Hi,
I haven't followed this discussion very closely, but have a few
comments on one of Jonathan's suggestions.
> If we include data quality variables, we are dealing with more than just
> "errors". I would suggest calling the pointer attribute "ancillary_data"
> rather than "error_variables".
If an attribute contains a list of netCDF variables, I suggest it's
better to use "variables" in the attribute name, because otherwise
it's just a string with no indication it is intended to refer to other
variables. So I would much prefer "ancillary_variables" to
"ancillary_data". Similarly, if an attribute contains a list of other
attribute or dimension names, using a suffix of "_attributes" or
"_dimensions" in the attribute name makes the intention clear. If an
attribute can only name one variable, dimension, or attribute, then
you could use "_variable", "_dimension", "_attribute" to indicate
the intention that only one component is named, as in one-to-one or
many-to-one mappings.
I suspect this meta-convention hasn't been followed in all the
coordinate variables conventions that have been proposed (although
"_coordinates" is a good prefix, since it can name variables or
dimensions), but if followed, it would permit extra checking analogous
to what declarations in programming languages provide, that a variable
name in such a list exists. And it would make clear that such an
attribute exists for structural purposes, to establish additional
relations among the components in a dataset. It could even be used to
prevent components becoming separated from closely associated
components when processing splits a file, since it would permit
automatic determination of closely associated variables.
> * Give the ancillary variables a standard_name prefix of "ancillary_data_for_".
> They should also have all the metadata of their parent variables, so they are
> independently fully self-described.
Instead of prefixing a variable name with "ancillary_data_for", I
would instead recommend using an attribute named something like
"for_variable" or "for_variables" that names the variable or variables
to which the information applies.
--Russ
_____________________________________________________________________
Russ Rew UCAR Unidata Program
russ at unidata.ucar.edu
http://my.unidata.ucar.edu
Received on Tue Apr 08 2003 - 08:55:11 BST
