will be the persistence format for the complete CDM (with previous caveats).

In short you dont need a storage format (writing) for data access (reading).

hope that makes sense

John

==============================================================================
To unsubscribe netcdf-java, visit:
http://www.unidata.ucar.edu/mailing-list-delete-form.html
==============================================================================



More information about the Netcdf-java mailing list