Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!batcomputer!cornell!vax5.cit.cornell.edu!umh
From: umh@vax5.cit.cornell.edu
Newsgroups: comp.unix.questions
Subject: Can I pass variables to rlogin?
Message-ID: <1991Apr14.191051.4037@vax5.cit.cornell.edu>
Date: 14 Apr 91 19:10:50 EDT
Distribution: comp
Organization: CIT, Cornell University
Lines: 29

I am very surprised this is not part of FAQ? Is no-one else as ambitious as me?
The problem is that we have a number of machines (Sun3, SparcStation, RS6000)
on three different NFS networks, all running X-Windows. Now often one will try
to rlogin from say an RS6000 to a Sparc in order to execute a Sparc binary. And
well and good- we type rlogin host and get a terminal to the Sparc. However, if
the binary we wish to execute is an X program we then have to type, at the
Sparc, setenv DISPLAY rs6000:0, which becomes a right pain the 10 000th time
you've done it.

So my question to the net is how can one get around this? At present I have
this really ugly kludge with temporary files that .login and .xinitrc look at
and possibly change. This works so-so across a single NFS network, but to make
it work across three requires rlogin be aliased to something with rcp in it,
there are horrible synchroniztion problems- etc. 
The sort of nice way I'd like to do this would be either
be able to pass environment variables to rlogin (like passing #defines to cc)
so I could say rlogin host -Dfrom=`hostname` (then alias that, of course)
failing something that general, can I at least somehow test in my .login if
this .login is being started as an rlogin from a remote machine, and what that
machine's name is?

Or am I barking up the wrong tree? Is this nicely fixed by some X magic, not
using UNIX at all?

If anyone has been able to do something like this, please tell me. I'll
summarize any methods I find, of general interest.

Maynard Handley