🚀 go-pugleaf

I've just upgraded from GNU Finger 1.35 to 1.36 and have some comments
and question.  Overall, I was able to get Finger installed and
working, but to do so I had to re-edit the config files several times
and remake several times to get it right, mostly because the only way
I could figure out the impact of several config options was to try it
out.  It would have helped if I had read the info section on config
files before installing, but I didn't realize that and just trusted
the support/INSTALL file to explain everything.  Also, I was
completely unable to get mugshot support compiled.

In the hope of easing installation for others in the future I offer
these comments.

Finger Version: 1.36, from prep.ai.mit.edu.
System: A Sun SPARCstation 1+: SunOS 4.1.2 2 sun4c as server; a small
network of Sun Sparcs (1+'s and 2's) as clients.

Compilation Software:

    gcc           2.2.2
    GNU make      3.62
    X11	  	  Release 5, patchlevel 16, from export.lcs.mit.edu

Problems and suggestions for improvement:

o The same exact paths have to be edited into several files: config.h
  and include/fingerpaths.h.  This is extremely error-prone, not to
  say tedious.  I thought GNU autoconf was supposed to isolate all
  configuration options in one place.

o More explanation needed for paths in include/fingerpaths.h.  Most of
  the default values are for files in /usr/local/etc/fingerdir/.  But
  are these files all used only by the server?  Only by clients?  By
  both server and clients?  Should this directory by NFS-mounted and
  shared, or local to the server (or client) machine?  Does it matter?

  There's one comment on the ttyloc file that suggest it be in a
  per-client directory, but is this the only file for which this would
  be a good idea?  I'm going to have to guess.

  Since GNU Finger is specifically designed to support a networked
  environment and most networked environments share disks, it would be
  extremely helpful to have explicit instructions for each directory
  mentioned in fingerpaths.h as to whether that directory should be:

    o local to the server only
    o local to *each* client excluding the server
    o local to *each* client including the server
    o shared by all clients (excluding the server)
    o shared by the server and all clients

  I'm not sure all these combinations are needed for GNU Finger...

o Paths to X11 include files and libraries are hard wired into the
  Makefiles; we don't keep ours in those places.  Should be
  configurable.

o Bunch of #includes of the form #include <X11/foo.h> seem to really
  want to be of the form #include <X11/Xaw/foo.h> maybe (I'm no X
  expert, although I did install our X11R5 setup and dozens of other
  clients successfully).

o Where is X11/Scroll.h?  We have X11R5 installed and I selected the
  R5 define in config.h, but the compilation blows up looking for
  Scroll.h.  Is this include really supposed to be in R5?  I set -I to
  point to our local include dir, where X include files live in subdir
  X11 but no go.

o Info file only describes the shared form of ttylocs file.  Had to
  poke through the source to find out how to set up the local form.

o Is the ``what'' field known to work on Suns?  It's always blank for
  us.

o Are any of the finger executables supposed to run setuid?  I notice
  that .fingerrc's are executed by user ``nobody'', which I guess is
  potentially more secure, but prevents stuff like:

    chmod go-r my-data
    echo "grep 'Public Part' my-data" > ~/.fingerrc

  since user nobody can't read my-data.

In general I think GNU Finger is a tremendous program, thanks a bunch!

--
Keith Waclena                             	k-waclena@uchicago.edu
Member EFF                                      MIME / RFC 1341

And remember, it's not only the YP master, but it's also a client.