I think this represents a sensible approach. I think that the name_spacing is implicit and will have to remain so.
I think a strong statement detailing the responsibility placed on the data provider to have checked that the metadata is unambiguously consistent with all standards stated in the 'Conventions' global attribute would be useful alongside any statement that multiple conventions may be defined for a data file
cheers
mark
Dear Mark and Dave
I agree with Dave's answers. If two conventions are used together, it is the
responsibility of the data-writer to guarantee that the metadata supplied is
consistent if there are any overlaps in meaning. A particular case of that is
if the two conventions define attributes with the same names. It has been
suggested that conventions could signal their own name-spaces e.g. CF
attributes could all be prefixed with "cf_" (like the cf_role attribute, which
has been introduced in the new CF section 9). That could help with preventing
collisions of namespaces, but
* it would be cumbersome for writers of files that adhere to only one
convention, which is the usual case, and awkward for programs that read files,
since they would have to check for every attribute by two different names
(with and without the prefix, considering all the data that already exists
without prefixes).
* it doesn't help if the two conventions are inconsistent in their metadata,
whether or not they use similarly named attributes, and this is the more
serious problem, I would argue.
Therefore I don't think this is really a magic solution to get rid of the
potential difficulty. Rather, the writers of conventions have to be aware of
other netCDF conventions that might be used with theirs, and try to use ones
that already exist instead of defining new ones for a given purpose.
Best wishes
Jonathan
_______________________________________________
CF-metadata mailing list
CF-metadata at cgd.ucar.edu
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
Mark,
That is an important question. I suggest that there is only one
practical interpretation for multiple conventions: 100 percent
compiance with the requirements of each listed convention. This leads
to immediate answers for your questions, as follows.
On Wed, Dec 28, 2011 at 9:59 AM, Hedley, Mark
<mark.hedley at metoffice.gov.uk> wrote:
>
> I am interested in the implications for defining multiple conventions in the same file. As a data creator, what am I asserting by defining my file with multiple conventions?
>
> It could be said that the conventions attribute provides an implicit name space for the controlled terms it defines. This enables a data consumer to assign meaning to the terms defined by the convention which exist within a particular file.
>
> If I have one convention, I can pattern match all the attributes and explicitly link them to a convention. All the ones that don't match are user defined, and not part of the convention. Does this scale to having multiple conventions defined?
Yes, assuming that there are no incompatibilities between attributes
in use in any given file.
> Do conventions maintain mutually exclusive vocabularies? I don't think they do.
Agreed. Each shared term must be used in a mutually compatible way,
or else it must not be used in the context of two incompatible
definitions.
> Where vocabularies share terms, is there oversight that ensures that shared terms are defined the same way?
Only good judgement on the part of convention designers. In my
limited experience, there are a few popular conventions that have had
considerable influence on the development of other conventions. This
helps to engender conforming usage of attributes. Such conventions
include Unidata's attribute recommendations (mainly User's Guide
appendix B), COARDS, and CF; perhaps others.
> If I assert that two conventions are being used, does that mean that I have checked that my file contains no attributes which are ambiguously defined?
Not necessarily that checks were actually performed, but only that you
are asserting full compliance with each convention.
> If I want to use a term which is ambiguously defined, can I do this effectively?
By "ambiguously defined" do you mean defined incompatibly between two
conventions? I would say No, this would not be allowed if your goal
is to have the term interpreted unambiguously by all possible readers.
> There seems to be significant potential for confusion here; I think care is required.
Requiring independent conformance with each listed convention should
alleviate confusion.
--Dave
Received on Thu Dec 29 2011 - 06:27:05 GMT
