xrsh 5.92-8 / usr / share / doc / xrsh / FUTURE-WORK

Here are some unstated goals for the xrsh script to consider in
designing enhancements:

1. Keep xrsh as one script.  Configuration and installation are more
difficult with ancillary files.  As one script, it can be copied
easily from system to system.

2. Keep xrsh portable.  Avoid features which might not work on all
UNIX variants.  E.g. some rsh don't have a -n option.  Some
/bin/sh don't have shell functions.

3. Keep the number of programs fork()ed by xrsh to a minimum.  xrsh is
used a lot and always interactively so speed is important.  This is
why I play all those games with IFS, to avoid forking awk or sed.

Here are some things which might be done in the future or have been
recently tried. 

1. Parse the DISPLAY environment variable without forking any extra
processes like sed.  Perhaps use IFS in the shell.
<done, but doesn't work in Ultrix /bin/sh>

2. Pass other environment variables besides $DISPLAY.  Users might
want to pass an arbitrary set.
<can now pass XAUTHORITY using XRSH_AUTH_TYPE and arbitrary
  environment variables using XRSH_ENVS_TO_PASS>

3. Don't use backslashes for quoting, only  use only ' and " for portability.
<What hosts don't handle backslashes?  This turns out to be very hard.>

4. Pass  $XUSERFILESEARCHPATH to the remote host?
<I'm not sure this is always what you want to do, but you can
  do it if you want to using XRSH_ENVS_TO_PASS.>

5. Add command line flag to describe the remote machine (like -sysv or
-irix).  This would get all the config options correct without having
to guess.
<I hope not to have to do this -- I hope to make the one script portable.>

6. If you have to rsh the xhost to the serverhost, run it in the
     background and use wait before the final rsh to improve speed.
<I did this, but it made some system V hosts (HP-UX 7.0) go insane and
say "wait: pid 232 is not a child of this shell" over and over again.
I finally deleted it in the interests of portability.  Timing showed
it wasn't a major win in speed anyway>

7. Expand DISPLAY environment variable  from foo to foo.org.com so
that it works when you leave your local site.
<hard to do in a site independent manner.  Now uses ip addresses if the 
remote host name contains periods like foo.bar.com>

8. Decide rsh vs. remsh vs. rcmd at install time instead of run time.
<disadvantage:  one script can't be transferred from one machine to another>

9. make first line be : instead of #!
<Can't -- without #! some systems use $SHELL to eval scripts.  May not
be the right thing>

10. Use IP address when setting $DISPLAY for remote hosts that can't
resolve local hostname.
<This is done for remote host names with . in them>

11. Have an .xrsh.rc file which can be used to tailor defaults for
specific hosts.

12. If using xhsot authorization, optionally disable access to the
host after running the command.  Of course, on hosts where rshd
doesn't return, the xhost -host would never get executed until the
client exitied.  This feature was requested by Rick Calder, NCR
Systems Support Center - New Jersey [att!]rick!rick rick@rick.att.com
attmail!rcalder.


13. (From kmp@harlequin via gildea@expo.lcs.mit.edu) in order to preserver
users quoting for arguments to client, don't preparse $@ into a variable
and then pass that variable along to the client via rsh.

Instead, of setting and resetting the cmd line args for IFS hacking, push 
a marker on the arg list and push other things on the front.  Then 
shift things off the front and stop when you get to the marker.

at the end, just pass $@ along as the arguments to the client.

set new stuff XYZZY "$@"