[CF-metadata] Are There Best Practices for Developing New NetCDF Conventions?
Dear all
Two points about this question:
(1) I agree with Ken that CF is concerned with use metadata, but that there
is overlap between use and discovery metadata. If needs arise for metadata
that CF does not support at present but which could go in CF, do please
consider proposing extensions to CF. As you may know, we are at present
trying to agree definite rules and timescales for making changes to CF, as we
have not up to now had a clear procedure. I can appreciate that it is more
effort to propose and argue for an extension or change to CF than it is to
devise a convention within a smaller community. The discussions that ensue
can definitely be hard work, but I feel that it is worthwhile and leads to a
better result, and I think that the applicability of CF should be broadened
as much as possible. This is because many problems do turn out to be of more
general application than one might initially suspect.
(2) Although a syntax could be introduced to indicate in the names of
attributes which convention defines them (as John suggests), I suspect this
may be too much of an "easy way out". This is rather like discussions we have
had about standard names. So far we have maintained one standard name table
to serve all disciplines. It is more work to choose these names but it means
the names have to be clearly defined, and there is only one name for a given
concept, so it helps data exchange between different user communities. If you
allow conventions to have separate attribute namespaces, there is less
incentive to look at other existing conventions to see whether the same issue
has been addressed by someone else. The result will be the development of
synonyms and mutual unintelligibility, whereas if attributes have common
names across conventions, interoperability is promoted. So I think it would
be good if Unidata maintained a registry of attribute names defined by
known/registered netCDF conventions. This will assist in different conventions
sharing attributes, such as we already do with those defined by Unidata like
_FillValue and Conventions.
Cheers
Jonathan
Received on Thu Jul 12 2007 - 08:13:09 BST
This archive was generated by hypermail 2.3.0
: Tue Sep 13 2022 - 23:02:40 BST