The nbserverproxy allows Jupyter servers to proxy other ports on localhost to anyone with access to the Jupyter server.

We have a similar problem in dask. People often have access to the dask-scheduler dashboard, but don't have access to each of the dask-worker dashboards. However the dask-worker dashboards are visible to the scheduler.

I'm curious if functionality from nbserverproxy could be adapted to suit Dask's needs as well. There are some similarities

1. In both cases we have a server at an accessible location can see an inaccessible location
2. In both cases that server is a Tornado web server

And some differences

1. The accessible server (scheduler) is likely on a different host than the inaccessible servers (workers)
2. One is Jupyter, the other isn't

cc @yuvipanda, is this worth investigating? Is it worth generalizing nbserverproxy? "Nope, nbserverproxy is pretty tied to Jupyter internals" is a fine answer. This isn't a huge pain point right now, I'm just curious about the possibility.