Re: [thredds] GRIB variable name changes in 4.3

On 2/27/2012 5:28 PM, Benno Blumenthal wrote:
I have suggested a GRIB table registry in the past, not that I have gotten much enthusiasm in response.

I still think the registry is essential, however, enthusiasm or not.

While you are correct that your name conversion is problematic, as users of GRIB files we need to really solve the problem of what is in a particular GRIB variable. You could easily, for example, produce the wrong long_name if you don't really know what is in the variable, which will mislead the users. And ultimately the user needs to know what is there, particularly if it is hard to find out.

I know how hard it is to get GRIB variables right. So I understand why you want to change. I just wonder if we can address the underlying problem.

Benno


Hi Benno:

Yes, you go to the heart of the matter. How does one know what is in the variable? GRIB tables are very problematic, and they have been effectively mismanaged by the WMO community. A registry is needed, as well as a comittment by data providers to use it properly. In case others have not read my proposal, its mostly contained here:

      http://www.unidata.ucar.edu/staff/caron/papers/GRIBarchivals.pdf

Benno, if you have something written, can you post? I hope that everyone who has a stake in this will take the time to understand the issues involved.

John