@User: I do both. I like to use the python "subprocess" package where I can just run magic and feed it input from stdin. To do that, though, you want to run "magic -noconsole" to make sure that magic is communicating directly with the terminal that launched it. Or. . . I can generate a Tcl script for magic to process, and feed it to the magic command line. I don't have much preference of either method over the other.