On Tue, 11 Mar 2003 00:40:38 +0100
Daan Leijen
I do think it is an interesting subject, and I encourage you to pursue your ideas and create a working implementation to show and discuss. Maybe you can even work together with other people to find the right abstractions.
Oh, yes, my e-mail was meant to G.Russel AND of course anybody else who wants streams. And was meant to take the streams issue outside the GUI list. But, as I said, the GUI team has to leave the road open to a streamed implementation, wich means "abstract over your state". So, in conclusion, I am just looking for people to write this stream library together, not proposing to solve issues here; sorry for not having expressed myself well. Now, if there's someone who cares, I want to give my last contribute to this discussion, or better said monolgue :) Maybe I am wrong, but it seems to me that streams appear something like a research topic to the eyes of some people. I would like to point out that streams are nothing new, and nothing badly understood. The implementation details, and a general framework, are the goals of my proposal, not to study something simple, wich has been there for years. Streams are not a new topic in functional programming. Sometimes they are called "Dialogues". See for example this article http://citeseer.nj.nec.com/jones93imperative.html in wich SPJ and Wadler are *introducing monads*. The previous alternative where streams, indeed. The reason why monads have been introduced is that streams are uncomfortable in many situations, but think about the function "hGetContents" or "readFile", or "getContents". They are just useful in some situations, where you are coding a purely functional algorithm, but need to feed it with non-deterministic input. I repeat: streams are not FRAN, they are nothing new, just a common pattern in lazy functional programming. When a common interface is decided, by G.Russel, me and M.Chackravarty, by the GUI team or by anybody else, I *hope* the "standard haskell GUI library" will have a stream-based interface, since without it the entire program is being compelled in the IO monad, wich is not so easy to code, and not so readable. I hope to have not bothered or annoyed anybody, because I am here to cooperate (for what I can), not to annoy ;-) And, of course, I would have never proposed myself to do a job, if it was not easy to do, and if it was hard to understand from a theoretical point of view. Vincenzo