Some way to try it out Oscar without installing Oscar

[thofma]

Just so that we don't lost track of it. It would be cool to have a way to try out Oscar without installing Oscar. The canonical candidate would be to have a way to spin up an instance of a jupyter notebook.

• If the user has a google account, then google colab is an option. See for example https://colab.research.google.com/github/cs231n/cs231n.github.io/blob/master/python-colab.ipynb
• @benlorenz mentioned https://github.com/github/codespaces-jupyter (but I don't know how this looks in action).
• @aaruni96 will check whether there is something provided by mardi.
• something something binder

[benlorenz]

With github codespaces, and a github account, you can fire up vscode in the browser with the Oscar.jl repo and a terminal where you can use Oscar (from the Code button on the Oscar repo). This currently seems to use julia 1.8.4, I think we should update the .devcontainer/devcontainer.json.

It takes quite a while to ]up and precompile Oscar, maybe we can install a precompiled Oscar via maps? Also Oscar fails to load with the recog error due to missing software.
After manually installing build-essential I was finally able to start Oscar in the terminal:

This does work (without installing oscar locally) but is not really quick and nor easy to use I would say.

There is also a open in jupyterlab button for the codespace but that doesn't seem to work since jupyter is probably not installed in the container.

[aaruni96]

At NFDI jupyter hub, one can sign with a number of third party providers (Google / Orcid / github / university / many many others), and then run a jupyter notebook online.

Separately, these can be somehow (how exactly?) configured to provide a docker image based on a dockerfile : https://repo2docker.readthedocs.io/en/latest/config_files.html#dockerfile-advanced-environments .

So, we could have a docker file which specifies installing all the tools and precompiling Oscar, such that when jupyterhub starts, users can just do using Oscar, and don't need for the lengthy pre compile step to happen on the restricted resources available online (2GB RAM, 1 vCPU by default. 4 GB RAM, 1 vCPU on the "big" option).