IMO a bunch of CF describes how to encode complex types in the simple types provided by netCDF-C (e.g. geometries, datetimes, etc). Zarr doesn't have this limitation. I think this convention should move the world toward directly embedding complex types in datasets
IMO a bunch of CF describes how to encode complex types in the simple types provided by netCDF-C (e.g. geometries, datetimes, etc). Zarr doesn't have this limitation. I think this convention should move the world toward directly embedding complex types in datasets