This topic seems to be generating a lot of heat by friction among
the discussants. I offer the modest suggestion that what work is
is relatively unimportant, a matter of taxonomy at best. If one
looks at a particular transformation a system undergoes and asks
"What are the initial and final energies of A and B?" instead of
"How much work does A do on B?", is the question of what to name
the agent of change operating in the system really physical? I
can't think of a situation in which knowing how to classify such
a process really matters.