Keyword arguments for compile() and __import__(), as
discussed at PyCon. There is a slight speed difference
on my development system: __import__ is ~1% slower,
compile() is ~1.5% faster. I blame both on caches,
alignment and what not, so I bet they're totally
different for everyone. I'd say the advantage of
managing the growing set of arguments outweighs the
speed impact (if any.)