Open danielballan opened 4 years ago
It's to keep them consistent with the NSLS-II Jupyterhub.
Good point. We have dug ourselves into a bad hole here. Our deployment is based on the flawed assumption that you can use the same front-end code (as specified by the jupyter server environment on the jupyterhub
VM) with any back-end code (as specified by a given kernel on beamline machine).
As long as beamlines sometimes rely on very old kernels, we are stuck with very old front-end code to remain compatible, in particular with interactive matplotlib figures.
The new model that selects between images rather than kernels keeps front-end and back-end in sync. For beamlines that run their our jupyter servers (e.g. LIX) we may need to create a special case so that they can leverage the latest features in ipympl.
The current pins are very old, and ipympl in particular has improved a lot.