File synchronization ideas for remote jupyter server

[rchiodo]

Solves this: #1601

3 options I can think of

When connecting, ask user if they want to

Sync

Sync workspace to remote machine. This might use a zip or just the jupyter api to do one file at a time. Might be really slow.

Roll their own

Do their own syncing/Provide a root on the remote machine to run their kernels from. User provides a file path and we use that as the starting directory for a kernel.

List files in a separate panel

We add a separate entry in the file panel (like Don did with his extension before) that just shows the files on the remote box. User can open them and interact with them just like normal files.

This has the added benefit of you don't need to open a folder or anything in vscode.dev.

All 3?

We could ask any or all three of these. Or add them one at a time.

Might also be other options I'm not thinking of.

[greazer]

FYI, Shantnu did a bunch of investigation and prototyping for the Azure ML extension about a year ago. Here's what he recalls:

The file system was working fine using the VSCode's FileSystemProvider API. We just implemented a translation layer between the Jupyter's File System APIs and the VSCode's File System Provider. It was a very thin layer mostly just restructuring objects and calling the right functions

The terminal also worked using the VSCode's Pseudo terminal API. However, there were some problems with line breaks, something I think could be easily overcome if we would have investigated more.

Here's the protoype branch: https://github.com/microsoft/vscode-azureml/tree/shsuman/remoteFSAndTerminal,
Here's the folder that contains the entire code for this protoype: https://github.com/microsoft/vscode-azureml/tree/shsuman/remoteFSAndTerminal/src/azureML-remote/providers