[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 19981110: Indirect Access in netCDF-4

>> To: address@hidden
>> From: Mark Ohrenschall <address@hidden>
>> Subject: Indirect Access in netCDF-4
>> Organization: NOAA/NGDC
>> Keywords: 199811102203.PAA04191 netCDF 4

Hi Mark,

Sorry it took me so long to get to this ...

> In the Rew and Davis white paper UNIDATA'S NETCDF INTERFACE FOR DATA
> ACCESS: STATUS AND PLANS section 5.2 speaks of link variables, which are
> further addressed in section 6. NETCDF-4 AND FUTURE DIRECTIONS.

> I think I understand the discussion, but I'd like to be sure.  What I
> would like to do is to have raw binary data which is in a file that is
> separate from any application-specific meta-data.  Then I would like to
> provide several separate meta-data files, such as an HDF header, an
> IDRISI header, a GRASS header, etc., and also a netCDF header.  All
> these headers would reference a common raw binary data file, and HDF,
> IDRISI, GRASS, etc., and netCDF applications would be able to read this
> data the same as if the data were in the more traditional packaging.

> Does this fall under your plans regarding indirect access and link
> variables?

Not exactly.  In order for netCDF to read the data, it would have to be
in a machine-independent form, rather than in raw binary form.  I was
actually thinking the link would be either to a variable in another
netCDF file or be an expression involving variables in other netCDF
files.  I think it's unrealistic to imagine that netCDF, HDF, IDRISI,
GRASS, etc. could all access the same binary data, since they use
different data models.  Maybe some least common denominator data model
could be made to work, but it would be missing much of what makes these
different data models useful.

> Do you plan to implement this in release 4 of netCDF?

> Do you have a release date in mind?  If not, do you expect this release
> anytime reasonably soon?

No, we're not working on this now.  We didn't get the resources we
requested in our latest 5-year proposal, so we've had to put some plans
on the back-burner.  Currently our netCDF development is mostly for the
Java.  The MultiArray interface that Glenn Davis has developed has some
of the features of the link idea and of your idea above, in that it
presents alternate interfaces to the same data without making multiple
copies.  Furthermore, the MultiArray package is independent of netCDF
(but used by netCDF) so that other packages such as HDF could present
multidimensional data as MultiArrays too.  For more details on this, see

  
http://www.unidata.ucar.edu/packages/netcdf/java/docs/ucar/multiarray/package-summary.html

or just

  http://www.unidata.ucar.edu/packages/netcdf/java/

for more on netCDF in Java.

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu