Sorry for the delayed reply, I've just gotten back from sea and have not
had internet access for most of the past month.
I'd be happy with application/x-netcdf; once your mime software knows you
have a netcdf entity, you can detect that it's cf with whatever code
you use to
read netcdf.
> However, netCDF can be used by distant discipline or different data
> model, and it is likely that different software is preffered for
> different conventions.
Including this in the mime media type descriptor is redundant, and redundant
information always seems to cause problems. It also introduces the question
of how to decide where to stop, as far as defining tags for different
implementations
of netcdf.
If it's really necessary to include the cf tag, I'd prefer
application/x-netcdf-cf,
since netcdf is really the primary media type. Among users of all the many
kinds of mime types, cf is likely to be less well recognized than netcdf.
My 2 cents -
Nan
> Dear Colleagues,
>
> I have a question: what should be MIME media type for CF-netCDF data?
>
> This issue was originated in the GALEON IE project,
> but I think it is better to consult entire CF community before taking
> much action.
>
> There seems to be several solutions.
> Examples begin with "x-" because there seems to be no urgent need for
> IANA registration.
>
> 1) application/x-netcdf; profile=cf-1.3
> - The idea is "netCDF is a format, CF is convention on the format."
> - Some implementation of MIME dispatcher doesn't relate applications
> to parameters.
>
> 2) application/x-cf+netcdf
> - The idea is "netCDF to CF is XML to XHTML."
> - CF and other netCDFs can be distinguished even if parameter is ignored
> - Some doubt necessity for creating "+netcdf" convention.
>
> 3) application/x-cf-netcdf
> - Improved version of 2), without controversial "+netcdf" convention.
> - We can tell CF and other netCDFs, too.
> - Similar example: image/tiff and image/tiff-fx
>
> I firstly suggested 2), but now I'm inclined to support 3).
>
> MIME media type is originally created for dispatcher software such as
> web browsers or email user agent (or file type association of explorer
> of MS Windows).
> Unfortunately many implementations cannot handle parameters of media type.
> So we cannot associate different software to netcdf dataset with
> different conventions,
> if we choose solutions with parameters (such as 1)).
>
> That might be no problem for most of the CF community,
> where most of data is in regular spatiotemporal grid and can be
> handled by one smart software such as IDV.
> However, netCDF can be used by distant discipline or different data model,
> and it is likely that different software is preffered for different conventions.
>
> I would appreciate much if I could have any frank response....
>
> Regards,
>
--
*******************************************************
* Nan Galbraith (508) 289-2444 *
* Upper Ocean Processes Group Mail Stop 29 *
* Woods Hole Oceanographic Institution *
* Woods Hole, MA 02543 *
*******************************************************
Received on Mon Aug 04 2008 - 12:32:29 BST
