The idea of pseudo-type-judgements being
proven by a combination of the compiler's
and programmer's effort seems like it might
really be useful.
I wonder if a careful semantic description
of how malloc should behave has been done?
Still not sure how to think about
recursivity with non-(finite or integer) types.