The PIPE filesystem arose in answer to the question: how can I save all or part of a Linux system onto/ via a DOS computer or a Novell fileserver? The PIPE filesystem was designed as a quick attempt to solve this problem In the PIPE filesystem either shell scripts or Linux programs can be stored. These programs are treated on the client side (eg DOS) like simple files. Opening these files via the client causes a popen of the programs. The server passes as the first parameter either CREAT READ or WRITE, depending on the mode of the corresponding openfile operation. This allows the PIPE filesystem to provide a direct interface between client applications and Linux programs. The problem stated above could then be solved with the following simple shell script, which was stored in the PIPE-filesystem: #!/bin/sh case "$1" in 'CREAT') ;; 'WRITE') cd /u3 && tar -xf - 2>> /tmp/tar.in # restore directory /u3/mar ;; 'READ') cd /u3 && tar -cf - mar 2> /dev/null # save directory /u3/mar ;; *) ;; esac Under DOS this 'Pipe File' can now be 'copied' into a local file using the Copy command (->save), or the local file can be copied into this 'Pipe File' (->restore). A simple print operation can be achieved with the following script: #!/bin/sh case "$1" in 'WRITE') /usr/bin/lpr ;; *) ;; esac Various unix programs can be invoked with the following script, after it has been linked with the required program name. #!/bin/sh case "$1" in 'READ') /usr/bin/`basename $0` ;; *) ;; esac I would appreciate hearing about further documented applications of the PIPE filesystem or suggestions for other ways of using it. Martin