Description
This is sort of a "meta-issue" - that is, it's not a concrete action, but something we need to resolve overall.
The basic issue is this: right now the astropy.uncertainty machinery is pretty much completely divorced from the nddata uncertainty machinery. That should be resolved. It's not clear we can truly join the machinery - they are quite different in implementation, but there are possible intermediate solutions of making them interact better (or have converter methods).
At creation, I pointing the milestone at 4.0, but that's a milestone for the plan, not necessarily completion.