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/1426#issuecomment-454165639,https://api.github.com/repos/pydata/xarray/issues/1426,454165639,MDEyOklzc3VlQ29tbWVudDQ1NDE2NTYzOQ==,6815844,2019-01-14T21:20:27Z,2019-01-14T21:20:27Z,MEMBER,I'll close this for the recent discussion about MultiIndex,"{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,231308952
https://github.com/pydata/xarray/pull/1426#issuecomment-309268208,https://api.github.com/repos/pydata/xarray/issues/1426,309268208,MDEyOklzc3VlQ29tbWVudDMwOTI2ODIwOA==,6815844,2017-06-18T10:11:33Z,2017-06-18T10:11:33Z,MEMBER,"@benbovy 
Sorry for my late reply.

I think I like your proposal, which bundles multiple concepts in xarray such as `MultiIndex` and `multi-dimensional coordinates` into one, which may result in simpler API.
But actually I don't yet fully imagine how your proposal works with `multi-dimensional coordinates`.
(maybe because I am not accustomed with `multi-dimensional coordinates` very well.)

Currently, 'rasm' example is like
```python
In [1]: import xarray as xr
In [2]: xr.tutorial.load_dataset('rasm', decode_times=False)
Out[2]: 
<xarray.Dataset>
Dimensions:  (time: 36, x: 275, y: 205)
Coordinates:
  * time     (time) float64 7.226e+05 7.226e+05 7.227e+05 7.227e+05 ...
    xc       (y, x) float64 189.2 189.4 189.6 189.7 189.9 190.1 190.2 190.4 ...
    yc       (y, x) float64 16.53 16.78 17.02 17.27 17.51 17.76 18.0 18.25 ...
Dimensions without coordinates: x, y
Data variables:
    Tair     (time, y, x) float64 nan nan nan nan nan nan nan nan nan nan ...
Attributes: 
   ...
```

Does your proposal (automatically) change this like
```python
<xarray.Dataset>
Dimensions:  (time: 36, xy: 56375)
Coordinates:
  * time     (time) float64 7.226e+05 7.226e+05 7.227e+05 7.227e+05 ...
    xc       (xy) float64 189.2 189.0 188.7 188.5 188.2 187.9 187.7 187.4 ...
    yc       (xy) float64 16.53 16.69 16.85 17.01 17.17 17.32 17.48 17.63 ...
  * xy       (xy) SuperIndex
  - x        (xy) int64 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ...
  - y        (xy) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 ...
Data variables:
    Tair     (time, xy) float64 nan nan nan nan nan nan nan nan nan nan nan ...
Attributes:
   ...
```
?

","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,231308952
https://github.com/pydata/xarray/pull/1426#issuecomment-305062228,https://api.github.com/repos/pydata/xarray/issues/1426,305062228,MDEyOklzc3VlQ29tbWVudDMwNTA2MjIyOA==,6815844,2017-05-31T02:12:13Z,2017-05-31T02:12:13Z,MEMBER,"@shoyer 
I personally think 2 is more intuitive for users, 
because it might be difficult to distinguish 
```python
<xarray.Dataset>
Dimensions:  (yx: 6)
Coordinates:
    y        (yx) object 'a' 'a' 'a' 'b' 'b' 'b'
Data variables:
    foo      (yx) int64 1 2 3 4 5 6
```
(which may be generated by indexing from `x` in your example) from
```python
<xarray.Dataset>
Dimensions:  (y: 6)
Coordinates:
 *  y        (y) object 'a' 'a' 'a' 'b' 'b' 'b'
Data variables:
    foo      (y) int64 1 2 3 4 5 6
```
What is the possible confusion if we adopt 2?","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,231308952
https://github.com/pydata/xarray/pull/1426#issuecomment-304902053,https://api.github.com/repos/pydata/xarray/issues/1426,304902053,MDEyOklzc3VlQ29tbWVudDMwNDkwMjA1Mw==,6815844,2017-05-30T14:47:36Z,2017-05-30T14:47:36Z,MEMBER,"@shoyer Thanks for the comment.

> It occurs to me that if we had full support for indexing on coordinate levels, we might not need a notion of a ""MultiIndex"" in the public API at all. 

Actually I am not yet fully comfortable with my implementation, 
and I like your idea as this might be much cleaner and simpler than mine.

If my understanding is correct, does it mean that we will support 
`ds.sel(x='a')`, `ds.isel(x=[0, 1])`  and `ds.mean(dim='x')` with your example data?
Will it raise an Error if `Coordinate` is more than 1 dimensional? 
How about `ds.sel(x='a', y=[1, 2])`?","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,231308952