To whoever's interested in helping:
I just pushed to the `report/hmdup` branch a (very much)
work-in-progress article where I try to document the type inference
algorithm of Typer (well, more like the one I'd like to use than the
one we do use, so it's a mix of "document" and "cleanup/fix" and
"after-the-fact design").
If you're interested in helping, I suggest you try and understand what
is currently there and then come back with questions about what you
don't understand. You'll also see a boatload of FIXMEs which you're
welcome to try and attack.
I think currently one of the more feasible and useful things would be
to work on the proofs that our inference for System F is conservative,
sound, and complete.
Stefan