Lessons learned

Impressions

JupyterHub is a powerful educational tool. With the release of JupyterLab 4 and real-time collaboration at v1.0.0, I can envision JupyterHubs becoming even more useful, for example enabling real-time collaborative technical support, or for small group exercises to be done in a shared notebook.
The CryoCloud JupyterHub performed flawlessly and did everything we need. I was surprised how smooth it was for attendees from around the globe to use QGIS in a virtual desktop environment.
Our attendees were very engaged, and this made the workshop very rewarding and lots of fun to administrate.

The material we developed technically fit in the schedule, but we wanted to provide more time for attendees to do hands-on work. Either increase the time for the material, or cut some material.
When sending 5-minute warnings at the end of breakout group exercises, ask attendees to commit their work to GitHub now, regardless of whether it’s done.
For the Symbolizing datasets together exercise, it would be helpful to include a deliverable in addition to the .qml files. Maybe a screenshot and/or discussion post. This would make it easier for us to review the outcomes from this exercise.
We should have prescribed a more clear structure for file names/structure in git repos. It can be initially hard to tell which notebooks correspond to each exercise in groups’ repos.
Archive all workshop materials after completion. We had a temporary workspace (qgis-data) which had faster storage that allowed e.g., QGIS to more quickly access data in QGreenland that participants used and stored some data on. This directory got cleaned-up after the workshop.

Let people join their own breakout rooms; trying to predict what e-mail people will join Zoom with and pre-configuring breakout rooms turned out to be a time sink. Inevitably, some will join from their phone, not sign in, etc. resulting in spending time manually setting up breakout rooms.
One problem we encountered was copy/paste from the clipboard does not work between the remote desktop and the user’s local desktop environment. In order to get around this limitation, workshop participants resorted to using a text file for transferring text between the remote and local desktop environments. For example, if a user wanted to move text from their local desktop to the CryoCloud desktop environment, they would save that text to a file (e.g., copy.txt) and then transfer that file to CryoCloud (either by uploading the file through the file browser interface in JupyterLab or using JupyterLab’s built-in file editor).

We created run-of-show documents in Google Docs that we could collaboratively edit to manage team checklists and roles.

TODO: Link read-only copies