Excerpts from C.M.Brown's message of Sat Sep 06 17:33:58 +0200 2008:
Nicolas,
Yes
I feel, also, that as a reasonably experienced Haskell user, I am getting confused with what should be fundamental Haskell concepts. Perhaps these concepts should be made much clearer to beginners in the first instance.
This article [1] seems a good introduction. Maybe one could sum-up what you think is missing from this article and extend it a bit.
For instance one could explain the point about the run-time system.
Although what you've told me may be true (and I thank you for your explanation), I'm still not at all convinced that it's useful to think of IO as being pure. It still produces an effect if I evaluate it, no matter how we sugar-coat these things.
I like the idea of the type system flagging the function as being effect-full by the IO - and find that more useful than thinking about runtime systems.
Once we've understood that monads (even the IO) is not an exception in the *language*, but that's the runtime system that consume and evaluate IO values starting from main. Then one could think in a more pragmatic way: IO is effect-full, and non IO is pure, that the type system help us to separate effects from pure code... -- Nicolas Pouillard aka Ertai