[CF-metadata] CF provisional standards

From: Jonathan Gregory <j.m.gregory>
Date: Fri, 17 Nov 2006 13:17:19 +0000

Dear Bryan

> That's not to say we can't have a provisional copy of the convention/
> standard, which represents best practise until the next "official" version.

Indeed, that is what we have always done, and we should continue to do it.

> Any other modifications should be provisional and go through the
> procedure outlined in the white paper, which has been agreed!

In principle, yes, but we have not agreed, or even discussed, the timescales
for making changes. If the timescales are rapid, I think it's fine. My point
is that it does not serve the community well if our decision-making processes
are slow. Jamie's realization convention was proposed months ago. The design
of the forecast/analysis time convention, initiated by Ag, took place years
ago. The latter at least did reach a resolution which I would regard as a
decision, but no-one had time to write it into the standard.

> > If the CF process takes a year to decide something,
> > it is no use to turn round and say to someone, "Sorry, that was provisional,
> > your data is now deemed to be all wrong."
>
> In that situation all we are saying is that - nope, yours is the only
> code that can understand that data! If you want others to read it,
> without writing new code, you may need to rewrite it! That's fair!

No, I don't think it's fair. It is imposing a burden on people to redo what
they have done, because the CF community was too slow in making a decision.

> Actually, given I'm not aware of *one* codebase that supports *all* of
> CF, all experience tells us is that CF is "too complex" or "not worth
> implementing yet in its entirety". The lack of a complete implementation
> argues against your point!

I disagree. We have only discussed and agreed things when people said they
needed them. I think it is unlikely that there is any single codebase that
supports all of it, but I expect that nearly all of it has been used some-
where to *write* data at least, if not to read it. Because data exists for
longer and is harder to rewrite than software, we have to be very cautious
about changing a convention once it has been agreed.

> I think I'm actually arguing for finding a CF version 0.9 which is
> somewhere between COARDS and CF 1.0, but which at least some folk have
> completely implemented).

I don't think we should do that. We have more than enough to do with new
conventions without revisiting all that we have previously decided over many
years, and which is now in use.

Best wishes

Jonathan
Received on Fri Nov 17 2006 - 06:17:19 GMT