NetCDF-C discussion of dimension scope/visibility in a group hierarchy

[ethanrd]

There was a change back in 2021 to how netCDF-C handles dimension scope/visibility in a group hierarchy. It was a change to the NUG data model description that was not discussed widely or propagated to the netCDF-Java library. We are considering reverting back to the historic NUG description that dimensions can only be used by variables in the group where the dimension is defined or child groups.

If anyone has comments or has used the new dimension scoping allowed in netCDF-C, please go to netCDF-C issue #3157 "Visibility (scope) of dimensions in a group hierarchy".

[pvanlaake]

This is a problem that would only pop up if lateral search is necessary to locate dimensions referenced in variables. The use of lateral search is already discouraged for new data sets:

They are allowed mainly for backwards-compatibility with existing datasets, and may be deprecated in future versions of the standard.

Perhaps CF-1.13 is the future?

That may also be an opportunity to flag the other issues that were brought up in #333, including a number of outright errors (like the quote from NUG). There was general support among the participants in that discussion to do something with section 2.7, and outside such as in #420, but the PR fizzled out with one sceptical voice. Should that effort be picked up again?