One possible way out of this trap would be if, perhaps, the variant of
Haskell you picked were actually GHC's core language. That could
(...)
It still seems far too ambitious for GSoC, though. And I remain
unconvinced how useful it really is likely to be. I'll grant there
are other people that care a lot more about ML than I do.
Core to F# gets GHC on the CLR, which if nothing else makes the Windows support situation for GHC look quite a lot better. (I am still bemused by MSR not being interested in better Windows support for GHC....)