momonorphisation and imperative/weak type variables

[Stephen Weeks]

> I'm assuming that the old SML only noticed refs
> in its imperative/weak type variables, but not effects like that.

Right.