Add a remote kernel powered by Binder about jupyterlite

[psychemedia]

This would be interesting... for simple jobs, things could be run in the browser. If a more complex computer backend is required, then a remote kernel could do the work. (Note that in the general case, it would also be handy to connect to a local (desktop) kernel. Cf. how VS Code Jupyter extension lets you provide a kernel connection URL and run using that.

I also note a VS Code extension for working with pyolite that claims to build on jupyterlite - not sure how that manages the connection.

[jtpio]

Thanks @psychemedia.

Ideally some parts from JupyterLite would go to upstream JupyterLab, which might require some refactoring of the underlying @jupyterlab/services package. This could allow having in-browser kernels in a normal JupyterLab installation too (hybrid).

Otherwise for this specific issue, it will probably be implemented as a third-party serverlite extension once #104 has landed.

[psychemedia]

Just to generalise this a bit more, it would be be useful to connect to start an arbitrary remote kernel, eg via BinderHub, a user's account on a JupyterHub server somewhere, or via a Jupyter Enterprise Gateway.

The user could then bring their own client simply via a browser (eg on a Chromebook, tablet or phone) and access arbitrary hosted Jupyter served computational environments.

[bollwyvl]

As JupyterLite is now being used to take pressure off binder, I think we probably won't be doing this.

[psychemedia]

That is a separate issue, surely?

[jtpio]

Probably could be converted to a discussion in the "Ideas" category.