Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!bloom-beacon!arktouros!dyer
From: dyer@arktouros.MIT.EDU (Steve Dyer)
Newsgroups: comp.sys.ibm.pc.rt
Subject: Re: BSD Unix on 6150-135?
Keywords: RT BSD
Message-ID: <9486@bloom-beacon.MIT.EDU>
Date: 27 Feb 89 23:05:04 GMT
References: <1966@nmtsun.nmt.edu> <28304@bu-cs.BU.EDU> <45@archet.UUCP>
Sender: daemon@bloom-beacon.MIT.EDU
Reply-To: dyer@arktouros.MIT.EDU (Steve Dyer)
Organization: MIT Project Athena, Cambridge MA 02139
Lines: 47

In article <45@archet.UUCP> wlm@archet.UUCP (William L. Moran Jr.) writes:
>I thought I would contribute my $0.02 to this. If you think that there
>is any chance that you may want or need to use a serial connection to
>the RT (i.e. serial line or modem), then don't switch to 4.3. It pains
>me to say this because I really despise AIX. The problem is that the
>async support on ACIS RTs is totally broken. This holds true for 
>the 025, 125 and 135 running any version of ACIS. *sigh* So, if you
>switch you'll be getting into an incredible mess.

Let me point out that this may be true for the planar serial ports
and the original 4-port async card, but it is certainly NOT true
for the newer buffered 4-port async card (and presumably for the
8-port card which is also buffered.)  These newer cards use versions
of the NS16--- (I forget the chip name) UART which incorporate
input and output FIFOs on a per-line basis.  The Sep 88 and later releases
of AOS 4.3 have an async driver which knows about the FIFO (and is
backwardly compatible with older cards.)

Anyway, the older ports were a disaster as reported.  However, with the
new card, I have a 19.2kb (!) SLIP line and 3 Telebits (1 dial out, 2
dial in which range from 1200b-19.2kb) attached to my Model *25*.
Approximately once a day when the system is VERY loaded, I'll get a
flurry of "asy0: overrun error" messages (describing the SLIP line)
but the TCP traffic is unaffected, as you'd expect.   (I wouldn't
run NFS over SLIP without setting udpcksum to 1, however!)

Having this happen once a day, it's essentially a non-problem in my eyes.
I expect it to disappear completely with the APC upgrade once/if it arrives.
I hope anyone who is shying away from using AOS 4.3 because of purported
serial line problems should consider the low cost of the buffered 4-port
card (< $500).  It's a nice hardware solution to the problem.  (Hey,
I'd prefer a buffer along with bidirectional DMA, but you take what's
offered.)

>On the brighter side, the 135 with a megapel running 4.3 is quite a
>nice workstation. The 4.3 works quite well (with the exception of the
>async support). Also, the sources are in slightly different form than
>the generic Berkeley distribution, but that's minor.

Even with the enhanced mono card and a Moniterm monitor, it makes
a nice workstation.  An APC certainly is more than the equal of a Sun 3/60,
and when I see how poorly it's been marketed (and how AOS 4.3 has been
generally unavailable) it makes me want to cry.
---
Steve Dyer
dyer@arktouros.mit.edu
dyer@spdcc.com aka ...!{harvard,linus,ima,m2c,rayssd}!spdcc!dyer