I'm not sure whether this is just an i386 problem, but...

I have a variety of BSD boxes, some running NetBSD 1.6, some 2.0.2,
one running 3.0 and others still running BSD/OS.   They all talk to a
networked printer, an ln17ps FWIW, with a printcap entry that looks
like this:

lp:\
        :lp=:rm=ourprinter:rp=PASSTHRU:mx#0:sd=/var/spool/lpd/ln17ps:\
        :lf=/var/log/lpd-errs:

A similar entry for a printer named lpx is exactly the same, except
for rp=TEXT rather than PASSTHRU.   This has worked for many years.

The NetBSD 3.0 box is located off-site in a secure facility.    I have
just got round to trying to set up printing facilities.   If I do
lpr -Plpx somefile.txt everything works as expected: the file is
queued in the right spool directory and so forth, then printed.
However, a plain "lpr somefile.ps" queues the file correctly, but
/var/log/lpd-errs complains that there is no device /dev/lp.   So
somehow, lpd knows how to read /etc/printcap (because queueing to lpx
works) but doesn't handle the lp entry.   It's not that there is a
competing printcap entry with /dev/lp specified.   The two differences
are that it's 3.0 and it's on a remote site.   I think I've excluded
the latter as the problem, because lpx works.

Any suggestions?   I'm a bit wary of moving my other machines over to
3.1 if there is going to be a continuing problem here - I can't just
rename the printer without going through any number of scripts and
editing them to specify a printer name.


-- 
Steve Blinkhorn <steve@prd.co.uk>