[Discuss] Depend on jupyter-client provisioners

[echarles]

There has been a few discussion on this issue jupyter-server/jupyter_server#112 (comment) with @chbrandt @nreith and @kevin-bates to identify ways to use the jupyter-client provisioning sytem to offer conda envs to the notebook users.

This issue aims to continue that discussion.

[chbrandt]

As per discussion in the other issue (referred above), I noticed nb_conda_kernels is using the "providers" endpoint of jupyter-client:

nb_conda_kernels has entrypoint at jupyter_client.kernel_providers, whereas jupyter_client expects an entry at jupyter_client.kernel_provisioners. There is no reference to "kernel_providers" in the docs anymore (https://jupyter-client.readthedocs.io/en/latest/search.html?q=kernel_providers). In nb_conda_kernels's discovery.py module says "Initial support for the kernel provider mechanism to be introduced in jupyter_client 6.0".

• jupyter_client.kernel_providers: https://github.com/Anaconda-Platform/nb_conda_kernels/blob/f461d2159a970fec323f0c225557e74c63c6a45b/setup.py#L16
• jupyter_client.kernel_provisioners: https://jupyter-client.readthedocs.io/en/latest/provisioning.html#discovery
• discovery.py: https://github.com/Anaconda-Platform/nb_conda_kernels/blob/f461d2159a970fec323f0c225557e74c63c6a45b/nb_conda_kernels/discovery.py#L2

[mcg1969]

#255 removes the unused discovery module. It was added in anticipation of a discovery mechanism that was planned for 60 but pulled before release.

[waaffles]

Hey all I'm resurrecting this discussion as I'm unsure on where exactly this all left off as far as

ways to use the jupyter-client provisioning sytem to offer conda envs to the notebook users