Path: utzoo!attcan!uunet!husc6!mailrus!nrl-cmf!ames!ucsd!brian
From: br...@ucsd.EDU (Brian Kantor)
Newsgroups: comp.sys.att
Subject: WIN TCP/IP for 3B15 considered
Message-ID: <828@ucsd.EDU>
Date: 21 May 88 15:02:58 GMT
Organization: The Avant-Garde of the Now, Ltd.
Lines: 79
We just installed the latest version of AT&T WIN TCP/IP for our 3B15,
and got it working, more or less. ARRRGH!
Things to watch out for while installing it:
1. Don't install it twice. If the first attempt at installing it fails,
go back and remove everything before you start over. Guaranteed it
won't install the second time without an intervening remove. If you had
the beta test version, you'd better yank that out first or all kinds of
things in the installation script will come unstuck. Make sure that
none of the codefiles it's going to try to write on top of are running;
perhaps going to single-user mode is worth doing, although the manual
doesn't suggest that.
2. If you already have any of the files used by this software, it will
just cheerfully overwrite them; this is particularly true of /etc/hosts,
/etc/networks, and /usr/lib/sendmail.cf. Do a full backup of your /usr
and root filesystems FIRST so you can put your versions back. You have
been warned.
3. After you cpio the installation script in, go modify it so you can
see what it's doing so you can undo anything you don't like it doing.
I stuck a 'set -x' in the top of it. I consider a "please wait" message
inadequate progress reporting, and the damn installation script seems to
remove itself after completion (at least, I couldn't find it) so there's
no reasonable way to find out what it did after the fact.
Things to do to make it work:
The script that goes in /etc/rc.d/rc2.net is broken. It starts the NI
interface, then checks to see what it's status is. If you are doing
ANYTHING except coming up from power-on, it may well find that the
interface is already running. Since it's looking for the "just started"
status return, it considers this an error, doesn't start the daemons,
and you're stuck. I just commented out the test, so we now just tell
the interface to start and keep on going. I've never had the interface
fail to start, so this gamble seems worth it. The manual doesn't seem
to list the status return codes for the nistart/nistatus programs, so
I can't fix the script to work properly.
Things you can't easily do anything about:
The version of sendmail delivered is 4.12; which according to the
Version log in our BSD sources was current on Sept 7, 1983 (yes, FIVE
YEARS AGO). It's so full of bugs that you're lucky if mail gets
delivered. We get around five or six messages a day from the mailer
daemon telling me that /bin/mail was invoked from sendmail with bad
command line arguments; luckily this is being forwarded to me on a BSD
system where I can forward it on to the proper recipient.
If you have any hope of using this machine for ethernet mail
you'd better plan on porting the current version of sendmail or maybe
replace it with MMDF or something. You can get the latest version of
sendmail by anonymous FTP from ucbarpa, but it's configured for Berkeley
systems and will take some work to port.
Forget about subnets. This software has no idea what a subnet is.
There isn't a nameserver. Even if you grab one of the ports of BIND to
SysV, you're stuck, since you can't recompile the software to take
advantage of it. And sendmail 4.12 never even heard of an MX RR.
-------
Understand, this isn't meant to be a flame (although that might be
appropriate...). Except for sendmail, what's arrived now seems to
be working reasonably well. It seems obvious to me, however, that
this is a product that hasn't been upgraded at all recently - I'm
wondering how much time elapsed between the time Wollongong
delivered it to AT&T and when we got it last week. I suspect it's
been a couple of years.
It's a good thing we don't have serious users on this machine (just a
few students doing independent research) or I'd have to fix this stuff
- probably by porting over as much of the 4.3 code as would work. But
it's a whole lot better than nothing, and much less hassle than doing it
ourselves would have been, so we aren't going to send it back. It sure
could have been nicer though. Caveat, as they say, emptor.
Brian Kantor UC San Diego
Path: utzoo!attcan!uunet!lll-winken!netsys!len
From: l...@netsys.UUCP (Len Rose)
Newsgroups: comp.sys.att
Subject: Re: WIN TCP/IP for 3B15 considered
Message-ID: <8478@netsys.UUCP>
Date: 21 May 88 16:54:09 GMT
References: <828@ucsd.EDU>
Reply-To: l...@netsys.UUCP (Len Rose)
Organization: NetSys,Germantown Md.
Lines: 10
No other comments to make but this one.. I have stashed
sendmail for system V on ames.. it is available by anonymous
ftp at ames.arpa ..
Len
--
Len Rose - l...@ames.arc.nasa.gov
or {ames,decuac,ihnp4}!netsys!len
"This better not be haga!"