I hope I can persuade you of the problem here; after the bind update released
today, I found environment leaks in the running named.

The reason (I think) is the way that the scripts use:
   /etc/rc.d/init.d/named condrestart >/dev/null 2>&1 || :

which bypasses the env-scrubbing behaviour of /sbin/service. This is bad, not
for any bind-specific effects I can put my finger on, but for the principle that
a daemon should *always* start in a consistent manner, irrespective of the "yum
update" it was restarted under or the user's particular locale.

In an unscientific survey, I count 32 uses of /sbin/service vs 8 calls with
/etc/init.d on the installed packages of my FC4 box.

Please consider using /sbin/service instead of /etc/rc.d/init/ in the RPM script.