I'm looking for references to chaosnet.
The MIT AI Memo No. 628 at
ftp://publications.ai.mit.edu/ai-publications/500-999/AIM-628.ps looks
like a scan from a tenth generation xerox copy and is unreadable. Does
anyone have a plain text copy of this document, or at least a better
scan? Or does anyone have a hardcopy version?
Also, what is the status of the Linux port of chaosnet? The site
ftp://ftp.heeltoe.com/pub/chaos does not appear to have anything there.
RCS will be holding a Symbolics open house on Oct. 19th and we would
like to get some info before then. We'll have two or three 3600 series
machines networked with chaosnet running on that day. It would be
interesting to try out the Linux software on our router.
RCS has started to collect some notes on the restoration of our
machines. They can be found at our "works in progress" website at
http://starfish.osfn.org/rcs/Symbolics/ So far it just contains board
lists and part numbers, but later we will post some images of the
machines and other info.
-mikeu
Michael Umbricht
--
The Retro-Computing Society of RI, Inc. (401) 861-1977
25 Eagle St Bldg 5 Ste 206 http://www.osfn.org/rcs/
Providence RI 02908 telnet://kronos.egr-ri.ids.net
In article <·················@osfn.org>, RCS / RI <···@osfn.org> wrote:
> I'm looking for references to chaosnet.
>
try www.spies.com/aek/pdf/mit/AIM-628_chaosnet.pdf
I also have some schematics for the unibus chaosnet interface
used in the CADR. It would be nice to find the schematics for
MIT's cable interface, since I hope to rivive the CADR that
I brought out to the west coast this past spring (If I can ever
find the software, and I can get the CADR CPU schematics)
In article <····················@il0502a-dhcp245.apple.com>,
Al Kossow <···@spies.com> wrote:
>In article <·················@osfn.org>, RCS / RI <···@osfn.org> wrote:
>
>> I'm looking for references to chaosnet.
>>
>
>try www.spies.com/aek/pdf/mit/AIM-628_chaosnet.pdf
It would sure be nice if you would donate that to MIT, to replace the
crappy one they've got. Theirs is close to unreadable.
--
Barry Margolin, ······@genuity.net
Genuity, Woburn, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
Hello,
In article <·················@osfn.org>,
RCS / RI <···@osfn.org> writes:
>
> I'm looking for references to chaosnet.
>
> The MIT AI Memo No. 628 at
> ftp://publications.ai.mit.edu/ai-publications/500-999/AIM-628.ps looks
> like a scan from a tenth generation xerox copy and is unreadable. Does
> anyone have a plain text copy of this document, or at least a better
> scan? Or does anyone have a hardcopy version?
J. Noel Chiappa told me early this year:
> The only thing I can quickly locate online is:
> http://its.svensson.org/sysdoc;chaord
It seems to not be online at the moment, and I don't have a copy here (maybe
at home). Try yourself.
Regards,
-is
Hi,
I wrote:
>> The only thing I can quickly locate online is:
>> http://its.svensson.org/sysdoc;chaord
>
> It seems to not be online at the moment, and I don't have a copy here (maybe
> at home). Try yourself.
of course, you can find it in the google archive, at least for a while:
http://www.google.de/search?q=cache:R_yy5MJQ7BUC:its.svensson.org/SYSDOC%253BCHAORD+sysdoc%5C%3Bchaord
-is
Ignatios Souvatzis wrote:
> Hi,
>
> I wrote:
>
>
>>>The only thing I can quickly locate online is:
>>> http://its.svensson.org/sysdoc;chaord
>>
>>It seems to not be online at the moment, and I don't have a copy here (maybe
>>at home). Try yourself.
>
>
> of course, you can find it in the google archive, at least for a while:
>
> http://www.google.de/search?q=cache:R_yy5MJQ7BUC:its.svensson.org/SYSDOC%253BCHAORD+sysdoc%5C%3Bchaord
>
> -is
link does not work
Hi,
In article <············@usenet.otenet.gr>,
ilias <·······@pontos.net> writes:
>> http://www.google.de/search?q=cache:R_yy5MJQ7BUC:its.svensson.org/SYSDOC%253BCHAORD+sysdoc%5C%3Bchaord
>>
>
> link does not work
Strange - it works for me. Anyway, but the original is online again.
Regards,
-is
In article <·············@dtpq.com>,
Christopher C. Stacy <······@dtpq.com> wrote:
>I think old computer stuff is fun, but I am still wondering:
>Why does anyone care about CHAOSNET at all?
Getting it right.
Followups reduced.
>>>>> On 30 Sep 2002 13:23:11 -0800, Eugene Miya ("Eugene") writes:
Eugene> In article <·············@dtpq.com>,
Eugene> Christopher C. Stacy <······@dtpq.com> wrote:
>> I think old computer stuff is fun, but I am still wondering:
>> Why does anyone care about CHAOSNET at all?
Eugene> Getting it right.
I can't quite decipher your response because it lacks any context.
My question was: in this day and age, for what purpose are people
so interested in CHAOSNET, such that they keep asking about it?
It's historically interesting, but the question comes up regularly
on comp.lang.lisp and other newsgroups which are not primarily about
footnotes in the history of computing.
By "Getting it right", maybe you are suggesting that CHAOSNET is
superior to what people run on their LAN's today (TCP/IP), or for
inventing some replacement for TCP/IP. I am not sure there are
many lessons to be learned that would be applicable to designing
today's kind of local area networks. CHAOSNET's main virtue was
that it was very simple.
CHAOSNET was the local area network protocol invented at the MIT
AI Lab for the Lisp Machines, back before Ethernet was available.
It also ran on ITS, of course, and was later ported to Unix, VMS,
and TOPS-20. It was made pretty much obsolete when Ethernet became
commercially available and TCP/IP was implemented on the Lisp Machine.
Here's CHAOSNET in a nutshell.
It was a 75 ohm 4 Mbit CSMA cable network.
Packets begin with 48 bits of hardware leader, consisting of the source
and destination address (what today we would call the MAC address),
and a checksum. There was a simple bridging (routing) protocol.
Packets can be uncontrolled (like UDP) or controlled (like TCP,
delivered in sequence and flow controlled)
The data portion of the packet is up to 252 16-bit words.
The first 128 bits of the data are the software header.
CHAOSNET had a flat 16-bit address space (8 bit subnet and 8 bit host).
It had 16-bit "indexes" which specified the unique port numbers on
each end (the address+index is globally unique on the host).
Request for connections to services are not by well-known port number,
but by well-known contact name strings (eg. "MAIL").
A protocol could also receive some initial data in the RFC packet.
Also, instead of opening a connection, lightweight protocols could
be defined to send back a single-packet response (ANS packet).
An RFC could also cause the server to send a redirection packet (FWD).
There were a few other simple packets similar to what ICMP does.
That's about it! If you get a copy of the AI Lab Memo #628,
you can read about the API and other details.
My personal theory is that people who hear the word "CHAOSNET" have
little idea what it was, but they just think the name sounds cool.
It's called that because of the contrast between a wire network with
packet collisions, versus ARPANET NCP or the token-ring experiments
going on downstairs at the Laboratory for Computer Science, and also
because one of the inventors (or his girlfriend, I forget exactly)
was reading the Roger Zelazny fantasy novels at the time.
"Christopher C. Stacy" <······@dtpq.com> wrote in message ··················@dtpq.com...
>
> It's called that because of the contrast between a wire network with
> packet collisions, versus ARPANET NCP or the token-ring experiments
> going on downstairs at the Laboratory for Computer Science, and also
> because one of the inventors (or his girlfriend, I forget exactly)
> was reading the Roger Zelazny fantasy novels at the time.
Interesting...
I rather dislike Ethernet for its collision-prone design, too. At home,
I used to run ARCNET rather than Ethernet, but it got to where the ODI
stack I had to use in DOS under Windows 3.1 just pushed me right over
the memory limit.
Also, I had to use a Linux box to extend its reach, since I couldn't
find any ARCNET active hubs I could afford, and passive hubs allow
only four ports.
-dq
······@dtpq.com (Christopher C. Stacy) wrote in message news:<·············@dtpq.com>...
> >>>>> On 30 Sep 2002 13:23:11 -0800, Eugene Miya ("Eugene") writes:
> Eugene> In article <·············@dtpq.com>,
> Eugene> Christopher C. Stacy <······@dtpq.com> wrote:
> >> I think old computer stuff is fun, but I am still wondering:
> >> Why does anyone care about CHAOSNET at all?
>
> Eugene> Getting it right.
>
> I can't quite decipher your response because it lacks any context.
As opposed to Eugene's past posts where he invokes legends by
first name, mentions secure areas at a couple national labs, and then hints at
something that only a half-dozen people in the world comprehend?
Nothing against your posting style, Eugene, I just want to let Mr. Stacy
know that he isn't the only one who cannot always decipher your posts :-)
(And I've worked at those national labs...)
Tim.
In article <····························@posting.google.com>,
Tim Shoppa <······@trailing-edge.com> wrote:
>······@dtpq.com (Christopher C. Stacy) wrote in message news:<·············@dtpq.com>...
>> >>>>> On 30 Sep 2002 13:23:11 -0800, Eugene Miya ("Eugene") writes:
>> Eugene> In article <·············@dtpq.com>,
>> Eugene> Christopher C. Stacy <······@dtpq.com> wrote:
>> >> I think old computer stuff is fun, but I am still wondering:
>> >> Why does anyone care about CHAOSNET at all?
>>
>> Eugene> Getting it right.
>>
>> I can't quite decipher your response because it lacks any context.
>
>As opposed to Eugene's past posts where he invokes legends by
>first name, mentions secure areas at a couple national labs, and then hints at
>something that only a half-dozen people in the world comprehend?
It is generally not the practice of fishermen to steal or comment upon
one anothers bait.
Legends? They ain't dead yet.
>Nothing against your posting style, Eugene, I just want to let Mr. Stacy
>know that he isn't the only one who cannot always decipher your posts :-)
;^)
>(And I've worked at those national labs...)
I know.
The competition.
······@cse.ucsc.edu (Eugene Miya) writes:
>In article <····························@posting.google.com>,
>Tim Shoppa <······@trailing-edge.com> wrote:
>>······@dtpq.com (Christopher C. Stacy) wrote in message news:<·············@dtpq.com>...
>>> >>>>> On 30 Sep 2002 13:23:11 -0800, Eugene Miya ("Eugene") writes:
>>> Eugene> In article <·············@dtpq.com>,
>>> Eugene> Christopher C. Stacy <······@dtpq.com> wrote:
>>> >> I think old computer stuff is fun, but I am still wondering:
>>> >> Why does anyone care about CHAOSNET at all?
>>>
>>> Eugene> Getting it right.
>>>
>>> I can't quite decipher your response because it lacks any context.
>>
>>As opposed to Eugene's past posts where he invokes legends by
>>first name, mentions secure areas at a couple national labs, and then hints at
>>something that only a half-dozen people in the world comprehend?
>It is generally not the practice of fishermen to steal or comment upon
>one anothers bait.
Translation(?): Those who do not learn the lessons of history are doomed to
repeat the same mistakes.
Hey, what do you expect from someone fond of machines like the Enigma? :-)
-Ric
--
There are 10 kinds of people in the world:
Those who understand binary, and those who don't.
Ric Werme http://ewerme.home.attbi.com/ | ······@xxxxx.com
see also http://dcyf.home.attbi.com/ | Change xxxxx to attbi
"Christopher C. Stacy" wrote:
>
> >>>>> On 30 Sep 2002 13:23:11 -0800, Eugene Miya ("Eugene") writes:
>
> [...] It's called [CHAOSNET] because of the contrast between a wire
> network with packet collisions, versus ARPANET NCP or the token-ring
> experiments going on downstairs at the Laboratory for Computer
> Science.
Am I confusing CHAOSNET with something else? I thought it had a token-
like collision avoidance mechanism.
I am thinking of a network that, like CHAOSNET, had an 8-bit host
number. Each NIC in the network I am thinking of had an 8-bit counter
running freely at some specified frequency. A NIC would only transmit
when the value of its counter equalled its own host number.
The hardware was equipped to detect collisions, and collisions could
occur, especially if the counters in the various NICs were not
synchronized; but that would be seldom because every time a packet
was sent, all of the other NICs would set their own counters equal to
the host number of the transmitter. (or maybe it was, host number+1)
I KNOW that the network I am thinking of came out of MIT. Is it
CHAOSNET, or is it something else?
-- Foo?
>>>>> On Wed, 02 Oct 2002 00:33:51 GMT, Foobar T Clown ("Foobar") writes:
Foobar> "Christopher C. Stacy" wrote:
>>
>> >>>>> On 30 Sep 2002 13:23:11 -0800, Eugene Miya ("Eugene") writes:
>>
>> [...] It's called [CHAOSNET] because of the contrast between a wire
>> network with packet collisions, versus ARPANET NCP or the token-ring
>> experiments going on downstairs at the Laboratory for Computer
>> Science.
Foobar> Am I confusing CHAOSNET with something else? I thought it had a token-
Foobar> like collision avoidance mechanism.
Foobar> I am thinking of a network that, like CHAOSNET, had an 8-bit host
Foobar> number. Each NIC in the network I am thinking of had an 8-bit counter
Foobar> running freely at some specified frequency. A NIC would only transmit
Foobar> when the value of its counter equalled its own host number.
Foobar> The hardware was equipped to detect collisions, and collisions could
Foobar> occur, especially if the counters in the various NICs were not
Foobar> synchronized; but that would be seldom because every time a packet
Foobar> was sent, all of the other NICs would set their own counters equal to
Foobar> the host number of the transmitter. (or maybe it was, host number+1)
Foobar> I KNOW that the network I am thinking of came out of MIT.
Foobar> Is it CHAOSNET, or is it something else?
Yes! CHAOSNET uses the TDM scheme you are thinking of (somewhat like a
token ring) in order to minimize collisions, and also uses carrier-sense.
Each interface has an time-slot ("turn") counter that runs when the
ether is not busy. Each interface only attempts to seize the wire
when its turn comes up: when the its counter equals its host address.
The turn counters are re-synchronized whenever a packet goes by.
The length of a turn must allow for a packet to be transmitted
and consider the propogation delay to the next interface's turn.
Host addresses (turns) are assigned linearly along the wire
according to a formula of those terms.
When collisions do happen, the driver software tries to retransmit
a few times, and may then give up (possibly losing a packet).
Hi,
······@dtpq.com (Christopher C. Stacy) writes:
> My question was: in this day and age, for what purpose are people
> so interested in CHAOSNET, such that they keep asking about it?
>
> It's historically interesting, but the question comes up regularly
> on comp.lang.lisp and other newsgroups which are not primarily about
> footnotes in the history of computing.
comp.lang.lisp: because of lisp machines.
But I'm writing for a different reason.
http://its.svensson.org/sysdoc;chaord has a paragraph "comparison with
LCSnet". It starts with "The LCSnet proposed protocol is called DSP."
Is there any online document about that?
[Followup-To set to afc only.]
Regards,
-is
Hi,
······@dtpq.com (Christopher C. Stacy) writes:
> My question was: in this day and age, for what purpose are people
> so interested in CHAOSNET, such that they keep asking about it?
>
> It's historically interesting, but the question comes up regularly
> on comp.lang.lisp and other newsgroups which are not primarily about
> footnotes in the history of computing.
comp.lang.lisp: because of lisp machines.
But I'm writing for a different reason.
http://its.svensson.org/sysdoc;chaord has a paragraph "comparison with
LCSnet". It starts with "The LCSnet proposed protocol is called DSP."
Is there any online document about that? (e.g.,
Clark, D., "Revision of DSP Specification", Local Network Note 9,
Laboratory for Computer Science, MIT, 17 June 1977., in case that applies).
[Followup-To set to afc only.]
Regards,
-is
"Christopher C. Stacy" wrote:
>
> I think old computer stuff is fun, but I am still wondering:
> Why does anyone care about CHAOSNET at all?
The Retro-Computing Society is a computer museum that restores
historically significant machines to working order and then displays
them for educational purposes. We runs demos on the machines for our
vistors to learn from.
The Symbolics 3600 systems in our collection originally (when they were
in production use) ran both TCP/IP and Chaosnet over Etherenet. We are
interested in seeing how Chaosnet worked and what features it supported.
Also, it seem like there is very little documentation about Chaosnet on
the web. We would like to document the protocol so that others can
learn about the history of networking.
-mikeu
Michael Umbricht
--
The Retro-Computing Society of RI, Inc. (401) 861-1977
25 Eagle St Bldg 5 Ste 206 http://www.osfn.org/rcs/
Providence RI 02908 telnet://kronos.egr-ri.ids.net
······@dtpq.com (Christopher C. Stacy) wrote:
>I think old computer stuff is fun, but I am still wondering:
>Why does anyone care about CHAOSNET at all?
Slightly off topic ... I vaguely remember that MIT's AI lab had a
TOPS-20 machine called MIT-OZ that wasn't able to support both ARPAnet
and Chaosnet without crashing. However, I don't remember many of the
details.
--gregbo
gds at best dot com
>>>>> On Fri, 04 Oct 2002 20:02:30 GMT, Greg Skinner ("Greg") writes:
Greg> ······@dtpq.com (Christopher C. Stacy) wrote:
>> I think old computer stuff is fun, but I am still wondering:
>> Why does anyone care about CHAOSNET at all?
Greg> Slightly off topic ... I vaguely remember that MIT's AI lab had a
Greg> TOPS-20 machine called MIT-OZ that wasn't able to support both ARPAnet
Greg> and Chaosnet without crashing. However, I don't remember many of the
Greg> details.
Yes, although I tried to have as little to do with that TOPS-20
machine as I could, so I don't remember what the bug was exactly,
or if anyone ever got it fixed. MIT had three TOPS-20 machines,
and the monitors were hacked up on all of them. The best people
to ask for those details would be Jeff Schiller and Rob Austien.
······@dtpq.com (Christopher C. Stacy) wrote:
>>>>>> On Fri, 04 Oct 2002 20:02:30 GMT, Greg Skinner ("Greg") writes:
> Greg> Slightly off topic ... I vaguely remember that MIT's AI lab had a
> Greg> TOPS-20 machine called MIT-OZ that wasn't able to support both ARPAnet
> Greg> and Chaosnet without crashing. However, I don't remember many of the
> Greg> details.
>Yes, although I tried to have as little to do with that TOPS-20
>machine as I could, so I don't remember what the bug was exactly,
>or if anyone ever got it fixed. MIT had three TOPS-20 machines,
>and the monitors were hacked up on all of them. The best people
>to ask for those details would be Jeff Schiller and Rob Austien.
I think it had something to do with the number of people connecting to
OZ using supdup. There were a lot of connections to OZ from people
using Lisp machines, terminal servers, and various other systems on
MIT's Chaosnet.
--gregbo
gds at best dot com
>>>>> On Mon, 07 Oct 2002 23:41:07 GMT, gds ("gds") writes:
gds> ······@dtpq.com (Christopher C. Stacy) wrote:
>>>>>>> On Fri, 04 Oct 2002 20:02:30 GMT, Greg Skinner ("Greg") writes:
Greg> Slightly off topic ... I vaguely remember that MIT's AI lab had a
Greg> TOPS-20 machine called MIT-OZ that wasn't able to support both ARPAnet
Greg> and Chaosnet without crashing. However, I don't remember many of the
Greg> details.
>> Yes, although I tried to have as little to do with that TOPS-20
>> machine as I could, so I don't remember what the bug was exactly,
>> or if anyone ever got it fixed. MIT had three TOPS-20 machines,
>> and the monitors were hacked up on all of them. The best people
>> to ask for those details would be Jeff Schiller and Rob Austien.
gds> I think it had something to do with the number of people connecting to
gds> OZ using supdup. There were a lot of connections to OZ from people
gds> using Lisp machines, terminal servers, and various other systems on
gds> MIT's Chaosnet.
That rings a bell.
"Christopher C. Stacy" wrote:
>
> >>>>> On Mon, 07 Oct 2002 23:41:07 GMT, gds ("gds") writes:
> gds> I think it had something to do with the number of people
> gds> connecting to OZ using supdup. There were a lot of connections
> gds> to OZ from people using Lisp machines, terminal servers, and
> gds> various other systems on MIT's Chaosnet.
>
> That rings a bell.
^G^G^G%% Decsystem-20 not running.
Does the Symbolics Alpha Emulator do Chaosnet ?
Some kind of Lisp to Unix raw NIC socket interface?
RCS / RI <···@osfn.org> wrote in message ······················@osfn.org...
>
> I'm looking for references to chaosnet.
>
> RCS will be holding a Symbolics open house on Oct. 19th and we would
> like to get some info before then. We'll have two or three 3600 series
> machines networked with chaosnet running on that day. It would be
> interesting to try out the Linux software on our router.
Do LispMs show better VM smoothness, even compared to the latest / last
generation Alphas ? (Would It be useful demonstration to add an Alpha
emulator to the demo net)
Do they do Object (rather than page) swapping / leaf unfolding - a la ST-80
LOOM ?
In article <·····················@news4.srv.hcvlny.cv.net>,
Michael J. Ferrador <···················@orn.com> wrote:
>Do they do Object (rather than page) swapping / leaf unfolding - a la ST-80
>LOOM ?
No. The virtual memory system is pretty traditional page swapping. There
are a few special features in the memory subsystem to support the ephemeral
GC, but that's about it IIRC.
--
Barry Margolin, ······@genuity.net
Genuity, Woburn, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.