Now we wait for any bugs to be reported...
It is always a tremendous amount of effort to make a netCDF release. There are so many people depending on it, and I really don't want to mess up up.The 4.1 release, which I did last Friday, was my sixth netCDF release. This release has a lot of new features - more than usual, since we have had more than the usual number of programmers working on netCDF. It hasn't been just me and (part-time) Russ, as in previous releases. This is the first release that contains work from Dennis, and what a large piece of work it does contain: the new opendap client.
In addition, the 4.1 release contains a new utility, nccopy, which copies netCDF files, changing format on the way if desired. (Russ did nccopy). There is also a new ncgen, finally up to speed on all netCDF-4 extensions, which Dennis did.
As for me, I put in a layer that can read (many but not all) HDF4 files, a layer that can use parallel-netcdf for parallel I/O to classic/64-bit offset files, a modification to allow most existing HDF5 data files to be read by netCDF. Finally, and at the last minute, I changed the default settings of caching and chunking so that the users producing the giant AR-5 datasets would get good performance. Also, I added the nc-config utility to help users come up with correct compiler flags when building netCDF programs, and two new libraries are part of the distribution, UDUNITS (Steve Emmerson) and libcf (yours truly).
Probably we should have released 6 months ago, and held some of those features back for another release. The more new features, the harder it is to test and release.
But it's out now, and we have started working on release 4.2. For that release we have more limited ambitions, and I hope it will be out this year.
Would sure be nice if NetCDF was more like other open source projects, public bug tracker, public code revision system :)
Posted by 75.71.9.118 on February 10, 2010 at 07:30 AM MST #
Is there an easy way in IDL to find out what is inside a netCDF file, and to read the data and metadata (attributes) from the file? The scientific data formats seem complicated.
Posted by 64.215.112.128 on May 23, 2010 at 07:15 AM MDT #
Also what about providing a development kit containing the header files and a binary version of nc-config? That would be very helpful for people that don't want to compile netcdf from source but are developing an app that need to be linked against netcdf. Thanks!
Posted by Herve Pages on June 23, 2010 at 10:09 AM MDT #
We are all in favor of making netCDF development more transparent to the user community!
We have recently moved to an svn server, and you can see all our code here:
https://sub.unidata.ucar.edu/netcdf/trunk/
Most of my development is done in /branches/ed, but everything is merged into the trunk, after passing branch testing.
We (the netCDF C/Fortran/C++ programmers) do not have a bug tracker of any kind. The use of a bug tracker implies a list of bugs to be tracked. Instead, we try to fix as we find, and therefor we have no active bug list. However, other Unidata programmers are currently discussing the idea of having some publically accessible bug lists for our products. So hopefully we will have some nice web-based bug tracking soon.
Posted by Ed Hartnett on November 28, 2010 at 09:09 AM MST #