Grex Coop10 Conference

Item 10: Yet another ISDN proposal

Entered by jared on Sat Jun 21 22:55:33 1997:

New proposal:

        So you've all read the previous ISDN proposal, here's the next one:

        Dual channel ISDN from behind the t1 that I'll have at home about
the beginning of september.  My t1 will be to a dual-homed hub between
digex and michnet.  This proposal is independent of CICNet and/or any of
my future employers.  Here's what you'd have to do:

        Order an ISDN line on your centrex into the pumpkin
        Order an ISDN line on your centrex to my place of residence (Haven't
found one yet, ugly long story, but it'll be in a2).
        Get two pipeline 50's (or whatever you want), attach them to my
ethernet port, and I'll route you some ips.

        Comments?  I can go into technical details elsewhere as to exactly
how this will work, along with possible network diagrams, etc..

68 responses total.

#1 of 68 by srw on Sun Jun 22 05:59:16 1997:

It's a generous offer.


#2 of 68 by valerie on Sun Jun 22 13:11:59 1997:

This response has been erased.



#3 of 68 by janc on Sun Jun 22 14:54:46 1997:

Terrific offer.


#4 of 68 by jared on Sun Jun 22 15:27:34 1997:

That really depends on the place that I find, I need to keep looking
before my sublet runs out in another month..  More details once I sign
a lease..


#5 of 68 by scg on Sun Jun 22 17:13:07 1997:

This certainly seems like a great offer.  I've got the same questions that
Valerie asked about stability of where you're going to be living, and also
questions about the stability of the deal giving you the T1.  Is that deal
such that you could discuss it publicly, or is that something better discussed
in private?


#6 of 68 by jared on Sun Jun 22 17:18:50 1997:

I can discuss publically where my t1 will go, but other network details
past that router are kinda tricky and complicated..

My t1 will go to dorians t1 he has at home, and he will be dual homed (at
least) between a few providers.. I can draw a map at the BoD mtg or afterwards,
etc. for people to look at.


#7 of 68 by dpc on Tue Jun 24 18:26:01 1997:

Very nice!!


#8 of 68 by kaplan on Sat Jun 28 15:58:06 1997:

I don't know what you mean by "dual homed".  Does that mean that if Michnet
stays poorly connected to the net, you'll be linked to both Michnet and a less
poorly connected provider?  Who is dorians? Would grex have to trust him to
not packet sniff grexers' passwords?


#9 of 68 by jared on Sat Jun 28 18:27:46 1997:

I give you my word that I and Dorian don't have the time to waste
sniffing grex passwords for grins and giggles of breaking
into peoples accounts.

At the BoD meeting, the connectivity was explained, here it is in
technical terms:
Dorian will be dual-homed (connected to two ISPS) speaking eBGP to both
of them, so packets to grex will take the shortest ASN path to get
to grex, and the same outbound shortest ASN path will be taken also to
return your packets to the site.

Non techie terms:
Grex's ISP will have two major backbone ISP connections (initally),
and be able to be reached as long as nothing bad happens to any of
the direct-connect phone lines that will interconnect dorian's
and my houses together.


#10 of 68 by scg on Sat Jun 28 19:01:19 1997:

In other words, Jared and Dorian will be a lot better connected than most
local ISPs are.


#11 of 68 by jared on Sun Jun 29 04:36:06 1997:

BTW, i do understand the need to bring up the security of the networks
that you will be attached to...  Please feel free to bring up any
more you might have either in e-mail or here.. to give you a brief
idea of what I'll be doing (by default), is source filtering, and spoofing
filtering.. that's not really a discussion here, but notable since those
tend to be the typical attacks you might see in either direction, either
from grex out or elsewhere in.


#12 of 68 by awijaya on Tue Aug 19 16:10:11 1997:

Hello Jared Mauch, can you explain the router/modem required for connection?
Perhaps you can send uuencoded pictutres in several parts?
Perhaps I can get some real estate investment in Michigan?/Property?
Regards ( AW)


#13 of 68 by scg on Tue Aug 19 19:09:39 1997:

For a picture of an Ascend P50, see http://www.ascend.com/202.html.


#14 of 68 by jared on Thu Aug 21 19:49:06 1997:

yup


#15 of 68 by dpc on Fri Aug 29 18:25:49 1997:

Can anyone give us a reasonable estimate of a date by which our
new ISDN line will be open for users?


#16 of 68 by kaplan on Fri Aug 29 22:32:03 1997:

Being a pesimist, I expect the ISDN link to be up and running in one month.
It would be reasonable to expect it sooner.


#17 of 68 by scg on Sat Aug 30 05:11:37 1997:

Ameritech has installed the lines at both ends.  I'm planning to wire from
the Pumpkin's phone room to the Pumpkin sometime this weekend.  The person
doing the wiring on the other end should be doing it sometime in the next
week, I think.  After that, it will be a matter of configuring the routers
and plugging them in.  Once I get my hands on the routers that probably means
about half an hour or an hour of work.  I think they were ordered yesterday,
so they should be showing up any time.


#18 of 68 by janc on Sun Aug 31 12:58:24 1997:

Sounds right.  The most optimistic estimate would be sometime in the latter
half of this week.


#19 of 68 by valerie on Sun Aug 31 13:41:04 1997:

This response has been erased.



#20 of 68 by jared on Sun Aug 31 20:12:33 1997:

I can probally get the connection up a bit sooner even without dorioans end
being done by passing it to unused isdn lines at work even for just outgoing
traffic


#21 of 68 by senna on Sun Aug 31 23:53:17 1997:

I have a question that I don't believe has been answered:  Will we be
increasing the number of telnet ports available when the new connection is
installed?  


#22 of 68 by richard on Mon Sep 1 01:28:46 1997:

I hope so, then we can tal about losing the damn countdown que!


#23 of 68 by orinoco on Mon Sep 1 02:24:50 1997:

Amen to that.


#24 of 68 by robh on Mon Sep 1 02:49:00 1997:

I doubt we'll ever get rid of the queue, since any increase in telnet
ports will eventually be swallowed up by increased usage.  (I remember
when 48 ports seemed like way too many...)  Hopefully an increase in
ports will at least mean that we won't see the queue used for a while.


#25 of 68 by scg on Mon Sep 1 04:16:53 1997:

In addition to the bandwidth problems, there are also system resource problems
with increasing the number of telnet ports too much.    We should discuss that
when the new connection goes in.  I won't say we definitely should increase
the number of telnet ports right away with the new connection, but we
certainly should once we have the new computer online.

Richard, do you have a better solution for handling situations where we have
more users wanting to get on than we can support?  Please explain, in detail,
how it would be implemented and what advantages *and disadvantages* it would
be likely to have.


#26 of 68 by davel on Mon Sep 1 11:24:22 1997:

Right.  I basically could not ever get in, at all, via telnet before the
queue.  I suppose I could write a Procomm script that would attack-telnet,
but that hardly seems like a better solution.


#27 of 68 by janc on Mon Sep 1 14:57:09 1997:

This was discussed in another item.  Basically the conclusion was that we will
probably increase the number of people who can simultaneously telnet in, but
we will not eliminate the queue.

The increase in the number of slots will be modest at first.  We need to see
how well Grex's CPU can handle the added load.  If we can get the 670 running
as well, we will be able to make more substantial increases.


#28 of 68 by valerie on Mon Sep 1 15:08:02 1997:

This response has been erased.



#29 of 68 by richard on Mon Sep 1 22:53:42 1997:

#24...see I dont or didnt knwo this...I never had a problem telnetting
in prior to the countdown...neverused any programs or anything...rarely
ever got the "all ports are busy"  And if I did, it was just an extra
clicl of the mouse to get right back in.  I had worse problems telnetting
into mnet than I neverhad into here.

Now I come on in the morning and Im number forty in the countdown and 
get timed out half the time when I get to ogin because I
dont want to stare atmy grex screen like a zombie waiting for the que.


#30 of 68 by senna on Mon Sep 1 23:58:22 1997:

i hated the old system... took me ages, if I got in at all.  essentially, I
had no choice but to telnet to grex every five seconds waiting for a port to
free up.  At least with the queue I can leave for a bit.


#31 of 68 by valerie on Tue Sep 2 00:09:21 1997:

This response has been erased.



#32 of 68 by orinoco on Tue Sep 2 00:55:31 1997:

I also have never run into the problem of telnet not accepting me prior to
the queue.  Now I dial in almost all the time, so this really doesn't affect
me much, but...


#33 of 68 by janc on Tue Sep 2 16:04:05 1997:

There may be some modicum of truth to that.  With the old system, if people
got "all ports busy" a few times, they probably went away.  Now, if they get
"you are number 10 in the queue" they are more likely to wait around and
eventually log on.  More people logging on means more people on the system,
so the periods of full utilization are longer than they used to be.

Basically, if you make it easier to wait for a connection, more people will
wait, so there will be fewer times when there is no wait.  This definately
means better utilization of Grex, with more users on more of the time, but
it could lead to the perception that Richard and Orinoco report of Grex being
harder to get onto.


#34 of 68 by richard on Tue Sep 2 16:18:56 1997:

#33...thats what I mean, if it increases usage load overrall then
the que defeats its own purpose.  Plus nobody wanting to run
new use is going to wait around if they are down in the que somewher
like fifteen or twenty.  At some point, the curiousity factor is outweighed
by the time/trouble it takes to get in.

check how often new user was run on average before the que was installed
against how often it is run now...bet you'll find a difference


#35 of 68 by valerie on Tue Sep 2 19:09:32 1997:

This response has been erased.



#36 of 68 by aruba on Tue Sep 2 20:34:27 1997:

Re #34:  You miss the point of the queue, Richard - the point was to be more
fair to people trying to telnet in, the same way that it's fairer (and more
civilized) to stand in line at the bank that it is to mob the tellers, with
the biggest person getting in first.


#37 of 68 by robh on Tue Sep 2 22:00:19 1997:

Yep, and the manager(s) of the bank should ignore the ranting of
the big people.  >8)


#38 of 68 by senna on Wed Sep 3 01:28:05 1997:

That's funny, my rampant use of newuser has slowed down quite a bit :)  


#39 of 68 by valerie on Wed Sep 3 02:08:31 1997:

This response has been erased.



#40 of 68 by dang on Sun Sep 7 17:10:24 1997:

Actually, if the queue is causing grex to be used more fully more of the time,
then I would maintain that it is working great.  I would be happy of grex was
full *all* of the time.


#41 of 68 by valerie on Tue Sep 9 17:14:12 1997:

This response has been erased.



#42 of 68 by dang on Tue Sep 9 23:50:08 1997:

Because that would mean that people are using it.  (Okay, so that's obvious,
but...) It means that Grex still has room to grow, that people are
communicating.  It means that Grex is a little slower, and so it increases
the effect of that filter that selects for people who *like* it here, rather
than people who use Grex just for email.  It means a higher chance that people
will wander into conferences, and so more people to talk to here.  It makes
me feel better about donating my time to Grex.  Afterall, I probably wouldn't
want to donate my time to a system that isn't used.  It means that I feel
better about donating money for better hardware and better internet
connections.  They're likely to be used.  I've never said to myself "I think
I won't Grex today, because it's too slow."  I have a multitasking computer,
and a multitasking brain.  If Grex is slow, I do something else while I'm
waiting.  So, more people on Grex is a good thing, in my opinion.  If there's
a time of day when not many Americans are on, I'm happy to see people of other
nationalities from other parts of the planet filling Grex up.  This is also
why I think the queue is a great thing.  It allows everyone equal access to
Grex, even while Grex is full.  It allows Grex to be fuller.


#43 of 68 by mta on Wed Sep 10 00:24:57 1997:

Amen, Daniel!


#44 of 68 by davel on Wed Sep 10 10:58:14 1997:

Well, *most* of the time lately Grex hasn't been slow enough to scare people
off, I think.  But there comes a point when that *is* an issue, somewhere
before you start getting 10-second lag for every character you type.


#45 of 68 by senna on Wed Sep 10 23:41:45 1997:

I've only occasionally been "scared" off of grex when lag is really, really
bad, usually because there's very little else to do due to computer problems
of my own and because I'm spoiled by dialin speeds.  I usually multitask my
eyes out when I'm bored.  It works for me, and I'm not the only one who does
it.


#46 of 68 by valerie on Fri Sep 12 21:36:21 1997:

This response has been erased.



#47 of 68 by senna on Sat Sep 13 02:42:54 1997:

I've had to do it;  It's annoying.


#48 of 68 by valerie on Sat Sep 13 13:25:34 1997:

This response has been erased.



#49 of 68 by richard on Sat Sep 13 20:18:00 1997:

Hopping into and out of mail on grex, its like wearing twenty pound 
weights on each leg.  You can do it, but it is unbearably slow.

I cant imagine using grex for my regular email...I type so fast that 
grex (or rather Pine on grex on rare occasions when I use it) takes 
forever to catch up.  I've found myself three and four paragraphs ahead. 

I wouldnt recommend grex for email to anyone...use a .forward file and 
do email elsewhere, even if its m-net or nether or some horribly 
over-commercialized place on the web like Hotmail.   Grex has to be one 
of the worst places to do email on the entire 'net.


#50 of 68 by richard on Sat Sep 13 20:22:01 1997:

Of course once grex is ISDN-connected, that may change.  

(I actually multi-tasked grex this morning...I hadnt read Agora in two 
days and had 25 items to read, so I did a Backtalk vs. Picospan contest 
and ran Agora on both at the same time, just to see which got me through 
it faster.  And..drum roll...Backtalk won.  Normally Picospan would've 
won easily but I guess it was a bad morning to be telnetting *shrug*)


#51 of 68 by scott on Sat Sep 13 23:06:21 1997:

Richard, Grex can do mail faster, with a different mail program.  Pine is the
slowest thing ever written.  "mail" (the program) may be less friendly, but
it is a heck of a lot faster.


#52 of 68 by orinoco on Sun Sep 14 00:55:56 1997:

Yeah, I've always used mail, and it's slow to load, but no slower than
anything else is on grex.  Certainly not unbearable.


#53 of 68 by valerie on Mon Sep 15 02:14:53 1997:

This response has been erased.



#54 of 68 by senna on Mon Sep 15 06:14:50 1997:

I type fast enough to have similar problems with richard when I type
constantly, but usually I type a bit slower because my thoughts aren't all
that gathered until I put them out.  It's nice to be able to type fast enough
to keep up with my thinking.  I used to outrace grex enough that my old telnet
program would lock me out and freeze.  It was the original impetus for getting
multiple logins.


#55 of 68 by valerie on Mon Sep 15 14:09:57 1997:

This response has been erased.



#56 of 68 by richard on Mon Sep 15 23:39:06 1997:

why would running two editors at once hurt my .cf participation file?


#57 of 68 by mdw on Tue Sep 16 00:42:53 1997:

Running two editors at once won't, unless you're editting your
participation file.  If you are editting your participation files by
hand, then one editor is more than sufficient to damage the contents.

I presume, though, that you mean running PicoSpan and Backtalk.
PicoSpan reads your participation file when you enter a conference, and
writes it out when you leave.  If you run 2 copies of PicoSpan on the
same conference, neither will know what the other has read, and when you
leave the conference, the items that were marked seen for you will
reflect whichever copy of PicoSpan last wrote the file out.  If you were
to leave the conference "at the same time" in both copies, then it is
possible, although unlikely, that you could end up with a corrupted
participation file (in practice, running out of disk space, or an I/O
error, are more likely causes of a damaged file).

When you use backtalk, the rules are different.  I don't know exactly
what Jan did, but most likely, most of the time, when you're reading
stuff via the web, there is nothing at all running on grex.  Only when
you do something that causes your web browser to GET or POST a form on
grex, is there any opportunity for backtalk to update your participation
file.  Most likely, therefore, backtalk updates your participation file
*each time* you read a new item.  So, if you had 2 copies of backtalk
running at exactly the same time, it's possible (but unlikely) that they
could trash your participation files.  There are actually more ways for
it to fail, because backtalk has to both read *and* write the file.  So,
bad things could happen if the second session is reading the file before
the first session finishes writing it.

If you ran both backtalk and picospan at the same time, but only did
things with one at a time, then, most likely, PicoSpan will not "see"
any changes made by backtalk while PicoSpan is in that conference, and
when PicoSpan leaves that conference, any items marked "seen" only by
backtalk, will revert to whatever state PicoSpan has remembered.


#58 of 68 by mta on Tue Sep 16 02:11:51 1997:

Which will look close enough to "trashed" from your perspective that 
it's better not to do it.


#59 of 68 by srw on Tue Sep 16 18:35:37 1997:

The main reason not to fo it is that neither will know what the other 
has seen. The result is that it is not updated properly. This is not 
really *corrupt*, but it is bad. 

Marcus is right that backtalk updates the participation file more often 
thatn picospan. Piocspan keeps its updated copy in core, and only writes 
it out when you switch conferences, exit, or explicitly request it. 
Picospan is thus more efficient with respect to file accesses, but is 
more prone to leaving the participation file out of date when there's a 
crash or other catastrophe. 


#60 of 68 by dang on Wed Sep 17 22:21:17 1997:

(Actually, picospan writes my participation file at the end of every item for
me, so backtalk and picospan would both be writing once an item.)


#61 of 68 by valerie on Thu Sep 18 04:04:11 1997:

This response has been erased.



#62 of 68 by janc on Thu Sep 18 14:45:45 1997:

Backtalk does read and write the participation file each time you read
an item.  It holds a lock on the participation file between those two
events, and I'm reasonably sure the locking is compatible with Picospan.
I presume Picospan, like Yapp, only locks while reading or writing, not
between.  Thus, if you read with both at once, some of the things you
read with Backtalk may show up unread again, but everything you read
with Picospan should be OK, since Backtalk will not clobber Picospan's
data.  I've often used both to read at once (using "set autosave" in
Picospan) and haven't had any big problems.  It seems to work better
than I would expect it to.  I don't know why.


#63 of 68 by janc on Thu Sep 18 14:48:18 1997:

Oh, by the way, using Backtalk and Picospan simultaneously to read
*different* conferences is 100% safe.


#64 of 68 by mta on Thu Sep 18 18:05:45 1997:

Which is they way I've always done it.  It's very efficient use of time. 
 Dunno, though, how efficient a use of Grexes resources it is, so I've 
stopped.


#65 of 68 by mdw on Fri Sep 19 15:24:45 1997:

PicoSpan never locks participation files, and won't respect
any file locks that might be present.


#66 of 68 by janc on Sun Sep 21 14:26:34 1997:

That makes things a bit scarier.  I would thus advise not using Backtalk
and Picospan on the same conference at the same time unless you don't
care too much about your participation files getting trashed.  The same
goes for two Picospan invocations, but not for two Backtalk invocations.


#67 of 68 by kaplan on Wed Sep 24 22:39:58 1997:

Backtalk in anonymous mode would be perfectly safe.  Run Picospan and as many
anonymous Backtalks as you want.


#68 of 68 by valerie on Tue Sep 30 22:31:14 1997:

This response has been erased.



There are no more items selected.

You have several choices: