[CF-metadata] Conventions vs. Community Profiles

[John Caron]

Derrick Snowden wrote:
> Hello,
>
> I am interested in creating a new file format for use in an 
> operational data collection scheme.  The file will contain XBT 
> profiles collected under the auspices of the JCOMM Ship Observations 
> Team so will have many international users.  I'd like to use CF 
> conventions but I'm struggling with the difference between a 
> "convention" and a "profile".  As I understand it, CF provides 
> guidance for how to encode data.  I.e. given a certain data structure, 
> CF shows how that can be encoded in an efficient and hopefully 
> interoperable way.  As this new XBT file will be used by a large 
> community with common goals we would like to recommend/require the use 
> of certain attributes and variables.  This strikes me as more of a 
> community content profile similar to the way ISO19115 provides the 
> structure of a metadata document while the North American Profile 
> recommends the usage of certain elements.  My question is, how does 
> one go about developing a community profile?  Does this end up being 
> another convention or is it distinct in it's representation in the 
> file.  For example, would the global attributes look like
>
> .Conventions = "CF-1.4, XBT-1.0" or
>
> .Conventions = "CF-1.4"
> .Profile = "XBT-1.0" or .Community_Profile or whatever?
>
> Clearly anything can be done, my question is does anyone care to 
> comment on what should be done?  I know other communities such as Argo 
> and OceanSITES have made some choices but I'm not sure if they're the 
> right ones.
>
> Thanks for your time.
>
> Regards,
> Derrick
>
> _______________________________________________
> CF-metadata mailing list
> CF-metadata at cgd.ucar.edu
> http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
>   
Hi Derrick:

CF conventions are a way to describe the semantics of data encoded in 
netcdf files. The current set of conventions is mostly oriented towards 
gridded data, such as model output. Some new proposals for observational 
data are in the pipeline, and you might want to have a look at them:

  https://cf-pcmdi.llnl.gov/trac/wiki/PointObservationConventions

Generally, netCDF Conventions describe the possible ways data can be 
stored (eg variables and dimensions) , and the required metadata for 
interoperability (eg attributes). Typically a data provider for a 
particular dataset chooses a particular storage method and metadata, 
often represented as a CDL template. This can be considered as 
describing a subset of the universe of netcdf files, which is what i 
think "profile" means.

The CF Conventions try to constrain the storage choices to a reasonable 
balance between efficiency for the writers and interoperability for the 
readers. We havent seen too much profiling within it, and the less the 
better. However, as the CF specification grows and covers more types of 
data, no doubt we will see some of this.

With regard to

  Conventions = "CF-1.4, XBT-1.0"

I'd ask, what semantics does "XBT-1.0" imply that "CF-1.4" doesnt ? 
Because something not  in CF means that its not interoperable with CF 
clients.