[WIP] Multiple ContentsManagers

[jasongrout]

@danielballan and I were talking about some usecases he has for Jupyter, and one that came up is that he'd like to provide two different types of contents managers (perhaps one as a server extension). We came up with the following back-of-the-envelope list of bullet points which we can flesh out:

• jupyter-js-services allow to configure the contents manager api endpoint
• The services object allowing to configure more than one contents manager

In JupyterLab, having the file manager be associated with a separate contents manager.

Also, a contents manager supporting getting a slice of a file would be great too.

I'll flesh out more later - just recording the short list of abbreviated bullet points now...

[blink1073]

I don't inderstand the need for multiple instances here, why not override the server end instance with an extension to provide the extra capabilities?

[danielballan]

@blink1073 Good question. Overriding the existing instance is a reasonable alternative, but I think two separate managers is a cleaner solution to this particular use case. We have a database of "experiments" of different kinds that can be represented well by a ContentsManager. Mixing them in one manager seems confusing to me. On the client side, I want a separate "Experiment Browser" to make it clear that these experiments are files at some special mount point. On the server side, experiments have different rules that files (always read-only, for example).

[jasongrout]

@blink1073 - we were playing with the idea of having two contents manager (each backed by different storage) and different file browsers for each system.

[blink1073]

That's fair, thanks for clarifying.