Design of module communication

[freeman-lab]

Opening for discussion...

In writing the CLI and implementing a couple of the submodules (as described jupyter/enhancement-proposals#5), there's a choice to be made about how best to interface with the various components.

These two drafts of the CLI reflect two choices, curious to get feedback: #1 #2

The question is whether we have a top-level CLI parse options / arguments and make POSTS/GETs to the various servers that constitute the system (a la #1), or should each submodule define its own option / argument handling and treat the top-level CLI more like a metamodule (#2).

Although it's somewhat cosmetic, it does affect how we define interfaces for the submodules, so maybe worth deciding early.

cc @rgbkrk

[rgbkrk]

I'd just make a top-level CLI and we'll hack on it together, adding features as we go along.