Re: [Toss-devel] No nominal subtyping with higher-order types

[Lukasz S. <luk...@gm...>]

Actually it means that there should be the covariant unification that
finds the greatest lower bound (with <= being ISA) and invariant
unification that just finds the substitution that makes terms equal;
and the covariant unification should use the invariant unification for
"funtype".