John A. De Goes

Also the APL folks have a point: doing everything at the level of arrays gives you lots of benefits. Could take a numerics lib in that direction...

@NeQuissimus That looks good to me. You'll have to see how that plays with `instanceOf`.

@vmarquez This needs some TLC when you get a chance.

> As @edmundnoble pointed out, Profunctor with Category is also called a Prearrow (an Arrow without first), so maybe FreePrearrow? 👍 We should also consider taking the time to rename...

The law is wrong for 2 reasons: the reason you state, and another reason, which is that it may execute some part of `f`, which makes it not equivalent to...

I will add a note here to extract a law from this: ``` fork f >>= \f -> interrupt f t *> join f == terminate t ```

@durban My first goal is to eliminate `bracket` as a primitive. If I can't, then it will replace `always`. If I can, it will go alongside `always` so it can...

@durban I found it only needed `IORef` and `ensuring` (e.g. `always` above). You can see the implementation [here](http://github.com/scalaz/scalaz-zio). I still don't like `async` and `asyncPure` and `sync`.

Maybe `IORef` should be treated as primitive. It's the `IO` version of a mutable reference—hard to build much imperative code without it. You could create them from a type class....

@shajra Such laws can be expressed in terms of `MonadFree` with `liftF :: f a -> t f a`, and `foldFree :: (Monad g) => (f ~> g) -> Free...