Abstracting over things that can be unpacked

[Aleksey Khudyakov]

> As I mentioned further up in the email, I think this needs to be done at
> compile time. However, I'm not sure type classes are the right mechanism, as
> they don't guarantee that the polymorphism is resolved at compile time.
> Perhaps type families, in some form, is the right solution.
>
There is problem with type families. Currently GHC is unable to unpack them[1].

[1] http://hackage.haskell.org/trac/ghc/ticket/3990