I agree with all John Denker says about time scale; it is a point I
emphasize when I teach thermodynamics, but not classical mechanics.
Since he really only disagreed with my answer to his reformulated
version of my question we really have no disagreement. My point is
that there is no value to worrying about work *per se* done in such
a system, that's all. If the definition of work in mechanics depends
on time scale then I think it is not terribly useful. Do you feel
that the amount of work varies *continuously* in time, John? Can you
really say with a straight face that after the eddies have died out,
the canoeist has done *no mechanical work at all* on the water? In a
mechanics course I would never say that; I would come down firmly on
the side that says the canoeist does an amount of work equal to the
force he exerts on the paddle time the length of the stroke in his
frame. I would then go so far as to say that the paddle does an
equal amount of work on the water, and that no net work is done on
the paddle by the other two agents. That's fine for an introductory
course, but if I were pushed by the student to answer the question I
asked originally I would demur and tell him that the question really
isn't a physical one. I can be much more definite in answering the
question about the sound produced by a tree falling in the absence
of a human auditor.