John Clements writes: > Things get pretty messed up for > non-tree-like evaluations, though; what about exceptions, or (Heaven > forbid) continuations? That's why I said graph and not tree...