Closed rabernat closed 6 years ago
gcsfs is now at 0.1.2, may be no reason to install from pip rather than conda.
I suggest prototyping versions using our binder instance. If you run through the examples using unpinned versions (), you'll see a few issues that have snuck in over the past few months. We can pin versions in the environment yaml if we want to. We can also use the docker image built by binder/repo2docker directly in our jupyterhub deployment.
I would like to get some of the recent improvements to xarray and dask into pangeo.pydata.org.
Does anyone see any obvious conflicts in what I have proposed here?
Note: I don't propose to merge this before the tutorials @jhamman and I are doing this week. Just starting the discussion.