I finally (after about three more days of pain and effort) found a
solution to this problem on my own.  The problem is that the version of
magicfilter-t in the printconf RPM is not compatible with lpdomatic in
the foomatic RPM.

As near as I can figure (I'm neither an expert in printconf, foomatic,
nor python), the printing sequence in the updated Redhat 7.2
distribution runs as follows:

1) lpd reads /etc/printcap and sends options to
/usr/share/printconf/util/mf_wrapper;

2) mf_wrapper reads some m4 macros and constructs a
/usr/bin/magicfilter-t command with those options;

3) magicfilter-t, examining the filters passed to it, invokes
/usr/share/printconf/util/mf_postscript_wrapper;

4) mf_postscript_wrapper invokes /usr/sbin/lpdomatic.

"magicfilter-t" read options from its command line -- but does not pass
those arguments to mf_postcript_wrapper (which is written as though it
expect them).  Instead, it sets some environment variables, $ZOPT for
example.  

However, lpdomatic does not access those environment variables; it
expects the options to come in on its command line.

The following patches are a brute-force solution to the problem.  They
force the file /var/spool/lpd/<queue>/mf.cfg to pass the options from
mf_wrapper to mf_postscript_wrapper.

As for me, I'm switching to CUPS.  It's flakey, unreliable, and hard to
configure -- but it's all of one piece, and it knows how to interpret a
standard PPD file.

The following patches are meant to be applied in /usr/share/printconf. 
After they're applied, it's necessary to run printtool (or
printconf-gui), click on the "Driver Options" tab of the relvant print
queue, then click the "Apply" button.  After that, the -Z options will
work.

> diff -ru util/mf_wrapper.original util/mf_wrapper
--- util/mf_wrapper.original    Mon Feb  4 11:48:34 2002
+++ util/mf_wrapper     Mon Feb  4 13:52:47 2002
@@ -8,7 +8,16 @@
 MF_RULE_DIR='/usr/share/printconf/mf_rules'
 MF_RULES=$MF_RULE_DIR/mf[[:digit:]][[:digit:]]-*
 TMP_FILE=`mktemp /tmp/printconf.XXXXXX` || exit 1
-cat ./mf.cfg $MF_RULES | m4 > $TMP_FILE
+
+# Create an m4 rule to pass the options to lpdomatic
+# via magicfilter.
+OPT_FILE=`mktemp /tmp/printopt.XXXXXX` || exit 1
+echo -n "define(LPRNGOPTS, " > $OPT_FILE
+echo -n ""'`'"" >> $OPT_FILE
+echo -n "$*" >> $OPT_FILE
+echo "')dnl" >> $OPT_FILE
+
+cat $OPT_FILE ./mf.cfg $MF_RULES | m4 > $TMP_FILE


 # Find the effective language set to use.
@@ -45,12 +54,15 @@


 # Run magic filter
-/usr/bin/magicfilter-t $TMP_FILE $DEBUGSTRING $* < /dev/stdin
+/usr/bin/magicfilter-t $TMP_FILE $DEBUGSTRING < /dev/stdin
 RETVAL=$?

 if [ -f $TMP_FILE ]; then
        rm -f $TMP_FILE;
 fi
+if [ -f $OPT_FILE ]; then
+       rm -f $OPT_FILE;
+fi

 if [[ $RETVAL -eq 0 ]]; then
        TERMINATION=`grep '^[[:space:]]*#[[:space:]]*TERMINATION='
./mf.cfg | sed 's/^[[:space:]]*#[[:space:]]*TERMINATION=//;'`

> diff -ru util/printconf_backend.py.original util/printconf_backend.py
--- util/printconf_backend.py.original  Mon Feb  4 13:27:37 2002
+++ util/printconf_backend.py   Mon Feb  4 13:27:11 2002
@@ -260,6 +260,9 @@
 make_mfomatic_perl_str = r"""

 $postscript_wrapper =
'/usr/share/printconf/util/mf_postscript_wrapper';
+# This complex string, when appended to the postscript_wrapper,
+# passes the contents of the LPRNGOPTS variable to lpdomatic.
+$lprng_opts = "'``LPRNGOPTS''`";

 undef $/;
 umask 0077;
@@ -376,7 +379,7 @@
                $text_filter = 'text "" \r\n\14 ';
        }

-       $ps_filter = "filter $postscript_wrapper $rerender_ps --mfomatic
$gs_driver-$printer_id.foo";
+       $ps_filter = "filter $postscript_wrapper $lprng_opts
$rerender_ps --mfomatic $gs_driver-$printer_id.foo";
        # Yes, this is stupid. But we don't have reliable information
anywhere on #dnl
        # if a given printer supports pcl and pjl, so what ya gonna do?
#dnl
        $pcl_filter = 'cat';

Thanks for the patch!  I think that this part:

OPT_FILE=`mktemp /tmp/printopt.XXXXXX` || exit 1
echo -n "define(LPRNGOPTS, " > $OPT_FILE
echo -n ""'`'"" >> $OPT_FILE
echo -n "$*" >> $OPT_FILE
echo "')dnl" >> $OPT_FILE

would be better written as:

OPT_FILE=`mktemp /tmp/printopt.XXXXXX` || exit 1
echo 'define(LPRNGOPTS, `'$*"')dnl" > $OPT_FILE

If the $* macro gets interpreted within single quotes, then that's fine.

I was in constant pain trying to get the backtick within the string, and I
finally gave up.

Okay (it does).  Also, there was a modification needed in the $mf_type eq
'POSTSCRIPT' case too.  I've applied this in the CVS tree.

Thanks for getting this working!

Fix built as printconf-0.3.68-1.

*** Bug 56834 has been marked as a duplicate of this bug. ***

Hmm, this is causing problems.  lpdomatic's call to getopts doesn't understand
what's going on and complains.

The lpdomatic command-line ends up like this:

/usr/sbin/lpdomatic: called with arguments:
'--lprng','-Aroot@localhost+330','-CA','-D2002-03-06-12:32:12.224','-Ff','-Hlocalhost.localdomain','-J/usr/share/printconf/tests/testpage-a4.ps','-N/usr/share/printconf/tests/testpage-a4.ps','-Ptest','-Qtest','-a/var/spool/lpd/test/test.acct','-b18255','-d/var/spool/lpd/test','-edfA330localhost.localdomain','-f/usr/share/printconf/tests/testpage-a4.ps','-hlocalhost.localdomain','-j330','-kcfA330localhost.localdomain','-l66','-nroot','-sstatus','-t2002-03-06-12:32:12.000','-w80','-x0','-y0','/var/spool/lpd/test/test.acct','bj200-58720.foo'

but lpdomatic's getopts call is expecting this format:

   getopts("cw:l:i:n:j:h:z:o:Z:J:");

It's just the '-Z...' and '-J...' bits that we want, so I plan to just extract
those in the mf_postscript_wrapper like this:

        case $ARG in
                "--rerender-ps" )
                        RERENDER_PS="1" ;;
                "--mfomatic" )
                        MFOMATIC='1';;
                -Z* | -J* )
                        args="$args $ARG" ;;
                -* )    ;;
                * )
                        args="$args $ARG" ;;

Fix applied in 0.3.72-1.