response 34 of 65:

Aug 10 23:42 UTC 1999

So what's the current block to the number of connections allowed? Is it
bandwidth or proccessing power? If I remember right, Grex has around $5K in
its coffers and around one to three hundred dollars of surplus each month
(this may have changed due to the phone credit running out, so correct me).
$5K could easily buy a few more processors and memory chips (and whatever else
is needed, such as faster disks for /usr and /var), and $100/month can buy
a dedicated dialup (though at pretty small speeds, something like 14.4 or 28.8
kbps); a little math shows that the bandwidth would support around 7-8 more
connections, and I have never seen all of the dialups in use so one of those
modems could be used (or could be bought).

To be a bit more clear from what I had said earlier, maybe we shouldn't
actively try to grow in terms of users; Grex has more users every day, and
grows at a healthy rate. Upgrading the proccessing power for the increased
connections would probably allow for more mail spooling and such. A bit of
retrospect makes me think that I should have said that we should be preparing
for the growth we have, by doing stuff like installing a few extra disks for
the home filesystem as needed (The Pumpkin inventory shows that there are a
bunch of spare 2-gig drives), and eventually (In the far future, probably
around the time the commies take over Russia and they invade and Grex gets
blown to bits and outlawed) Grex will need to get around the 65K user limit...
All of this will tie in well with the transitions to the newer OS's which
we'll need to do some time.

I do agree with Steve that adding more connections right now without doing
anything isn't going to hurt; even at max capacity, Grex generally isn't slow
(although there are some periods of lag), so it wouldn't do anything too bad.
Then again, I remember times when there was a 70-user queue, I went to get
the mail, found out that my login window had expired (Did I mention that my
driveway is a fifth of a mile long?), gotten back into the 70-user queue, and
goten on in 10 minutes.

response 36 of 65:

Aug 11 18:09 UTC 1999

I got a busy signal for five minutes this morning, first time in a month.
I don't see any problems with grex as it is right now.

response 38 of 65:

Aug 15 03:04 UTC 1999

Yetch.

I can think of at least 3 other possibilities to investigate.  (1) DSL
(or whatever the symmetric version of ADSL is called).  (2) fractional
T1.  (3) stuffing a bunch of ethernet cards into a couple of cheap
486's, & teaching them to talk to a bank of ascend pipelines; the idea
would be to send 50% of the packets down one ethernet card to one
pipeline, and 50% to the other pipeline.  We could probably even invest
in "cheap" pipelines with this approach (the pipelines that don't know
about multiple mac addresses and routing), because we can do packet
encapsulation and hide the actual ip address from the pipeline.

The T1 approach is perhaps the most scary financially speaking; but we
have started to collect cheap T1 hardware, so this may yet eventually
become feasible.  The basic approach here would be to wait for a used
cisco router to fall out of the sky on us (which is surely getting more
& more likely as time goes on), and then figure out how we can pay for
the telco expenses, plus delicate negotiations with our ISP regarding
bandwidth and hardware at his end.

response 40 of 65:

Aug 16 00:32 UTC 1999

Well, my coworker who was interested in multiple ISDN lines was interested
in this in part because he has 384k DSL that isn't reliable, at which
point, two ISDN lines would make sense if you could get sane routers
at a sane price.  So DSL was a decidedly disinteresting option to him.

If you can get an ISP that will give you free IP conectivity if you pay
for the local loop to them, the next question to investigate is whether
colocation at that ISP makes any sense.  And I don't know what the
situation in Ann Arbor is with that.  (I was talking to an ISP in
Boston recently who's quite happy to host some random IRC server
for someone, and I would be unsurprised if he would be equally happy
to host something like grex.)

Marcus, there's also something called `frame relay' that you neglected
to mention.

response 42 of 65:

Aug 16 05:59 UTC 1999

I also didn't mention fiber optic cable, although we do have a fddi
interface and could, in theory, run a fiber cable out to our ISP.  (If
someone rich died and left us a sufficiently large pile of money).
We've also considered using a radio link, or an open air infra-red link;
but these both have distance and weather limitations, and need expensive
equipment.

The major issue with colocation is having 24 hour access.  Most of the
time, grex runs quite happily unattended, but when there are problems,
we really *need* that access, and unfortunately, since all of our staff
are volunteers, that generally means late at night and on weekends, when
they have free time.  Another related possibility would be to try to be,
not quite "co-located", but "nearly" co-located, such that we can
basically toss an ethernet cable over the wall.  We investigated both of
these alternatives in the last go-around, and found a few almost-deals,
but there were the usual set of problems that prevented closure, so they
didn't happen (in one case, the ISP went out of business!)  Co-location
isn't really a viable alternative with our current ISP (I *think* the
other end of our ISDN line goes to the apartment of one of the owners,
and he understandably wasn't interested in giving us carte blance access
to his bedroom).

response 44 of 65:

Nov 9 14:43 UTC 1999

As mentioned in the minutes of the October board meeting, a "long-range
planning" meeting is scheduled - hopefully not tentatively, this time -
for Sunday November 21.  Mark (aruba) volunteered to host it.  Details
of time and place will be forthcoming.

It's to be a brainstorming session on Grex's long-term plan, open to the
public.  I'm hopeful that as many interested folks as possible can make
it.  I do suggest that people review the issues brought up in this item
and make further comments here.

I frankly don't know if any concensus for significant change in what
Grex does will come out of this meeting or not.  But the face of online
communications has changed radically since Grex opened for business in
1991, so I agree with others who feel that a look at what Grex can and
should be doing in today's online world is timely.

No "official" action can be taken at the meeting itself; as I say, it's
a brainstorming session.  Any proposals for significant changes that
come out of this meeting would of course be subject to thorough
discussion and review in Coop.

response 46 of 65:

Nov 12 04:40 UTC 1999

Well, so long as an exception is made for Arlo.  I'm not sure we can
convince him only to talk when he has the bottle, and I'm not sure
anyone else wants to handle the bottle after he's done with it.

One thing Grex might consider getting into:  allowing people to create
their own conferences or groups of conferences.  These wouldn't be "Grex
conferences" as such, and they wouldn't show up on the Grex list of
conferences.  Anyone could create one for any purpose, publicizing it
anyway they want.  You wouldn't have to talk to anyone about it, just
type a command and it appears.  We might or might not allow closed
conferences to be created - probably not.  These conferences would be
automatically deleted if they were idle for too long.

response 48 of 65:

Nov 17 03:41 UTC 1999

Ummm...Somebody ought to fix the MOTD so that it reads to look in item # 92,
not item #93 for the information on the FOG meeting....:)

response 50 of 65:

Nov 17 20:39 UTC 1999

Oops, my mistake.

response 52 of 65:

Nov 23 05:35 UTC 1999

Don't forget the cost of the equipment to connect the T1 to, at Grex's end.

response 54 of 65:

Nov 28 07:10 UTC 1999

In terms of monthly charges, there are three components to a T1 that cost
money.  You two local loops (one for each end), and then channel milage
charges for the interconnection between the two local loops, if they go from
different Ameritech COs.  Then you need to pay for any services that ride on
top of the T1, such as phone lines or Internet access.

The local loops come in a number of forms.  A DS0 is one 64K or 56K channel.
A T1 local loop is 24 DS0s.  A DS3 local loop is 28 T1s.  An OC3 local loop
is three DS3s.  And so forth.  All of those local loops except the DS0 can
be ordered clear channel or channelized.  The simplest form of T1 has a T1
local loop on each end, and 24 channels going between the two local loops.
If you have multiple T1s on one of the ends, you can use a channel of a
channelized DS3 local loop on that end instead.  If you wanted to use only
half a T1 for data, and the other half for phone lines, you could order a
channelized T1 local loop for your end of the circuit, a full T1 local loop
at the other end of the Internet connection, and then interconnect the first
12 channels of the local loop at the other end to 12 of the channels on the
channelized T1.  Then you could order 12 phone lines to go on the other 12
channels of the channelized T1.  Alternatively, you could bring multiple
fractional T1s in from various different locations and combine them all on
the single channelized T1, as long as you didn't need more than 24 channels.
There are any number of ways you could configure such a setup.

response 56 of 65:

Nov 28 16:33 UTC 1999

The phone company generally does not supply the equipment for the customer's
end.  If you're ordering the Internet service from an ISP, they may or may
not supply the equipment, but they generally charge more if they do.

At work, we've got a T1 that does part data and part voice.  We've got a 
channel bank plugged into it that spits out 12 of the lines as analog POTS 
lines vi a RJ-11 jacks, and then also has a v.35 port on it, which we plug 
into the router to handle the other 12 channels.  There are also various 
routers and CSU/DSUs that have drop and insert functionality, such that they 
will take the full channelized T1 in one port, use part of it, and send the 
rest out another T1 port.  All of these solutions are probably beyond the
scope of Grex's budget.

response 58 of 65:

Dec 1 05:17 UTC 1999

Yes.  I had occasion to look it up this morning, for a class I'm teaching.
Unfortunately, I don't remember the exact price.  However, while I saw lots
of mention of "two-way Internet", their emphasis seemed to be on "one-way
Internet":  A cable-modem in, and a 14.4 kbps (up to 33.3 kbps) modem out.
Of course, they are aiming at the residential market, not the business
market.  Two-way Internet Service provides 1.5MB in and 300Kbps out.

For a home without cable TV, it's $44.95 per month, and they only
support Win95/98 in Ann Arbor/Ypsilanti; no NT or Macintoshes need apply.
Installation is $99.95.

Or so says their web page, http://www.mediaone.net, although it might be
easier to start at

        http://www.mediaonerr.com/homeframe.cfm

(Make sure JAVA and Javascript are turned on before you start.  I'm not
sure which one it uses, but without them, nothing works.)