We gotsta get Usenet news back online, pronto!

Feb 3 05:27 UTC 1995

I would really appreciate it, if we could expedite the process of getting
Usenet news back on line.  Telnetting to msu or unc is slow and cumbersome,
and often not reachable (busy).  Besides which, we need posting capability.
I know there have been offerings of ways to e-mail postings to a service that
will post for you, but I've tried it 2 or 3 times w/o success.
I am at a critical point in developing self-employment, and I can fan the fires
if I could post.  It's getting near the point where I'm going to have to join
AOL or some other commercial outfit, and I really can't afford it.
I know staff has alot of critical jobs in front of them, not the least of which
is getting more phone line, but I would hope they could front-burner this
pressing need.  It seems like it's been six months that we've been w/o 
Usenet.
Thanks-  Chicago Taxi Willie

response 2 of 124:

Feb 3 07:21 UTC 1995

   At the moment, two things stand in the way of getting news up.
1) router problems, which Marcus and I are working on right now, and
2) which is the disk problem.
   When we're stable in those two areas, we can determine where we
want to go next.  I believe that news is the right project after
this, but its going to take work to get there, no matter how we
decide to do things.  News has gotten so incredibnly resource
intensive that we *can't* run it on Grex anymore.  We'd only
hurt Grex if we tried.  So we are going to have to get another
machine in place for that.
   But first, the other two things need to be solved.  Right now I'm
going to go back and see whats happening with FreeBSD.

response 4 of 124:

Feb 3 08:54 UTC 1995

   Just about all, if I remember right.  If we want to start thinking
about this, we should think of about $1500.
   (More later when I'm not splitting my mind between the
router and Grex).

response 6 of 124:

Feb 4 15:27 UTC 1995

good idea!

response 8 of 124:

Feb 4 19:38 UTC 1995

        If we get usenet back it might almost pay for itself in ioncreased
        memberships.  I bet there are many non-members on the system now who
        would become
members if they had usenet access.  I know we aren't a service provider and
people *ought* to become members (as I have) to supports Grex's community
service.  Nonetheless I bet the first statement is the *truth*.

response 10 of 124:

Feb 4 19:51 UTC 1995

I don't see any reason to limit it to members, especially since we
plan to have it on our own disk (aren't we?). It is just looking up
a file. And doesn't our new policy permit posting to verified  users?
Of course, it would *still* pay for itself in increased members, as
we have shown that many persons voluntarily support a system they like.

response 12 of 124:

Feb 5 00:14 UTC 1995

In what meaningful sense (meaning equipment--if not the Sun 3/260,
then what) can we get news running again?

response 14 of 124:

Feb 5 00:33 UTC 1995

So are we talking a second Sun for mail, in addition to the 486
news server?  Or would the 486 be able to also handle mail?

response 16 of 124:

Feb 5 12:59 UTC 1995

I have to dispute Marcus's assesment in #13.
If for no other reason, than we were doing news over a crappy router
with just a Sun 3/260 and 40 users for almost 6 months until we turned
news off because of the disk problem.
1.) Yes, a second phone line for the news feed would be *nice*, but not
    critical. PPP and SLIP have a prioritization system built into them
    that causes telnet and rlogin packets to always go first. The upshot
    of adding news to the link would be that the news would come in slowly,
    but I think it's impact on the interactive login processes would me
    minimal.                                                         ^^be
2.) Yes, it would also be *nice* to have a 486 box for news, but it could
    be done on the Sun. This, I will admit, *will* load the Sun down to
    a much greater extent than #1, so we probably don't want to it. OTOH,
    there is another alternative. We already own enough parts to make 2
    more functional Sun 3/260's. At least 2. All we need is disk. This
    would be alot cheaper than building a 486 system.
3.) Upgrading the Sun to a 4/260 is simply not a viable short term solution.
    I've gone around and around with steve on this one. It took us *months*
    to get the current Sun properly set up. And that was with a significant
    piece of staff resources(Marc and myself) who wouldn't be available
    for a conversion. The only route for this is to set up one of the *other*
    Sun's with a 4/200 CPU, and then gradually build/install/port all the
    code we currently have on Grex over to the new system. This is doable,
    but will take awhile to get everything done right, so it's a long-term
    solution.
    However, I *will* agree that when we turn news back on, we make people
    use trn or rn for now. Tin is nice, but we don't have the resources
    for it.

Another item I'd like to point out is that the main and only place we
really need to worry about performance is interactive user processes.
I've heard suggestions that we need 1 machine for mail and another machine
for news and another machine for logging and maintainence tasks. The main
thing we need is another machine to unload *all* the background processing
onto to remove as much load as possible from the main box. When a user
hits <CR> and then sits drumming his fingers for 15 seconds waiting for
a prompt, that is *bad*. OTOH, if the mail/news/background-processing/
whatever box gets a little overloaded and it takes 1 minute to process
a piece of mail instead of 10 seconds, who cares? That kind of thing
happens in the background, asynchronously.

The main need is for a second machine to dump all our non-time-critical
processing onto. After that, as time and financial resources permit, we
*can* add more machines.

So, at this point, I believe staff's current priority list would be(someone
on staff hit me if this isn't right):
1.) Get the Router stable and functioning.
2.) Fix the damn disk problem once and for all.
3.) Once we are certain the disk is truly stable, we then need to go back
    and re-enable a number of things that were turned off while trying to
    fix the disk.
4.) We also need to restore a great deal of Grex's OS from a known good
    source. Much has become screwed up in the last 6 months.

After that's all done, my reccomendation would be to get one of the other
Sun's running, but that's just my call.

response 18 of 124:

Feb 5 16:51 UTC 1995

This discussion should be linked to coop (planning).

response 20 of 124:

Feb 5 17:18 UTC 1995

Steve(srw), I agree that news *processing* would "bog down Grex" but
the transfer of that news over the link shouldn't. IP traffic over the
link tends to be "bursty", but I suspect that our average utilization
of the link is not as high as we think. News would also come in 24 hours
a day, the link tends to be fairly quite in the early AM hours. It would
be useful to this discussion if we could get some 24 hour statistics on
average network load over the link, see just how much bandwidth we are
*really* using, otherwise it just becomes a "my hunch vs your hunch:
kind of argument.                                                  ^"

response 22 of 124:

Feb 5 21:46 UTC 1995

   Why is stating a thought a moot point?  I don't agree with what Bill
is saying, but I don't see any reason for him not to say things.

response 24 of 124:

Feb 5 23:48 UTC 1995

Larry, check out the history conference.  To get ther, type "j history"
at an Ok: prompt.