From  (Duncan Barclay)
Organization   Ragnet, don't ask!
Date           22 Dec 1997 09:27:27 GMT
Newsgroups     comp.unix.bsd.freebsd.misc
Message-ID     <67lbpv$3qq$>
References     1 2

In article <>,
	Tony Griffiths <> writes:
> Martin Dieringer wrote:
>> Hello,
>> How to connect two computers via a parallel cable?
>> I have a serial connection working, but would like to
>> have the higher speed of the parallel.
>> Is there something like slattach for the parallel port?
> Well I've seen references to running TCP/IP over the parallel port but
> don't know what, if anything, you need to do to enable this other than
> build in the || port driver (lp0).  Take a look in LINT for any clues.

All you need to do is get a laplink cable (or build one, look
in /usr/src/sys/i386/isa/lpt.c for a pin out, i think there might
be something in the handbook as well) and plug it into the
parallel ports. Then ifconfig it just like a normal point to
point (not PPP) interface in rc.conf

network_interfaces="lo0 lp0"
ifconfig_lp0="inet netmask 0xffffffff"
                   ^ IP address for this machine
                                  ^ IP address for other machine

and do the opposite on the machine at the other end of the plip link.

A couple of warnings:
  - sometimes when using a fast and slow machine (eg.P100 and 486/25)
    the fast machine crashes when you do big transfers. I think it
    is due to buffer overruns on the slow machine side and the fast
    machine doesnt deal with this well. The work around is to get the
    slow machine to be the initiator end of any xfer, i.e. slow machine
    is saying "give me next block" or "here's the next block". Be warned
    things like ftp and rcp have different initiator ends.
  - NFS works (bearing in mind the above, use the old reduce the
    block size hack, although I never tried it). However, due to
    the way an RPC is implemented in FreeBSD, portmap/mountd attaches
    itself to the first network interface (for things like getting
    a hostname etc.) which is often lp0 (depends on kernel config). This
    is a pain because the RPC code can't handle a point to point
    interface (like PLPI or PPP) and you don't get a running mountd.
    Workaround is to add a static route use

    # to enable rpc to work with a PLIP network (rpc binds to the first
    # network interface it finds an lp0 is found v. early)
    route_loop_lp0=" 1"

    in rc.conf
    Search the hackers mailing list for a bit more on this.

> Given the cost of Ethernet NICs, is this really worth it?  You can
> connect two systems back-to-back even with twisted-pair if the cable is
> a cross-over.  The same is true for FDDI (CDDI?) as long as Tx is
> connected to Rx and vice versa.  No hub is required!

Ahh, but what about a really cheap 486/25 laptop that you pick up from
the scrap heap at work. I did a full install onto an old Compaq Contura
over PLIP and used to for weeks.

PLIP runs fast (in this scenario about 70kB/s) and is easy to use.


Duncan Barclay          | God smiles upon the little children, | the alcoholics, and the permanently stoned.