Fri Jan 24 15:29:55 EST 2020

Heterogeneous deployment: chain of caches

This might be a very special case, but it seems to work alright:

- distributed highly-available store that gets resynchronized get, put
  and reconnect.  this level is multi-directional (each node can

- in parallel, a read-only content-addressable store.  can be
  referenced by the objects in the former, but is lazy.

- local updates to store can be watched, and can cause further
  unidirectional propagation, e.g. load from CAS, then wait for
  locally managed nodes to appear for deployment.

For bluepill, I've noticed that running a trans-atlantic GDB RSP
session is not a great idea.  Move the file first, then run gdb