Remote Network Commands

Jens Hartman shows how to use rlogin, rcp, and rsh to transfer and manipulate data on different computers from across the network.

The rsh command is the most powerful remote command, as it can be integrated into a pipe. This enables you to execute complex command sequences between different machines. Without any command, rsh rlogins to the other machine. When I am logged into orange:

rsh banana
logs me remotely to banana, while:
rsh -l fred apple

does the same on apple, where the username is different from the one for my current shell on orange.

Both stdout and stderr from the remote machine are piped to the local machine. After establishing the connection, neither /etc/profile nor .bash_profile (for bash) nor .login (for csh and tcsh) are scanned. This can be confusing in the beginning, as not everything you defined as variables and aliases for a login shell are present. This is different from rlogin, which gives you a real “login shell”. The most common problem is that you use a command that cannot be found because the path variable has not been set correctly. In these cases, you could set your path in your shell initialization file, such as .bashrc or .cshrc (there is no such file for /bin/sh, however). A more general solution is to simply use fully qualified pathnames for commands.

The easiest use of rsh is a simple command like:

rsh banana ls

You should get a listing from your home directory on banana. If you want to use options that start with “-”, the syntax would be:

rsh banana `ls -al'

Everything in between the quotes is executed on banana.

The listing was sent to your local stdout, your screen, just as if ls had been executed locally (on orange). Below are some examples of creating a tar archive on banana from a directory called bin, with the output going to stdout and the archive file being placed on orange:

rsh banana `tar cf - bin' | dd of=archive.tar


rsh banana `tar cf - bin' > archive.tar

Of course this will work for different usernames or in the other direction too:

rsh -l fred apple `tar cf - bin' > archive.tar       tar cf - bin | rsh -l
fred apple dd of=archive.tar

The dd command is very handy here, because it copies stdin to the file specified by of. If I were to use > to redirect the output, I would end up on orange again, but I want the file to be written to apple instead, which dd does correctly.

rsh Wizardry:

Now for some examples of what happens on which machine. Start with something simple:

        ps -aux | grep root

This shows all processes root owns on your local machine. The next command will show all processes on banana which are owned by root.

        rsh banana `ps -aux' | grep root

In this case, the ps command is executed on banana, but grep is executed locally (on orange).

The next command does the same thing, except that the grep command is executed on banana, as is the ps command.

rsh banana `ps -aux | grep root'


rsh banana `ps -aux' | rsh banana `grep root'

The next example shows how to split stderr and stdout (this works only for sh or bash, not for csh or tcsh!):

 rsh banana `dd if=bin.tar 2>fehler' > test.dat

Here we have dd that writes its error output to the file fehler on banana, but which transmits its standard output to orange into the file test.dat. The secret here is the use of quotes. Because 2>fehler is inside the quotes it is executed on banana. Things can get very tricky. Not only can you make full use of shell commands, you also can run them on different machines:

rsh banana `tar xf bin.tar `rsh banana `tar tf \ bin.tar' | grep gj2.c`'

Here I have a tar archive bin.tar on banana. In it there is a file called usr/local/src/gj2.c. First, there is a command expansion in between ` ... `. This expansion returns all the filenames from the archive which contain the string “gj2.c”. First, tar tf returns the list of files in the archive and then grep (running on orange) performs the pattern match. This yields usr/local/src/gj2.c and usr/local/src/gj2.c~. Now the first tar (tar xf bin.tar) extracts these files on banana.

Imagine that you have a tape drive attached to banana and another to orange. You want to make a copy of a tape. Dumping the contents of the tape to the disk drive and copying it back to another tape would be a solution, but would require enough free disk space to hold the entire contents of the tape. Another solution would look like:

rsh banana dd if=/dev/rst0 ibs=1024 | dd \
of=/dev/rmt0 obs=1024

Here /dev/rst0 is a SCSI tape drive on banana and /dev/rmt0a SCSI tape drive on orange. Now you want to process your data with a special program called demux. After processing, your data has shrunk considerably to about 200 megabytes. As we operate on binary data, porting our program to banana would be very time consuming (orange is still my Linux machine). On the other hand, orange doesn't have 200 megabytes of free space. We do the following:

rsh banana dd if=/dev/rst0 ibs=1024 | demux | \
rsh banana dd of=file.dat

Now we read from banana, process on orange and write back to banana.

In the next illustration you want to access a printer connected to banana. We have a PostScript file,, that we want to send on a printer called p_a4:

dd | rsh banana `lpr -Pp_a4'

You might want to have a look at the file before you print it:

dd | rsh banana `xv - -display orange:0'

This will only work under X-windows. On orange, you would have given a command like xhost +banana first.