Re: Referential Transparency (was Re: OO in Haskell)

[Adrian Hey]

On Thu 07 Oct, Jan Skibinski wrote:
> I think that the monadic IO provides us with such a
> simplification. As long as we realize what are its limitations
> and as long as we stay within reasonable limits of the concept
> we should be fine here. The operative word here is "realize".
> Do we really know those limitations? 

Yes, I agree. As a pragmatic aproach to simple imperative IO I'm
content with IO monads and I don't really care if we have referential
transparency or not, just so long as I can do IO.

But, single threaded monadic IO is very restrictive. I think it's important
to understand whether or not we really do have referential transparency
with monadic IO, if other models of interaction between program and
outside world are (like those in Concurrent Haskell) going to be rejected
because we 'lose referential transparency'.

If we lose something we never had, then what's the problem I wonder?

Regards
-- 
Adrian Hey