On 1/31/13 12:58 PM, Schultz, Martin wrote:
> Hi Phil,
>
> while I agree with you that the only clean solution would have to be
> defined in the netcdf API itself (not in CF), I don't think that
> groups are what we are looking for here. This namespace thing is
> entirely related to attributes, while groups are a variable concept.
> What would help were "hierarchical attributes", so that one could
> define metadata "groups" just as in an XML file. Possibly something
> like
>
> float myvar(dim1, dim2)
> cf_attributes:
> standard_name: "..."
> units: "..."
> OSD_attributes:
> measurement_principle: "..."
>
> etc.pp. -- only this would make reading netcdf files a wee bit more
complicated ;-(
> ...
This is an intriguing idea. It could be streamlined if we were willing
to limit the selection of attributes from different standards so that
the source of any given term would be the same throughout a file.
A global attribute could contain a map to the sources of all the terms -
actually this would probably not be needed for all the attributes, just
ones that need to have their name space identified.
In a file that follows CF, UDD and OceanSITES, it would be something like:
// global attributes:
:Conventions = "CF 1.4, OTS 1.2, UDD 1.0" ;
:Metadata_Conventions = "Unidata Dataset Discovery v1.0";
:format_version = "OceanSITES 1.2?.
:Attribute_Conventions =
"CF: standard_name coordinates axis ancillary_variables flag_values
flag_meanings
UDD:naming_authority creator_name publisher_name license
geospatial_lat_* geospatial_lon_* time_coverage_*
OTS: site_code platform_code format_version source data_mode references
area data_assembly_center
update_interval uncertainty accuracy precision resolution ;
:acknowledgement = '....'
:naming_authority = '...'
:geospatial_lat_min= '...'
:geospatial_lat_max= '...'
:geospatial_lat_units= '...'
That long string called attribute_conventions would make it clear
where to look for the definition of any attribute, in cases where there
might be ambiguity. It could also encompass global attributes, which
are at least as likely to collide as variable attributes.
It wouldn't let you use a term with one definition on one variable
and another definition somewhere else - would that be a problem
for aggregation software? It seems like a sensible limitation.
Best of all it could be completely ignored by existing software with
no problem, until it's needed. It could be a low cost insurance policy
against future development of terms in CF or other standards.
Regards - Nan
>> Perhaps the only solution lies in using netCDF-4's existing namespacing
>> mechanism, namely groups (which idea has been proposed before). I
>> suspect this too will have it's adherents and detractors!
>
--
*******************************************************
* Nan Galbraith (508) 289-2444 *
* Upper Ocean Processes Group Mail Stop 29 *
* Woods Hole Oceanographic Institution *
* Woods Hole, MA 02543 *
*******************************************************
Received on Mon Feb 04 2013 - 12:07:52 GMT
