Hi Jonathan,
I think you present a nice summary of the pros and cons of formula_terms
vs. ancillary_variables with standard name modifiers. I especially agree
with the following:
> Either mechanism could be
> used in practice. Neither is exactly analogous to what we need, which is to
> record parameters (global or a function of independent coordinates) that are
> needed to define the quantity in the variable.
If forced to chose between these options you've convinced me that
formula_terms is more appropriate. But I feel it would be better to define
a new attribute, or possibly modify (again) the standard name string to
point to additional metadata required to unambiguously define a quantity.
This issue seems too important to try and force a solution that doesn't
quite fit into the existing attributes. It's important for ease of use
that the attributes have clearly defined purposes. I find having the
purpose of an attribute depend on its context (e.g., attached to a
dimensionless vertical coordinate, or not) to be confusing.
Brian
On Mon, Oct 25, 2004 at 08:55:38AM +0100, Jonathan Gregory wrote:
> Dear Derrick
>
> > This could be solved but next, we'll see neutral
> > density (commonly called gamma) which is defined with respect to an
> > independent set of 3d data, not just a formula.
>
> In a similar way, formula_terms can point to a 2D field of pressure data in
> order to calculate p from a sigma coordinate variable. However, a standard
> name modifier could also indicate a field of data. Either mechanism could be
> used in practice. Neither is exactly analogous to what we need, which is to
> record parameters (global or a function of independent coordinates) that are
> needed to define the quantity in the variable. By contrast, standard name
> modifiers supply metadata about the values of the variable, while formula terms
> supply parameters needed to compute other quantities from the values in the
> variable.
>
> I think I prefer formula_terms because formula_terms are defined explicitly
> for each standard name i.e. what terms it can have and what they mean. By
> contrast, standard name modifiers are defined in a way which is not specific to
> the standard names with which they might be used. Hence I think that with
> formula_terms it will be easier to make the dataset more self-describing,
> unless we substantially depart from what I understood as the intention of
> standard name modifiers. In answer to
>
> > where is the practical division between relying on a
> > user's intuition or knowledge and creating "fully self describing" data sets?
>
> I think we should rely as little as possible on intuition.
>
> Cheers
>
> Jonathan
> _______________________________________________
> CF-metadata mailing list
> CF-metadata at cgd.ucar.edu
> http://www.cgd.ucar.edu/mailman/listinfo/cf-metadata
Received on Tue Oct 26 2004 - 11:24:43 BST
