An open letter
Noah Friedman
The following is a personal essay and clarification of some of the
things that have been going on around the FSF. To some extent, IRC has
been affected by what we do. This is not an official statement by the FSF
and the opinions expressed here are not necessarily representative of the
organization as a whole or of any of its members (except for myself).
But first, a statement of fact. The user "belladona"
(ta...@gnu.ai.mit.edu) on IRC has not, and never did, give the root password
to the FSF machines to anyone on IRC. She was teasing naive people who
asked for the root password by giving them false ones. Some of these
characters were naive enough to go around walloping "Hey! I've got the
root password!" without even checking to see if it worked. And it doesn't
help that apparently clueful (I guess I was mistaken) people went around
spreading this rumor without checking their facts.
Our machines are (were) not particularly secure. It was trivial to
obtain root access without the password.
Friday afternoon, around 3:00 PM, staff members in the office pulled the
FSF machines off the net, turned off all accounts and made the machines a
bit more externally secure. There was no warning given to guests, staff,
or volunteers working from remote sites.
Starting a month or two ago the amount of destructive and annoying
behavior by some of the FSF guests began to increase. We received
complaints from all over the country about some of our guests breaking into
remote sites, sending abusive mail is massive quantities, and harassing
users on IRC. Probably these same guests were also responsible for
deleting files on our systems and bringing our machines down so that they
were completely unusable.
The staff members who work in the Cambridge office and the board of
directors (with the exception of Richard Stallman, who wishes it to be
known that he does not agree with or accept the decisions we made) decided
that it would probably be necessary to remove the anonymous open accounts
from our systems. We discussed plans for doing so, but only by voice or in
person. We had to implement "email-silence" because we knew that some of
the crackers on our systems were reading our mail. It would not have been
a good idea to let them on to what we were up to. I hope this explains why
we were unable to warn people what was about to happen.
Friday afternoon someone (possibly more than one person) did something
so pointlessly destructive that the people in the office decided to carry
out the actions we had planned on for a later date. I won't go into the
details. The people who are responsible know what I'm talking about.
This won't affect our policy of giving people guest accounts. We like
having guests on our machines and I know that many, many people have
benefitted by the fact they they can use them. Of the hundreds (possibly
thousands) of people who used our machines, probably only 6 or 7 caused any
trouble. Unfortunately, these 6 or 7 people were persistent enough and
obnoxious enough to spoil things for everyone else.
All we've done at this point is to remove the anonymous accounts, and
disabled the other accounts until users can change their passwords. The
crackers had modified various programs on the system and recorded the
passwords of most of the accounts, and this makes it necessary to insure
that they are changed before the accounts are re-enabled. Hopefully, this
policy will allow us to restrict access to our machines by people who are
bent on causing damage.
We do not encourage cracking. We never did. Our open access policy was
originally a way of expressing to crackers that they didn't need to be
antisocial and that "breaking in" wasn't necessary. We welcomed them
(along with anyone else who knew about us) to use the computers here and
tried to encourage them to do something constructive. A lot of the time we
succeeded.
The FSF has always tried to encourage people to do beneficial and
constructive things. That's why the GNU project was started. The idea was
to provide a complete operating system which everyone could use as a base
for writing and sharing software freely. At the same time, the FSF wanted
to teach people that it was possible to share computing resources in an
open environment where people worked toward improving the system as a
whole. While the two issues are not completely intertwined such that one
without the other is impossible, they are reflections of the same general
philosophy. Security is an obstruction which prevents people from doing
this "without permission." By removing our anonymous accounts we now say
to the world "we have to assume that everyone is guilty and untrustworthy
until further inspection." What a sad statement about human nature that
is!
To the people who contributed to forcing us to change the way we think,
thanks heaps. You've caused more damage and unhappiness than you can
possibly imagine.
---
Noah Friedman
frie...@gnu.ai.mit.edu
System administrator, Free Software Foundation
(PS: if you're thinking of asking for an account at this point, don't.
There is too much work to do at the moment for anyone to take the time to
make them.)
Godzilla
frie...@nutrimat.gnu.ai.mit.edu (Noah Friedman) writes:
>
>---
>Noah Friedman
>frie...@gnu.ai.mit.edu
>System administrator, Free Software Foundation
>
See original posting
>(PS: if you're thinking of asking for an account at this point, don't.
>There is too much work to do at the moment for anyone to take the time to
>make them.)
>
Remember people (to quote Cliff Stoll):
"Networks are built on trust"
(quote Godzilla))
"Networks are rooted by morons"
bbs.Godzilla)spies.com
OBSCURED BY CLOUDS
>several weeks, I've received a large number of unsolicited /msg's from
>people I do not know at all and have never talked to before, asking me
>to join a channel, usually +hack. Most of these have come from users
>still accessing IRC from *.gnu.ai.mit.edu sites. Has anyone experienced
>the same thing,
Yes
>and has anyone done this and been victimized at all?
>The most persistent of these folks is one George McBey (g...@pogo.gnu.ai.
>mit.edu). I keep receiving 'msg's from him, largely saying "My old
>friend RadioKAOS! Please join +hack." After telling him that I do not
Yes. Same thing happened to me. I just ignored (not /ignored) gfm
thinking that there might have been another TheP...@the.gates.of.dawn
(IRC Name) somewhere out there in netland. I guess not...
--
Alexander Feygin _____________________________ And everything under
University of Illinois / The probability of someone \ the sun is in tune
email: fey...@uiuc.edu | watching you is proportional | But the sun is eclipsed
IRC: Tylenol or PiPeR | to the stupidity of your | by the moon.
______________________/ action. | --Pink Floyd
--anonymous | "Eclipse"
Sean P. Ryan, Hardcore Alaskan
unfamiliar commands that cause damage, I had a question. Over the past
several weeks, I've received a large number of unsolicited /msg's from
people I do not know at all and have never talked to before, asking me
to join a channel, usually +hack. Most of these have come from users
still accessing IRC from *.gnu.ai.mit.edu sites. Has anyone experienced
mit.edu). I keep receiving 'msg's from him, largely saying "My old
friend RadioKAOS! Please join +hack." After telling him that I do not
ignore whatever it was that I said, and continuing to ask me to join
+hack. It gets a little tiring, to say the least. The only time I join
a channel on request is when I receive an actual /invite, from someone I
know. Anyhow, before I completely ramble on here, any info would be
appreciated.
***************************************************************
Sean Patrick Ryan Bitnet: FSSPR@ALASKA
Internet: fs...@acad3.alaska.edu On IRC: RadioKAOS
Snail: P.O. Box 240554, Anchorage, AK 99524-0554
"And James here, he made a hundred dollar vow of faith. And he
vowed that that hundred dollars would turn into three thousand
through his gambling, and it came to pass. Praise Satan,
hallelujah!" - the sinner's Robert Tilton
Carl v. Loesch
>people I do not know at all and have never talked to before, asking me
>to join a channel, usually +hack. Most of these have come from users
>still accessing IRC from *.gnu.ai.mit.edu sites. Has anyone experienced
>the same thing, and has anyone done this and been victimized at all?
This message went to all people on IRC.
> I keep receiving 'msg's from him, largely saying "My old
>friend RadioKAOS! Please join +hack." After telling him that I do not
>know him, I get a /msg back indicating that he apparently has chosen to
>ignore whatever it was that I said, and continuing to ask me to join
>+hack. It gets a little tiring, to say the least.
The nickname was inserted for each person on IRC.
This is a new dirty technique of misusing ircII.
But it could also be done easily with a little program.
The dirty thing about it is: It generates a message for each person online,
that is about 220 messages. That is a huge netload which can cause network
links to crash! It is extremely asocial of these guys using these methods.
I have been /killing people doing this, but some other ops started flaming at
me, saying normal users finally found their way to /wall, and that's nice.
I also started talking to the guy, since he had an auto-unkill facility
and just kept sending laughs at me! He asked what other use there could
be for /on who? Was it meant to do anything else than that?
Yes it was. It was only meant to change the looks of the /who-output.
That's the way it is used in the scripts, part of the ircII distribution.
Now that we have come to the point, where every 5th individual on IRC thinks
he's got the right to bother all the other >200 individuals, we have a problem!
What hurts me is that I acted in similar ways myself at my first contacts with
IRC, but that shows me that a person isn't just a 'jerk' if he does these
things; he just doesn't realize! It's just a stupid computer he thinks, all
he says is the tiny ascii screen before him, he doesn't think of all the
computers having to spend CPU time for him, all the networks having to
send his messages, all the money these things have costed and still consume,
and most of all: all the other people who have to live with it, with the
netlags, crashes and everything else.
This is not just an IRC problem really, it's just the same with world-
distribution flamers on Usenet, and with GIF collectors and other cross-the-
nets-ftp-freaks.
At this point of net-degeneration I'd prefer to give the /wall command (which
only generates one netwide broadcast, just as much as a /nick change or an
/away message) to the public: everyone can /wall, and everyone /ignores all
/walls except those who enjoy sending and receiving them.
Did you notice that IRC is really going to the dogs now? :-(
--
o----------------------------------------------------------------------o
| Carlo "Lynx" v. Loesch is loe...@informatik.uni-oldenburg.de |
| o=-=-=-=-=-=-=-=-=-=-=o or irc-admin or ly...@dm.unirm1.it |
| :^D :^D :^D ;^) :^D or loe...@uniol.uucp or loe...@uniol.zer |
| ----------------------- or 244661 at DOLUNI1.bitnet / .earn |
`^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^'
It's too late to change events
It's time to face the consequence
For delivering the proof in the Policy of Truth
(Gore)
Greg Lindahl
>Over the past
>several weeks, I've received a large number of unsolicited /msg's from
>people I do not know at all and have never talked to before, asking me
>to join a channel, usually +hack. Most of these have come from users
>still accessing IRC from *.gnu.ai.mit.edu sites. Has anyone experienced
>the same thing, and has anyone done this and been victimized at all?
One user in particular, gfm@*.gnu.ai.mit.edu, has been pretty
enthusiastic about sending everyone messages like this. He seems to
have stopped for the moment. If you see further instances, please send
me email (gl...@virginia.edu) and I'll drop a note to the gnu admins.
Benjamin Bernard
>mit.edu). I keep receiving 'msg's from him, largely saying "My old
>friend RadioKAOS! Please join +hack." After telling him that I do not
gfm is only one of several IRC users who enjoy harrassing people. His
particular approach is to send apparently personal messages to each user
on IRC, using specific information available from the /who reply. This
mis certainly annoying, and also quite effective. Many people HAVE joined
+hack as a result of his garbage messages. It gives him his jollies. The
best approach seems to be to ignore him, and advise anyone else around you
(on your channel, querying, sitting next to you, whatever) to ignore him as
well. If he really bothers yoiu, please send email to the admin of his site
(gfm is on pogo.gnu.ai.mit.edu, as noted above) describing his actions. Maybe
we can get his admin to take some sort off action.
Ben
Mike Pelletier
(Carl v. Loesch) writes:
>
>Now that we have come to the point, where every 5th individual on IRC thinks
>he's got the right to bother all the other >200 individuals, we have a problem!
You're really taking this one (rather big) incident over the weekend and
blowing it way out of proportion. Take a deep breath and a step back, and
calm down, and re-assess things. I would hardly consider us to be at the
point you describe.
>This is not just an IRC problem really, it's just the same with world-
>distribution flamers on Usenet, and with GIF collectors and other cross-the-
>nets-ftp-freaks.
It's good that you realize that, so that you'll begin to see that this isn't
a problem that IRC has any more chance of changing in any sort of technical
way than any of the other mediums do, and that it just comes with the
territory. Perhaps the first step would be for people to start a dialogue
with the abuser, instead of just assuming that he is an evil jerk and /killing
him. Most humans that know how to use a computer will listen to the voice
of reason.
>Did you notice that IRC is really going to the dogs now? :-(
Just like it was last year, and six months ago, and two years ago...
Funny, IRC seems to be in a perpetual state of going to the dogs!
L I G H T E N U P ! ! !
--
Mike Pelletier | "Wind & waves are breakdowns in the commitment of
The University of Michigan | getting from here to there, but they are the con-
College of Engineering | ditions for sailing. Not something to eliminate,
Student/Systems Admin | but something to dance with."
Noah Friedman
>In article <THB-=+K...@engin.umich.edu>, ste...@engin.umich.edu (Mike Pelletier) writes...
>>Perhaps the first step would be for people to start a dialogue with
>>the abuser, instead of just assuming that he is an evil jerk and
>>/killing him. Most humans that know how to use a computer will
>>listen to the voice of reason.
That's what I and others here have tried to do for over a year, and not
just on IRC. Rms has done it for decades. Most of the time it works, but
there seem to be a few hard-headed people who refuse to listen.
>Well, I already brought this point up in my earlier post, but in case
>you missed it, I'll repeat. At least in the case of George McBay, I did
>/msg him after his first /msg to me, but he ignored me and continued to
>send me more of these type of /msg's. Obviously he reads this group, as
>I have seen a post from him recently. But he has chosen to ignore this
>whole issue and the fact that numerous individuals have implicated him
>in this matter. Now, he's not the only one who has done this, but as
>far as I've seen he's the only one who did not quit after doing it one
>time. And that's certainly a problem as far as I'm concerned.
Gfm is the least of anyone's problems on the net right now. He may be
irritating, but everything he's done has been mostly harmless. He is not
responsible for convincing clueless users to run the command "/on ^msg
drdoom $1-" in their IRCII clients, for example.
I don't personally approve of his behavior. He's been warned about it.
But I think it's more important to handle other problems, like the recent
increase in cracking.
---
Noah Friedman
frie...@gnu.ai.mit.edu
(personal opinions only, as usual)
Sean P. Ryan, Hardcore Alaskan
>the abuser, instead of just assuming that he is an evil jerk and
>/killing him. Most humans that know how to use a computer will
>listen to the voice of reason.
Well, I already brought this point up in my earlier post, but in case
you missed it, I'll repeat. At least in the case of George McBay, I did
/msg him after his first /msg to me, but he ignored me and continued to
send me more of these type of /msg's. Obviously he reads this group, as
I have seen a post from him recently. But he has chosen to ignore this
whole issue and the fact that numerous individuals have implicated him
in this matter. Now, he's not the only one who has done this, but as
far as I've seen he's the only one who did not quit after doing it one
time. And that's certainly a problem as far as I'm concerned.
***************************************************************
Troy Rollo
stealth> You're really taking this one (rather big) incident over the weekend and
stealth> blowing it way out of proportion. Take a deep breath and a step back, and
stealth> calm down, and re-assess things. I would hardly consider us to be at the
stealth> point you describe.
IT wasn't one incident over the weekend. IT was a whole chunk of incidents over
a week or more. The only difference with some of the ones over the weekend is
that they were personalised.
stealth> >This is not just an IRC problem really, it's just the same with world-
stealth> >distribution flamers on Usenet, and with GIF collectors and other cross-the-
stealth> >nets-ftp-freaks.
The flaming analogy is possibly valid however the others aren't. Even the
flamers one is on shaky ground considering there has never been talk of
stamping out the news system.
The effects of doing something like this on IRC have a much greater load to
effort ratio than cross-the-net ftping or GIF collecting. Now, if
*everybody* were to start ddoing this, we'd have the same magnitude more
trouble with the critics of IRC.
stealth> with the abuser, instead of just assuming that he is an evil jerk and /killing
stealth> him. Most humans that know how to use a computer will listen to the voice
stealth> of reason.
You obviously don't know gfm and Dr Doom very well. They are evil jerks.
Or perhaps immature jerks would be a better way to phrase it. GFM gets his
kicks out of annoying other users, although it's never done in any way that's
even remotely clever. He started the /on MSG incidents.
Dr Doom is very much the same. In fact they could well be the same person.
--
__________________________________________________________________________
tr...@plod.cbme.unsw.oz.au
Troy Rollo
cvl> The dirty thing about it is: It generates a message for each person online,
cvl> that is about 220 messages. That is a huge netload which can cause network
cvl> links to crash! It is extremely asocial of these guys using these methods.
cvl> I have been /killing people doing this, but some other ops started flaming at
cvl> me, saying normal users finally found their way to /wall, and that's nice.
Debateable. Most of the walles that come from Ops are annoyances in themselves.
BUt I'd say that there's not really anything uncertain in the case of the
mass - MSGers - kill them. THe load's way over the top.
cvl> I also started talking to the guy, since he had an auto-unkill facility
cvl> and just kept sending laughs at me! He asked what other use there could
cvl> be for /on who? Was it meant to do anything else than that?
cvl> Yes it was. It was only meant to change the looks of the /who-output.
cvl> That's the way it is used in the scripts, part of the ircII distribution.
The same thing applies to people using /on ENTER and /on LEAVE for autogreets,
although it's not such a large problem in terms of efficiency, more an
annoyance.
All the features of IRC are there so you can customise your own
environment, not modify somebody else's (as autogreets effectively
do - after all, what more do I get out of receiving somebody else's
autogreet htan setting up one of me own to autogreet myself on
joining a channel?)
Now, in some cases a /MSG from /ON ENTER, LEAVE or even WHO, may
be a reasonable thing to do (/ON ENTER by a channel op who wishes
to explain lengthy rules for a channel which has a specific purpose,
for example. /on WHO if you have some kind of verification system
for people which uses output of /WHO), but in the vast majority
of cases, any kind of /MSG from a /ON is going to be an annoyance
to its recipient.
BTW, while I'm here, may I suggest an alternative to those of you who
are using /WAIT when performing special once-only processing on the
return value of a command. THere is a better way:
/on CTCP "* * ARGS *" /setargs $*
/alias setargs /if "$N=$0" "/assign ARGS $3-"
/alias WI /ctcp $N ARGS -WI $0\\n/WHO $0/ctcp $N ARGS
/on WHO * /dowho $*
/alias dowho /gowho$ARGS $*
/alias gowho /echo $[10]0 $[3]1 $2@$3 ($4-)
/alias gowho-WI /echo $0's real name is $5-
This hasn't been checked, but I think it demonstrates the point. /WI
translates a nick to a name.
It also gives an example to dweebs like Dr Doom and GFM of what the
/ON functions in IRCII were really meant for.
--
__________________________________________________________________________
tr...@plod.cbme.unsw.oz.au
Greg Lindahl
>It's good that you realize that, so that you'll begin to see that this isn't
>a problem that IRC has any more chance of changing in any sort of technical
>way than any of the other mediums do, and that it just comes with the
>territory.
It's true. However, you'll find that if a user starts posting
thousands of Usenet messages or sending thousands of unwanted mail
messages, their admin will do something about it. Equivalent abuses in
IRC are sometimes ignored. Most of the time the tools IRC has are
capable of dealing with the problem: /ignore, /kick, and channel
modes. But occasionally they aren't -- just like other mediums.
> Most humans that know how to use a computer will listen to the voice
>of reason.
Most will. Some won't. This is the second time that we've had people
around irc who were mindlessly distructive. Both times, it was timely
admin intervention which helped out. Thanks, Noah.
Mike Pelletier
>In article <THB-=+K...@engin.umich.edu>, ste...@engin.umich.edu
(Mike Pelletier) writes...
>>Perhaps the first step would be for people to start a dialogue with
>>the abuser, instead of just assuming that he is an evil jerk and
>>/killing him. Most humans that know how to use a computer will
>>listen to the voice of reason.
>
>Well, I already brought this point up in my earlier post, but in case
>you missed it, I'll repeat. At least in the case of George McBay, I did
>/msg him after his first /msg to me, but he ignored me and continued to
>send me more of these type of /msg's. Obviously he reads this group, as
>I have seen a post from him recently. But he has chosen to ignore this
>whole issue and the fact that numerous individuals have implicated him
>in this matter. Now, he's not the only one who has done this, but as
>far as I've seen he's the only one who did not quit after doing it one
>time. And that's certainly a problem as far as I'm concerned.
I guess he's not "most" people. Either that or he was in no position
to read and respond to your message due to being in the middle of sending
a message to everyone on IRC, scrolling incoming text off his screen at
a rapid rate. Has anyone sent him mail that contains something other than
personal attacks?
Helen Trillian Rose
Troy> == Troy Rollo <tr...@plod.cbme.unsw.oz.au>
Troy> You obviously don't know gfm and Dr Doom very well. They are evil
Troy> jerks. Or perhaps immature jerks would be a better way to phrase
Troy> it. GFM gets his kicks out of annoying other users, although it's
Troy> never done in any way that's even remotely clever. He started the
Troy> /on MSG incidents. Dr Doom is very much the same. In fact they
Troy> could well be the same person.
Could be.. considering they both login from Northeastern from the same
machine at sometimes the same time ... but all of this is trivial
information :-)
--Helen
--
Helen Trillian Rose
<hr...@cs.bu.edu, hr...@cobalt.cco.caltech.edu, hr...@eff.org>
Joshua Geller
frie...@nutrimat.gnu.ai.mit.edu (Noah Friedman) writes:
|> Gfm is the least of anyone's problems on the net right now. He may be
|>irritating, but everything he's done has been mostly harmless. He is not
|>responsible for convincing clueless users to run the command "/on ^msg
|>drdoom $1-" in their IRCII clients, for example.
yeah, george isn't actually so bad. It is amazing how encountering a true
asshole gives one perspective on merely annoying people.
|> I don't personally approve of his behavior. He's been warned about it.
|>But I think it's more important to handle other problems, like the recent
|>increase in cracking.
He is even sorta humorous if you look at it just right.
josh
Joshua Geller
|>You obviously don't know gfm and Dr Doom very well. They are evil jerks.
|>Or perhaps immature jerks would be a better way to phrase it. GFM gets his
|>kicks out of annoying other users, although it's never done in any way that's
|>even remotely clever. He started the /on MSG incidents.
classifying gfm with drdoom is (imho) a mistake. gfm annoys people on IRC;
doom cracks systems with the intent of annoying people on IRC. big difference.
|>Dr Doom is very much the same. In fact they could well be the same person.
could be, I suppose, but I doubt it.
josh
Troy Rollo
joshua> classifying gfm with drdoom is (imho) a mistake. gfm annoys people on IRC;
joshua> doom cracks systems with the intent of annoying people on IRC. big difference.
And you have proven that he doesn't? IT could just mean he doesn't get caught.
One of my first encounters with GFM involved security breaches where the
ultimate aim was to annoy people on IRC. AS I said, he started the /on
incidents, before dr_doom even started to become a problem on IRC (and,
incidentally, before anybody else was abusing ON on anything like a
regular basis).
GFM also started the /ON WHO incidents of late, I believe. These are not
merely annoying, but are expensive in terms of bandwidth. Read "damagingly
antisocial."
He also has the attitude that he can do whatever he pleases and not expect
it to bother anybody. He also claims to be untouchable. In my first
encounter with GFM (long before he became a more general problem) he
claimed that the FSF had given him license to do whatever he pleased.
I am beginning to think this is true.
As for it being funny, well I don't consider preying on new and inexperienced
users in any way funny. To me, that sort of attitude is elitist.
--
__________________________________________________________________________
tr...@plod.cbme.unsw.oz.au
Greg Lindahl
> In my first
>encounter with GFM (long before he became a more general problem) he
>claimed that the FSF had given him license to do whatever he pleased.
>I am beginning to think this is true.
I don't think it is. GFM has received a "final" warning from FSF about
sending messages to everyone on irc. After the last warning he went a
week being nice; if he starts sending messages to everyone please drop
me mail with details.
ystein Tvedten
I agree with Joshua.. GFM is a nuisance, nothing else (at least not that I know
of.) He started first to bother me 1/2 year ago when he used some method of
detecting which channels someone went into (secret ones) and then jumped into
the channel after them. He then disappeared for a while (with minor apperances :)
and have lately gone over to /msgsing\/inviting everybody into a channel.
I *think* (IMHO) that he does it because it drives the OPS mad and he knows
that he irritates them with it.
BTW - what is the use of /killing someone when he/she has an auto-reconnecting
client ? If /kill is gonna work like it's supposed to do (at least the way I see it), it has no use nowadays...
Just my $0.02
Oystein Tvedten
Lancelot on IRC
Ex-IRC admin at IFI.UIO.NO
----
Oystein Tvedten @"I know not with what weapons World War III will
University of Oslo @ be fought, but World War IV will be fought
Norway @ with sticks and stones."
email <oyst...@ifi.uio.no> @ -----Albert Einstein
Charles Hannum
GFM also started the /ON WHO incidents of late, I believe. These are not
merely annoying, but are expensive in terms of bandwidth. Read "damagingly
antisocial."
This is one of the least annoying things I've seen on IRC, actually.
Let me list some others:
1) /ON JOIN messages (a.k.a. autogreets).
2) /WALL messages like "#foo# Just though I'd say hi."
3) Ops masturbating with /SQUIT and /CONNECT. (Pick good routes to
start with and LEAVE THEM ALONE for christ's sake!)
And as far as bandwidth goes, 47 people in your /NOTIFY list is almost
as bad.
Charles Hannum
Charles> 3) Ops masturbating with /SQUIT and /CONNECT. (Pick good routes to
Charles> start with and LEAVE THEM ALONE for christ's sake!)
I think you're a little ... um... uneducated about the linking.
Heh. I hope that was a joke.
In the
middle of the night, US time, when not many operators are on, if there
is a link split due to network death, there is noone around to reconnect
it. This means the server does it on its own.
Exactly. And the fucking server *should* do it automatically, whether
there's someone logged on or not. As it is now, IRC's linking is
braindead and not as flexible as it needs to be. More details in
another message...
But let's face it, if from MIT-> BU is going thru Finland, I'm going to
fix it whether you like it or not.
If the routes were set correctly to begin with, this would never
happen. This is stupidity on the part of the admins at both sites.
Charles> And as far as bandwidth goes, 47 people in your /NOTIFY list
Charles> is almost as bad.
yeah. about time to make this a server function.
Or shorten that /NOTIFY list...
(and yes, I *still* think AIX bites the big one :-)
When did I disagree? B-)
Troy Rollo
mycroft> And as far as bandwidth goes, 47 people in your /NOTIFY list is almost
mycroft> as bad.
Ah.. no it isn't. The /ON WHO /MSG generates one message for each user, then
transmits it across the net to each of them. Typically, this is 400 MSGs.
The load between the client and the server would be similar to the load
of 47 people in a /NOTIFY list per minute, but that's not the whole story,
being that those messages are then forwarded to other servers around the world.
--
__________________________________________________________________________
tr...@plod.cbme.unsw.oz.au
Troy Rollo
troy> transmits it across the net to each of them. Typically, this is 400 MSGs.
Oops! 200 :)
--
__________________________________________________________________________
tr...@plod.cbme.unsw.oz.au
Helen Trillian Rose
Charles> == Charles Hannum <myc...@kropotki.gnu.ai.mit.edu>
Charles> In article <17...@usage.csd.unsw.oz.au> tr...@plod.cbme.unsw.oz.au (Troy Rollo) writes:
Charles> GFM also started the /ON WHO incidents of late, I believe.
Charles> These are not merely annoying, but are expensive in terms of
Charles> bandwidth. Read "damagingly antisocial."
In the past three months, IRC traffic has pretty much stayed the same.
it has, however, gone down in ranking compared to everything else.
(February, I believe, we ranked third. May, we ranked tenth... so
comperably, we're eating less.)
(stats from nis.nsf.net in the "stats" directory)
Jan irc 6667 50,541,247 1.316 3,942,709,477 0.640
Feb irc 6667 107,960,902 2.776 8,849,122,531 1.397
Mar irc 6667 105,673,019 2.595 8,667,528,352 1.263
Apr irc 6667 114,403,671 2.653 9,518,768,576 1.368
May irc 6667 121,510,249 2.579 9,973,686,586 1.247
Charles> This is one of the least annoying things I've seen on IRC, actually.
Charles> Let me list some others:
Charles> 1) /ON JOIN messages (a.k.a. autogreets).
depends. /on join whois isn't a bandwidth chomper.
Autogreeters can be .. depends on the channel.
Charles> 2) /WALL messages like "#foo# Just though I'd say hi."
there aren't that many of these. some of them like "routing" *needs* to
be walled.
Charles> 3) Ops masturbating with /SQUIT and /CONNECT. (Pick good routes to
Charles> start with and LEAVE THEM ALONE for christ's sake!)
I think you're a little ... um... uneducated about the linking. In the
middle of the night, US time, when not many operators are on, if there
is a link split due to network death, there is noone around to reconnect
it. This means the server does it on its own. It *can* deal. Some
servers are setup with a fast reconnect time, so that if it misses a
link, within 5 minutes it will try to reconnect it back.
Some servers do do this every twenty minutes. It's usually the "twenty
minute" servers that lose badly.
But let's face it, if from MIT-> BU is going thru Finland, I'm going to
fix it whether you like it or not. (unless it's a problem inherent in
the network, in which case, I'm subscribed to all the appropriate
mailing lists to keep up to date on *all* problems that might occur,
both on NEARnet, and on the NSFnet and ALTERnet.) Usually the least
painful way is to send a WALL or WALLOPS with "routing .. MIT-BU via
Finland.. fixing" and then to do it *immediately*. No, the best way is
*not* to tell the users "I'm going to fix it in five minutes" because in
that five minutes we waste *more* bandwidth than one WALL.
Charles> And as far as bandwidth goes, 47 people in your /NOTIFY list
Charles> is almost as bad.
yeah. about time to make this a server function.
--Helen
(and yes, I *still* think AIX bites the big one :-)
--
Helen Trillian Rose <hr...@eff.org, hr...@cs.bu.edu>
Electronic Frontier Foundation NeXT Mail-> <hr...@black-cube.eff.org>
"This machine was behaving badly so I had to shoot it. MK. I messed up."
(Mitch Kapor <mka...@eff.org> on the misbehaviour of the NeXT at EFF :-)
Eric Olson
>In article <17...@usage.csd.unsw.oz.au> tr...@plod.cbme.unsw.oz.au (Troy Rollo) writes:
> GFM also started the /ON WHO incidents of late, I believe. These are not
> merely annoying, but are expensive in terms of bandwidth. Read "damagingly
> antisocial."
>This is one of the least annoying things I've seen on IRC, actually.
>Let me list some others:
> 1) /ON JOIN messages (a.k.a. autogreets).
Hear, hear. I generally leave a channel immediately. The only thing
that can drive me out quicker are the huge ascii graphics which people
seem compelled to send all the time.
> 2) /WALL messages like "#foo# Just though I'd say hi."
> 3) Ops masturbating with /SQUIT and /CONNECT. (Pick good routes to
>start with and LEAVE THEM ALONE for christ's sake!)
That's the point. We pick good routes in the configuration files, but
SOMETIMES LINKS ARE DOWN! Capiche? And IRC---now stay with me on
this one---tries to reconnect automatically any way it can! So---let
me know if I go too fast---left alone in the REAL WORLD long enough,
the routes will deteriorate and become more and more random. Unless
you want to take all of IRC down once a week so it can restart using
its "optimal" routes, you'll have to settle for operators patching it
by hand.
>And as far as bandwidth goes, 47 people in your /NOTIFY list is almost
>as bad.
Ayup.
Eric Olson <e...@kaja.gi.alaska.edu>
Darren Reed
[...]
>But let's face it, if from MIT-> BU is going thru Finland, I'm going to
>fix it whether you like it or not. (unless it's a problem inherent in
>the network, in which case, I'm subscribed to all the appropriate
>mailing lists to keep up to date on *all* problems that might occur,
>both on NEARnet, and on the NSFnet and ALTERnet.) Usually the least
>painful way is to send a WALL or WALLOPS with "routing .. MIT-BU via
>Finland.. fixing" and then to do it *immediately*. No, the best way is
>*not* to tell the users "I'm going to fix it in five minutes" because in
>that five minutes we waste *more* bandwidth than one WALL.
hmmm...well if you use *.fi/*.edu on the finland links then loops cant
form. Its a very nice way of preventing loops using those wildcard
links (whatever the drawbacks are that is a useful feature).
-avalon
Darren Reed
>From article <MYCROFT.91...@kropotki.gnu.ai.mit.edu>, by myc...@kropotki.gnu.ai.mit.edu (Charles Hannum):
>mycroft> And as far as bandwidth goes, 47 people in your /NOTIFY list is almost
>mycroft> as bad.
>Ah.. no it isn't. The /ON WHO /MSG generates one message for each user, then
>transmits it across the net to each of them. Typically, this is 400 MSGs.
>The load between the client and the server would be similar to the load
>of 47 people in a /NOTIFY list per minute, but that's not the whole story,
>being that those messages are then forwarded to other servers around the world.
I wouldnt like to do that on pre19+...it has inbuilt delays which are supposed
to stop that sort of flooding. at 5 msgs/10 seconds 400 messages take a
long while to get through (providing the delay works).
-avalon
Greg Lindahl
>This is one of the least annoying things I've seen on IRC, actually.
>Let me list some others:
>
> 1) /ON JOIN messages (a.k.a. autogreets).
> 2) /WALL messages like "#foo# Just though I'd say hi."
> 3) Ops masturbating with /SQUIT and /CONNECT. (Pick good routes to
>start with and LEAVE THEM ALONE for christ's sake!)
The first 2 are annoying but waste a small amount of bandwidth
compared to a single gfm-style global message. This is because channel
messages are efficiently transmitted, whereas multiple different
messages have to go out multiple times.
I'm not sure what you mean with #3. We do pick good routes, but the
Internet is subject to random failures. We've been attempting to keep
the number of squit/connects to a minimum; if you'd like to help then
become and operator and whine at DCLXVI every time he decides to "fix"
the working net.
4 wrongs don't make a right.
Helen Trillian Rose
>But let's face it, if from MIT-> BU is going thru Finland, I'm going to
>fix it whether you like it or not. (unless it's a problem inherent in
>the network, in which case, I'm subscribed to all the appropriate
>mailing lists to keep up to date on *all* problems that might occur,
>both on NEARnet, and on the NSFnet and ALTERnet.) Usually the least
>painful way is to send a WALL or WALLOPS with "routing .. MIT-BU via
>Finland.. fixing" and then to do it *immediately*. No, the best way is
>*not* to tell the users "I'm going to fix it in five minutes" because in
>that five minutes we waste *more* bandwidth than one WALL.
Avalon> hmmm...well if you use *.fi/*.edu on the finland links then
Avalon> loops cant form. Its a very nice way of preventing loops using
Avalon> those wildcard links (whatever the drawbacks are that is a
Avalon> useful feature).
yeah, but if I lose Australia off of near.ugcs.caltech.edu then I
*can't* /conn coombs.* because there isn't a server *called* Coombs in
my conf file .. since coombs and three other servers resolve to *.au
Same if I did that to Finland, which is precisely why I haven't
implemented it on bucsd yet. Or do you have a solution?
--Helen
Darren Reed
> Avalon> hmmm...well if you use *.fi/*.edu on the finland links then
> Avalon> loops cant form. Its a very nice way of preventing loops using
> Avalon> those wildcard links (whatever the drawbacks are that is a
> Avalon> useful feature).
>yeah, but if I lose Australia off of near.ugcs.caltech.edu then I
>*can't* /conn coombs.* because there isn't a server *called* Coombs in
>my conf file .. since coombs and three other servers resolve to *.au
>Same if I did that to Finland, which is precisely why I haven't
>implemented it on bucsd yet. Or do you have a solution?
2.6pre19 should check the server name for connect's from both parts
of a C: line. ie from near's conf line, you would have to say
"/connect 130.56.96.2" since you've got the IP addr for coombs.anu.edu.au
there and not the domainname :-)) That should be fixed in pre19
(its sooo long since pre19 was released i forget all the bugs that
were fixed from pre18 - we need another server version release :-))
-avalon
Joshua Geller
(Eric Olson) writes:
|>That's the point. We pick good routes in the configuration files, but
|>SOMETIMES LINKS ARE DOWN! Capiche? And IRC---now stay with me on
|>this one---tries to reconnect automatically any way it can! So---let
|>me know if I go too fast---left alone in the REAL WORLD long enough,
|>the routes will deteriorate and become more and more random. Unless
|>you want to take all of IRC down once a week so it can restart using
|>its "optimal" routes, you'll have to settle for operators patching it
|>by hand.
Close. Not everyone has good routes in their configuration files. It only
takes a few servers with bad routes in their configuration files and a good
net split to totally hose routing. Totally hosed in this context implies long
inefficient routes (like, say, hardy to henson (80 miles apart) through japan
and the united states). The situation is better than it was, but by no means
perfect.
josh
Joshua Geller
/connect [*.au *.jp *.kr] works just fine.....
at least from hardy....
josh