Proposal: Adding Kleisli composition to Control.Monad

[Neil Mitchell]

> > forever even has a different type than repeatM_.  More reason to have
> > both.
>
> Is anyone else feeling uncomfortably crowded by the current namespace
> exported by the haskell libraries?

I think a good naming criteria is "could anyone have a function
defined with this name which _didn't_ do exactly what this one does".
For example, repeatM_ - can anyone think of an alternative definition
which didn't do the obvious thing? However forever could have loads of
interpretations.

Thanks

Neil