NOTICE: This version of the NSF Unidata web site (archive.unidata.ucar.edu) is no longer being updated.
Current content can be found at unidata.ucar.edu.

To learn about what's going on, see About the Archive Site.

[netcdfgroup] taking szip write capability out of netCDF snapshot code...

Howdy all!

Recently I announced the availability of two new C functions,
nc_def_var_szip and nc_inq_var_szip, (and equivalent code for F77 and
F90) to allow full netCDF-4 support for szip compression.

Since netCDF-4 can already read szip files, nc_inq_var_szip is
not controversial. It simply reports existing szip settings. The
controversial part is allowing the user to *write* szipped netCDF-4
data.

After I added this we had a discussion within the netCDF programming
team, and have decided to take out the capability to write szipped files
from netCDF-4. Our discussion is summarized below for those users who
are interested in szip compression.

Szip Pros:

* Faster than zlib.
* Requested by some users. (But one, NASA GMAO, has since renounced
  using szip because it provides too many data distribution problems.)
* Already in use in HDF5 world.
* No licensing restrictions for reading.
* Read-access already possible in netCDF-4 with szipped HDF5 files.
* Source code is freely available - no danger of szipped data being lost
  to science.

Szip Cons:

* No existing Java version.
* License restrictions for commercial writers of data.
* Some (or most) netCDF-4 installations will not be able to read szipped
  files without rebuilding netCDF.
* Will not (and should not) be used by CMIP5 effort and (probably) other
  important archives. 
* Due to licensing szip will not be available in stock Fedora
  distribution. Fedora is a very popular Linux distribution, and at least
  some other free software distributions will probably feel the same
  about szip licensing problems.

Conclusion:

I will take out the def_var_szip function in all APIs. NetCDF users will
be able to use existing HDF5 szipped data, but not create it with the
netCDF API.

Thanks,

Ed

-- 
Ed Hartnett  -- ed@xxxxxxxxxxxxxxxx



  • 2009 messages navigation, sorted by:
    1. Thread
    2. Subject
    3. Author
    4. Date
    5. ↑ Table Of Contents
  • Search the netcdfgroup archives: