Wow.. quite many responses.. I'll have to rearead them thoroughly.. ;) I think GHCI would be great (expect it's start up time compared to bash ;) I wouldn't use hs-plugins because it needs quite a lot of time to compile pieces of code.. I got the idea to use something like Don Libes expect/pexpect (python) as layer between user/ghci ? So you can use short characters beeing expanded? Problem: How to get to know when ghci has finished executing a command? I don't like the approach to putting every command in a function by default. But I want to do so for some commands. Eg sox/ aplay, For most small problems you can't get a shorter command than using bash.. So why not be able to switch. Meta-b -> bash Meta-h -> haskell Meta-9 -> plan9 And after executing a command with bash take over the environment variables? One thing I'd really like to improve is history/ completion. So it would be cool to extend ghci to be able to use something like class Completion Completable where getCompletions :: Completable -> firstChars -> [String] or doCompletion :: Completable -> firstChars -> IO Completion or something more general perhaps even providing some list to select individual filenames... I'd also like to do somehthing like prog1 | prog2 | prog3| xargs mplayer .. Then prog1-3 and mplayer may be running at the same time. I still want to be able to send keys to prog1,2,3 and mplayer.. (eg to mute mplayer or to speed the music up, goto the next file, ....) That would be reallly cool and interactive ;) Marc