Jeff Whitaker

Seems like it

There's nothing in the python interface that is specific to CDF5 - it's all done under the hood in the C library. What do you mean that the header contains...

I noticed in your C example you are creating a file using NC_64BIT_OFFSET The problem in python is only triggered with NC_64BIT_DATA (the new CDF5 format).

Are you going to re-open https://github.com/Unidata/netcdf-c/issues/770?

you are correct that netcdf4-python is not parallel capable as of yet. I've never used the fletcher32 functionality, nor do I know of any users who have. It's not high...

There is very little documentation for fletcher32 on the netcdf library side either. I did find this: http://www.unidata.ucar.edu/mailing_lists/archives/netcdf-hdf/2008/msg00004.html and this http://www.unidata.ucar.edu/support/help/MailArchives/netcdf/msg10588.html which suggests that if fletcher32 is enabled, a checksum...

https://github.com/Unidata/netcdf-c/issues/1488

This is not a known problem, and from your description I cannot tell if it is an issue with the server, the C library or the python module. If it's...

The opendap features of netcdf4-python are just a thin wrapper around the functionality of the C library. Basically, we pass either a filename or a URL to the C lib...

Almost certainly a C lib issue. @WardF, should we create a unidata-c ticket?