Re: Significant milestone for netCDF-4/HDF5

Hi Russ,
    That's great!  We've almost wrapped up the float<->int conversions in the
development branch of HDF5, so it will be possible to drop the "shim" that Ed
has in place for this feature shortly.

    Quincey


> Ed just reported that he got his netCDF-3 C interface that uses the
> HDF5 API and storage layer to pass the current netCDF comprehensive
> test suite.  This uses an unmodified HDF5 1.6.1 library and a
> basically unmodified nc_test from the latest netCDF 3.5 release.
> 
> Ed's implementation demonstrates that the current HDF5 even without
> enhancements is adequate for a correct implementation of netCDF-3 and
> that the backward-compatibility approach is sound, since the new
> library can use the old interface and can access data from either
> current netCDF-3 or new HDF5 files transparently by just recompiling
> and relinking old netCDF-3 programs.  We weren't sure this level of
> both API and format compatibility could be achieved, but now we know
> it's practical.
> 
> I think this puts us ahead of schedule, providing a baseline
> implementation we can later compare with a tuned version that uses an
> enhanced HDF5 library.
> 
> Now that the netcdf-3 tests pass, we can start to investigate and
> write tests for some of the new functionality that this gives us with
> very little additional work, such as large files and new data types.
> 
> We can also begin planning improvements to both the current prototype
> and the HDF5 implementation to serve as a basis for the netCDF-4
> enhancements we have in mind.  We could also look at performance,
> taking care to avoid premature optimizations until we understand how
> to make the changes needed to support the netCDF-4 enhancements.
> 
> Thanks to Ed, Quincey, and others who have been involved in helping to
> achieve this significant milestone!
> 
> --Russ
> 
>From owner-netcdf-hdf@xxxxxxxxxxxxxxxx 03 2003 Dec -0700 13:47:50 
Message-ID: <wrxptf5mxvt.fsf@xxxxxxxxxxxxxxxxxxxxxxx>
Date: 03 Dec 2003 13:47:50 -0700
From: Ed Hartnett <ed@xxxxxxxxxxxxxxxx>
To: netcdf-hdf@xxxxxxxxxxxxxxxx
Subject: distribution musings - install three libraries to get netcdf-4?
Received: (from majordo@localhost)
        by unidata.ucar.edu (UCAR/Unidata) id hB3Klqks001858
        for netcdf-hdf-out; Wed, 3 Dec 2003 13:47:52 -0700 (MST)
Received: from rodney.unidata.ucar.edu (rodney.unidata.ucar.edu 
[128.117.140.88])
        by unidata.ucar.edu (UCAR/Unidata) with ESMTP id hB3Klpp2001853
        for <netcdf-hdf@unidata>; Wed, 3 Dec 2003 13:47:51 -0700 (MST)
Organization: UCAR/Unidata
Keywords: 200312032047.hB3Klpp2001853
Lines: 32
User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Sender: owner-netcdf-hdf@xxxxxxxxxxxxxxxx
Precedence: bulk


Howdy all!

As I prepare the fist tarball for netcdf-4 I am musing about some
interesting distribution requirements we have.

Here's the section from the requirements (I just added the last one
this morning.) The entire requirements document can be found as always
at http://my.unidata.ucar.edu/content/software/netcdf/netcdf-4/reqs.html

Distribution with both netCDF and HDF5 packages

    * The netCDF-4 library is distrubuted by both Unidata and NCSA.

    * The netCDF-4 library requires that HDF5 (version 1.8.0 or
      greater) be installed. 

    * The netCDF-4 library can coexist peacefully with the netCDF-3
      library, but both cannot be used in the same program. 

    * Potential users can select from a number of distributions,
      depending on whether they already have netCDF-3 and/or HDF5
      installed. Whatever the circumstance, users should be able to
      install netcdf-4 in one pass through the usual configure/make
      install/make check cycle. 

What do we all think of those requirements. Anyone object to any, or
want to add to them?

Thanks,

Ed