cboettig
commented
8 years ago @cgreene I'm quite interested in this as well. Berkeley is deploying this at some scale right now in the new data science foundations freshmen course -- the students work entirely on (Rackspace-hosted) Jupyter notebooks for homework & projects (with nbgrader), and the course textbook is also overlayed in a Jupyter notebook. See http://data8.org/ and http://data8.org/text.
It's an interesting model, though it opens several questions for me. Also I coming more from an R background, where Rstudio-server offers some parallels to the notebook but also differences (e.g. full IDE with version control).
I'm interested in both how well this works and how it translates into practices for later courses. How does one incorporate version control into the workflow? (Just terminal sessions running in jupyter?) Do later courses continue to work on hosted environments or perform local installations? Using virtualization (the Rackspace environment is fully dev-op'ed with Ansible & Docker) or native installations? (How) do students (or instructors) learn to extend such environments, or deploy them on the cloud?
I'd love to talk about technologies that facilitate classroom instruction from notebooks. These include Github classroom (https://classroom.github.com/), nbgrader (https://github.com/jupyter/nbgrader), Sage Math Cloud (https://cloud.sagemath.com/), etc.
I would benefit greatly from coming up with a workflow for distributing and rapidly assessing pre-class, in class, and brief homework examples. Bonus points if it provides the opportunity to teach the basis of version control + reproducible scientific discovery.