contification paper

[Stephen Weeks]

> I haven't touched anything except experiments.tex since I sent you the
> snapshot this morning.  So go ahead and take everything else.

Working away.

BTW, on the experiments section, I think that what we did in the closure
conversion paper worked very well.  We ran the numbers on all twenty or so
benchmarks, but then picked 5-10 representative ones about which we could say
interesting things.  Other qualities in a good representatives are: large
# lines, worst performing, best performing.