Hi,
I got this idea that the python{2,3}-ipykernel packages could recommend some modules useful to use in the Jupyter Notebook.
Recommends: * numpy * matplotlib
Not sure if should be recommended or suggested: * pillow * scikit-image * scikit-learn * scipy * statsmodels * sympy * pandas
What do you think? It would mean that a user who dnf installs notebook gets some opt-outable things useful in the notebook. And who uses the Python 2 kernel, will get the Python 2 version.
This would also drag the things to the Python Classroom Notebook which I think is good - it will make the Notebook much easier to use for teaching Python for science - without the need to start with virtualenvs first or install additional packages. But of course it would also make the image bigger.
Also see [1] for relevant issue just for the Python Classroom Lab.
[1] https://github.com/fedora-python/fedora-kickstarts-python-classroom/issues/1