However, I think the viewpoint of operational semantics gives a more flexible and general implementation of coroutines, or in fact any monad, while being just as simple. I have written a few examples that complement my “Operational Monad Tutorial” which explains this approach in detail:

• A demonstration of a web session, very much like Jean-Philippe Gariépy suggests.

• A TicTacToe implementation that interleaves different players, very much like coroutines would allow us to do.

newtype Iteratee el m a = Iteratee{runIter:: m (IterV el m a)}
data IterV el m a = IE_done a (Stream el)
          | IE_cont (Stream el -> Iteratee el m a) (Maybe ErrMsg)

from: http://okmij.org/ftp/Haskell/Iteratee/IterateeM.hs

It looks to me (but I’m not sure yet) that your type Result i o m a is, more or less, the free monad over (Moore i o . m), where Moore i o is the functor that at x is (o,i->x) (as in Moore automata).

It surprises me that you’ve a nice simple (though, as you admit, contrived) example, that doesn’t involve (as I would’ve expected) the dual functor Mealy i o = o->(i,x). But then, my expectations were probably wrong; coroutine interactions are much more “between equals”, than actor/reactor, client/server or something asymmetrical like that kind.

http://blog.sigfpe.com/2009/11/programming-with-impossible-functions.html

I tried to code a very simple version of http proxy and here is what’ve ended up with: http://haskell.pastebin.com/XfTY6vCE

Unfortunately my Haskell level is less than beginner. My version does not work correctly and I’d be glad if someone point me what’s wrong with it.

Thanks.