yes, I am using netsend/receive, but the program on the other end, piping the pdreceive into the program and forwarding the output via pdsend back to PD, needs to be started. If this were possible from shell, or otherwise from inside PD, then I could have small units that started the piping chain within the PD patch rather than having to start it myself from the command line.
i've made a little test script ( here attached ) that does the lauching of pdreceive and then communicates with pd.
it DOES stay alive and the problem is rather that the process is not destroyed when i destroy the shell object, that's why <defunct> processes appear ( the father ( shell object ) is dead and the script is still alive ).
then you cannot launch a new instance of your script if the former one holds an exclusive ressource ( like a port number ).
in my example, this luckily works but this is surely unclean...
cheers,
yves