serialization and globalization

[Stephen Weeks]

> In general, then, it seems that I can't depend on serialization of
> closures.

I would like to strengthen (weaken?) this statement:
	I can't depend on serialization of mutable objects.

Of course, since the globalization optimization happens in the Cps
long after closure conversion, my example has nothing to do with
closures, it simply has to do with the fact that I am serializing a
mutable object.  The real thing that has screwed me is that MLton has
changed the code to refer to a (Cps) global.