Proposal: NF newtype

[David Feuer]

On Tue, Feb 10, 2015 at 1:38 PM, Roman Cheplyaka <roma at ro-che.info> wrote:
> Since deepseq is in no way magical, it can't provide any guarantees (you
> can define an instance that does whatever it wants).
> And neither can NF, being based on deepseq.
>
> All we can say is makeNF = NF . force, by definition.

If we want to have an unsafeMakeNF, which I believe we probably do,
then we need to specify what constitutes correct usage of that
function.