[CF-metadata] Thoughts about CF future

[John Caron]

Since I am not able to come to the GO-ESSP meeting to discuss these 
issues in person, I wanted to lay out my thoughts about where CF could 
go in the future. These are personal opininions, and not in any way 
official Unidata positions.

We (Unidata) have been pushing CF as the recommended way to write 
gridded files to anyone who asks. We have had a lot of success, 
including with ESRI who are working on reading CF compliant netdf files 
into ArcView. CF has really done a great job of clarifying the semantics 
of data.  In my view, the current edges of CF that need work are:

    1. staggered grids, eg as output by the WRF model.
    2. need to specify very detailed info as used by GIS, for example, 
ellipsoids and datums etc.
    3. ongoing need to add new projections and vertical coordinate types
    4. ongoing need to develop standard names and hook these into other 
ontologies

My primary concern is creating a process for evolving CF, what Bryan 
called "governance". I would like to see a formal decision making group 
be identified, with super-majority but not necessarily unanimous consent 
required. I would also like to have ad-hoc subgroups form to work on 
special topics, for example "Regional/local models" and "GIS". These 
subgroups would work out a concrete recommendation, then submit to the 
entire group for feedback and eventually a decision by the governance group.

Another issue that has recently come up: CF has a small section about 
both Station and Trajectory data, that is good as far as it goes, but 
needs more semantics to be complete. I would recommend that CF restrict 
itself to gridded data, and withdraw the Station and Trajectory data 
sections. However, if there is enough interest, i would recommend 
forming a subgroup that would finish specifying it.

There is also development at Unidata that effects CF, most notably the 
"coordinate system" object model, prototyped in the Netcdf-Java library 
(which we may also add to the Netcdf-4 library). This parses CF 
conventions (among others) and presents a unified coordinate system API 
to the user. We also are working on "scientific data type" APIs that so 
far include Grid, Point, Station, Trajectory and Radial data types. This 
work is not as far along as the coordinate systems, but as part of it we 
have been evolving our view of what makes good file formats in those cases.

Whatever is decided, we will continue to support CF as an important 
semantic description of data. I appreciate all the time and intelligence 
this group continues to provide.