[CF-metadata] CF Conventions 1.2

From: John Caron <caron>
Date: Thu, 08 May 2008 09:23:10 -0600

Jonathan Gregory wrote:
> Dear Phil
>
>> Perhaps the recent
>> flurry of CF proposal activity in part reflects a general desire to
>> 'play catch-up'.
>
> Yes, I think that is the case. It certainly is the case for the two proposals
> I have made, on the axis and cell_methods attributes. These were discussed on
> the email list and in abeyance for a long time because we had no way to adopt
> them formally until we agreed the new rules.
>
>> I can produce some simple test files for the changes at CF 1.2. But
>> the question of what constitutes application conformance is, I
>> suggest, not easily defined. For instance, I could create a noddy
>> netcdf file with two new grid mapping attributes, as follows:
>
> Yes, I think such a file would be useful, because it does at least provide
> input data that the cf-checker can check for conformance, and other
> applications could likewise check that they can read in and interpret, if they
> are interested in these features. I agree with you that what "compliance"
> actually means for an application is ill-defined. This is an issue which has
> come up before, of course. Since most of CF is optional, in one sense (but not
> a very useful sense) an application is compliant even if it ignores all that
> optional metadata. On the other hand I am sure no application currently exists
> which interprets all the metadata. But I don't think that means the metadata is
> not useful. It can still be read by humans, it describes the data properly,
> and we only add features when people have a need for them (usually people who
> intend to produce data).

As a tool developer, a real netcdf file that has the new features(s) in it is extremely useful. In
fact I dont even try to implement a feature until I have a real example of it.

So on a practical level, requiring an example netcdf file before the final acceptance of a feature
seems to me to be reasonable. Proving that software "correctly conforms" is difficult in the general
case.

We have a repository at Unidata of sample CF files, but they dont document which features they use.
It would be very useful to start that documentation, and tie in back to CF section numbers or anchors.

I propose we start a repository of sample files, ideally on the CF site, documented as to what CF
features they use. It would be good if that documentation is a wiki (or equivilent), so that the
initial person can make a start, then others can augment and comment on.
Received on Thu May 08 2008 - 09:23:10 BST