Hello Joel, Tuesday, December 06, 2005, 1:27:58 PM, you wrote: JR> #1 reading messages from a socket and posting to #3, JR> #2 reading messages sent by #3 and writing to the socket, JR> #3 reading messages sent by #1, processing them and posting to #2. what you get by dividing this into 3 threads? i think that this have meaning ONLY if you then join all socket reading threads together and use one select to wait on them all JR> data Event JR> = Enter JR> | Exit JR> | Quit JR> | Timeout String JR> | Connected JR> | Disconnected JR> | Error String JR> | Cmd Command JR> | Custom Dynamic -- can't pattern-match on this? JR> deriving Show using "Dynamic" have meaning only if you don't know at compile time what messsages can be sent. is that really the case? JR> Last but not least, to be able to send messages to any thread I would JR> need to keep those around in some sort of a table. I would need to JR> create records and keep the thread id, the mailbox and possibly some JR> sort of a per-thread string so that threads can update me on their JR> doings. imho, you are think in Erlang style, which is ultimately dynamic and run-time oriented. what you really need - is an abstraction "Poker server" which have interface consisting of several functions, which includes ability to create new server, send it a fixed (at compile time) set of messages, and that's all (may be i don't know about something more). plus abstraction "Logger", which have facility "log to me", this facility passed to routine which creates "Poker server" data Logger = Logger {log :: String -> IO () } createLogger = do c <- newChan forkIO $ loggerThread (readChan c) return $ Logger (log = writeChan c} data Server = Server { send :: Event -> IO () , kill :: IO () } createServer logger socket = do c <- newChan t <- forkIO $ serverThread (log logger) socket (readChan c) return $ Server {send = writeChan c, kill = killThread t} main = do l <- createLogger s <- mapM (createServer l) [1..1000] ... -- Best regards, Bulat mailto:bulatz@HotPOP.com