In the previous section, we saw that the concurrent.futures package makes it very simple to farm out computational jobs to worker processes. If the program we need doesn't fit into the send out jobs and collect the results model, we're probably better off working at a somewhat lower level of abstraction.
So, let's now move on to look at another package that helps us handle multiprocess programs that don't fit that model, but the pieces are only partly independent of each other. From time to time, they need to pass information between themselves, not just back to the controlling process. We can't do that with concurrent.futures because it just doesn't fit into the model that concurrent.futures uses to describe parallel processing.
Alternatively, what if we need to be able to cancel a job after a worker process has...