Hi Ronny,
> I am an undergrad student doing some research on netcdf.
> A question I couldn't get a definite answer to, from docs and the interwebs,
> is: "Do conventions (CF / any) apply to the new netcdf4-model in exactly the
> same way they do for the older model, or are there differences to consider?"
>
> with kind regard
> Ronny
>
> ps: Please Don't forget to use reply-all, since I am not subscribed to the
> list.
That's a good question. There is a draft document on developing
conventions for netCDF-4 here:
http://www.unidata.ucar.edu/netcdf/papers/nc4_conventions.html
One way of using netCDF-4 is to continue to use the classic data model
of netCDF-3, taking advantage of the performance-oriented features of
netCDF-4, such as compression, chunking, and efficient schema changes,
but ignoring new features of the enhanced data model such as compound
types and groups. In that case any conventions appropriate for netCDF-3
apply unchanged, because the conventions apply to the data model rather
than its encoding in a specific data format.
Additional conventions will be required for new features of the netCDF-4
enhanced data model. For example, a convention for how to assign
attributes to members of compound types has been developed and described
in the draft document referenced above. Others will result from greater
experience using the new features.
There's still a "chicken and egg" problem. Data providers won't make
much use of the new data model features until applications and utilities
have been adapted to handle them. Developers of conventions won't
propose best practices for handling the new features until there is
experience with real data. Developers of applications and utilities
won't upgrade their software until there are conventions that make it
practical to interpret the semantics of the new features. There is slow
progress in all these areas, but such transitions take time.
--Russ
