Wed Apr 24 09:12:11 EDT 2019

Lifted closures, restarts, caches

If a process can be restarted from a specification, it should be
treated as a cache.  Caches are essentially "not there" at the
conceptual level.  They are an implementation detail.

The lessen is to create as many processes as possible that behave as
caches, i.e. they have no essential internal state.

Is there only one kind of "actual" process?  The one that holds
intermediate user state?  I.e. the stuff between two transactions?