[comp.sys.apollo] apollo netnews summary

watcher@athena.mit.edu (chris ross) (09/26/89)

Several weeks ago I submitted an article requesting information about
UseNet software for Apollos.  I received responses from the following
people:

	Tim Barrios <asuvax!gtephx!barriost@ncar.UCAR.EDU>
	David Boreham <davidb@brac.inmos.co.uk>
	Keith Cantrell <cantrell@attctc.DALLAS.TX.US>
	Dave Clemans <dclemans@mentor.com>
	Dennis Cottel <dennisc@peanuts.nosc.mil>

I attempted to reply to all messages received, but some replies seem
to have disappeared into limbo (i.e., no response from the other end,
and no mailer errors.)

The impression I get is that both B and C news will work under SR10.1
with only minor modifications.  The most notable problem is that of
multiple writers to the active news file -- Aegis does not permit
cowriters access from different nodes.

I'd like to describe my own experiences with the news software, except
I haven't put it together yet; my responsibilities as an application
programmer are taking up all my time at present (and the "official"
sysads either don't know C or Unix or don't care about UseNet.  Grrr.)
In addition, given the number of bugs I've found in SR9.7 and 10.1, I
may wait until SR10.2 to tackle it.

Listed after my .sig are excerpts from the messages I received.  The
total volume is sufficiently small to not require a detailed summary
on my part.

Thanx to all who replied.  I'll be in touch when I get it running.

________________________________________________________________
chris ross  <0>  uunet!codex!watcher  or  watcher@athena.mit.edu

====================

From: asuvax!gtephx!barriost@ncar.UCAR.EDU (Tim Barrios)
Date: Thu, 24 Aug 89 08:22:43 mst

we are using a version that our local Apollo office got for us from
the Seattle office (i think).  it has some problems.  for example,
notice that this mail (and my postings) come from my actual node
name ('louisiana') rather than the name of our usenet site ('gtephx').

	[in a later message]
    the actual node only appears on postings, i guess.  we have fixed our
    sendmail file (which is why it doesn't show up on the mail) but there
    is a special constant in the news configuration where you can tell it
    all of your TCP/IP nodes should come from the same news site.

i did re-compile the source to avoid these problems by changing certain
defaults but then the 'rnews' of incomming news was taking more than
a minute or so to incorporate each news item into our news.  the made
it so that we could never catch up on incomming news items.

    [in a later message]
    i did a LOT different in that version.  the biggest difference is that
    the version i got from Apollo was compiled under BSD and I wanted mine
    compiled as SysV since we are primarily a SysV site

====================

From: David Boreham <think!ames!hplabs!davidb@inmos-c>
Date: Thu, 24 Aug 89 21:03:09 bst

 We have SR10.1 and 9.7 nodes. The 9.7 do not run news and are
 due to be upgraded soon. The Apollos run the NNTP version of 
 rn. This compiles fine under SR10.1 (not under 9.7), you need
 to correct a couple of bugs in the source code---Sun's compiler
 is less strict than Apollo's. We got the gode off the net (patches
 to the standard rn source). 

 This NNTP rn gets its stuff over TCP/IP from a big SUN server
 which holds all the news and does the UUCP business.

 The only problem I found was that rn does a fancy hack to 
 find the current directory by hopping up the filesystem
 untill it can't go any higher. Then it cd's back to where
 it thinks it started from. Unfortunately, with the apollo
 network root '//', this scheme does not work. All you need
 to do is select the "don't-use-the-clever-scheme-use-pwd-unstead"
 option, then correct the bug in the "pwd" code and you're going.

====================

From: attctc!digi!kcantrel@EDDIE.MIT.EDU (Keith Cantrell)
Date: Wed, 23 Aug 89 18:23:36 CDT

If you are planing to receive news from one of your 10.1 machines, then it does
not take much to get just plain bnews-2.11 from run.  The same goes for 'rn'.
The first problem that I found with both of these were that they expect that
if you have the environment variable "ORGANIZATION" set to something, that
that is what you want to be in the "ORGANIZATION:" field of the header message
of any notes that you post.  Unfortunately Apollo defines it to be the
organization part of you login.

[notes about locking the active file; see next message]

====================

From: dclemans@mentor.com (Dave Clemans @ APD x1292)
Date: 23 Aug 89 15:42:15 GMT

The big problem you see with the news software on Apollos (provided you
set it up so that all the Apollo nodes look like one big "pseudo" machine,
which is what most people I've heard of have done) is that the Usenet
software expects standard Unix file locking semantics, that is no locking
at all.  Apollo's enforce exclusive file write access across a network.

We run Cnews as the Usenet "transport" service.  I didn't have any real
problems in bringing the recent release of Cnews up; it's been quite
stable.  Note that if you compile Cnews as a Sys5 program you WANT to
use one of the public domain dbm library emulators; what the Cnews
build script gives you by default to replace the standard dbm libraries
is an extremely slow sequential search.

I have all the "standard" news reading programs up; readnews, vnews,
rn and vn.  To work around the file locking problems they actually read
an "active.copy" file instead of "active".  "active.copy" is updated
from "active" every 15 minutes by a cron job that runs on our mail/news
gateway node.  rn was also modified to not hold the active file continually
open (which would prevent active.copy from being updated from active...)

I tried once to compile nn (under sys5) but it had problems with hanging
after reading one character.  I haven't had time to look at it to see why,
or to try to compile it instead as a bsd program.

====================

From: dennis@peanuts.nosc.mil (Dennis Cottel)

Although our experience won't be directly applicable, you might like to
know what we are doing.  On our network of 40 SR9.7 nodes, we are using
two newsreaders via NNTP to our organization's central non-Apollo news
server.  One is rn; sources came from our local standard BSD4.3 rn
distribution modified to run with NNTP, with some minor customizations
from Woody Kellum obtained over the network.  The other is Kim Storm's
nn which ported straight onto the Apollos right out of the box.

[note: dennis forwarded woody's changes to me in another message;
 contact me if you want a copy]