From owner-freebsd-current  Fri Dec  6 19:28:40 1996
Return-Path: <owner-current>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.4/8.8.4) id TAA26845
          for current-outgoing; Fri, 6 Dec 1996 19:28:40 -0800 (PST)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.8.4/8.8.4) with ESMTP id TAA26812
          for <FreeBSD-current@freebsd.org>; Fri, 6 Dec 1996 19:27:48 -0800 (PST)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.8.3/8.6.9) id OAA13164; Sat, 7 Dec 1996 14:23:43 +1100
Date: Sat, 7 Dec 1996 14:23:43 +1100
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199612070323.OAA13164@godzilla.zeta.org.au>
To: grog@lemis.de, phk@critter.tfs.com
Subject: Re: Problems with PLIP driver?
Cc: FreeBSD-current@freebsd.org
Sender: owner-current@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

>>1.  There seem to be an abnormal number of errors on the line.  In the
>>    following, freebie is my FreeBSD machine, and papillon is the
>>    laptop:
>
>Well, if there is a significant speed difference between the machines,
>you will have a problem.

`speed' probably means the speed of the parallel port bus interface.
I haven't noticed any problems for connections between a P5/133 and
a 486/33, but I only ran PLIP for 10 minutes to test it.  The P5's
ISA i/o is just as slow as the 486/33's.

>>4.  These problems seem to affect the Ethernet link (ep0) as well.

>The problem is that your speed is probably around 70Kbyte/sec, that
>means that the cpus are running for 22 msec with irqs disabled.

This also loses one or two clock ticks.  I haven't noticed any
problems with the splhigh()s replaced by splimp()s.

>It works sufficiently well that it is usable, but it is impossible
>to get it to really work well and preserve a decent transfer-rate. 

Both together, that is.  The old version of Linux PLIP that I'm
familiar with does extra i/o's for debouncing.  It was about twice
as slow as FreeBSD PLIP (due to about twice as many i/o's).  It
could be made much more reliable by doing a complete handshake.
This would also allow it to run reliably with interrupts enabled.
Then it might be another 5 times slower.  OTOH, modern enhanced
parallel ports can go much faster.

Bruce