html_url,issue_url,id,node_id,user,created_at,updated_at,author_association,body,reactions,performed_via_github_app,issue
https://github.com/pydata/xarray/pull/2844#issuecomment-761843786,https://api.github.com/repos/pydata/xarray/issues/2844,761843786,MDEyOklzc3VlQ29tbWVudDc2MTg0Mzc4Ng==,3460034,2021-01-17T16:58:34Z,2021-01-17T16:58:34Z,CONTRIBUTOR,"I unfortunately have not been following along with the recent developments in this PR, so these may have already been previously covered. Sorry if that is the case!

However, earlier on in the development of this PR, there were some substantial concerns about backwards compatibility (e.g., for libraries like MetPy that currently rely on `grid_mapping` and the like being in `attrs`) and the improved propagation of `encoding` through operations (so that moving these to `encoding` doesn't mean they are unnecessarily lost). What is the current status with regards to these?","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,424265093
https://github.com/pydata/xarray/pull/2844#issuecomment-671094846,https://api.github.com/repos/pydata/xarray/issues/2844,671094846,MDEyOklzc3VlQ29tbWVudDY3MTA5NDg0Ng==,3460034,2020-08-09T20:03:09Z,2020-08-09T20:03:09Z,CONTRIBUTOR,"For what it's worth, here's a bit of an elaboration on my view from https://github.com/pydata/xarray/issues/4215#issuecomment-656789017 and https://github.com/pydata/xarray/issues/4121#issuecomment-639758583:

`ancillary_variables` really fit a role of ""non-coordinate linked variables,"" which doesn't have a good place in xarray's current data model. Both the current behavior (any links are completely lost if not manually handled) and linking as coordinates (conceptually confusing and [causing spurious proliferation](https://github.com/pydata/xarray/pull/2844#issuecomment-670808763)) are problematic. Short of a ""linked"" or ""subsidiary"" variable concept being added to xarray's data model (could be worth discussing on a new issue?), `cf-xarray` seems like the best place for managing `ancillary_variables`, either through
- A method that extracts `ancilliary_variables` from a Dataset given a variable name (https://github.com/xarray-contrib/cf-xarray/issues/5#issue-628569748)
- Parsing the `ancilliary_variables` attribute and replacing the string with a Dataset or list of DataArrays or Variables containing the actual ancillary variables (https://github.com/pydata/xarray/issues/4121#issue-630573329 and https://github.com/pydata/xarray/pull/2844#issuecomment-670991386)
- A method or addition to `__getitem__` on the accessor returning a Dataset with the linked variables included (https://github.com/pydata/xarray/pull/2844#issuecomment-670996109)

For those interested, the issue for `ancillary_variables` in cf-xarray is https://github.com/xarray-contrib/cf-xarray/issues/5.","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,424265093