[ih] Internet-history Digest, Vol 10, Issue 34
John Shoch
j at shoch.com
Thu Jul 23 09:31:59 PDT 2020
If Dave F. has an expanded bibliography available on DCS, that would be
great! He knows a lot more than I do :).
In the meantime, I do have handy a few pointers for anyone interested in
some of the early work:
--My PhD thesis (in 1979, over 40 years ago!) was mostly about the
Ethernet, but the initial part included a review and taxonomy of almost 100
local computer network designs of the time, derived from hundreds of papers.
The full thesis can be found at:
https://ethernethistory.typepad.com/papers/1979-%20PhD%20Thesis.pdf
The annotated bibliography was later published as a separate tech report:
http://bitsavers.trailing-edge.com/pdf/xerox/parc/techReports/SSL-79-5_An_Annotated_Bibliography_on_Local_Computer_Networks_Oct79.pdf
--From the taxonomy, there were 19 DCS papers from 1970-1978.
Distributed Computing System (UC Irvine):
[Will, 1970] considered both control passing & empty slot
[Farber, 1970] very early proposal, fixed size blocks
[Farber, 1972]
[Farber & Larson, 1972a]
[Farber & Larson, 1972b]
[Farber & Heinrich, 1972]
[Loomis, 1973(7)]
[Farber, et a!., 1973]
[Rowe, et al., 1973]
[Farber & Vittal, 1973]
[Farber, 1974]
[Farber, 1975a]
[Farber, 1975b]
[Rowe, 1975]
[Lyle & Farber, 1976]
[Mokapetris, et a!., 1977]
[Farber, 1977]
[Mokapetris & Farber, 1977]
[Mokapetris, 1978]
--Detailed info on the 19 citations are in the body of the bibliography (in
either link above).
--The thesis included a "compare and contrast" discussion; the section on
circular networks first touched on the Farmer ring at AT&T, and then
includes these comments on DCS -- as I understood it at the time:
"b. Distributed Computing System (DCS) at UC Irvine.
Following the publication of the Farmer and Newhall paper in 1969, there
was further consideration of ring systems. In one early paper, Farber
compared the relative merits of the control passing and empty slot
techniques [Farber, 1970]; he initially endorsed the empty slot technique,
and went on to plan the development of the Distributed Computing System at
Irvine.
For the first several years the design was still based on empty slots
passing by each node [Farber & Larson, 1972a, 1972b; Farber & Heinrich,
1972]; only later did the detailed design return to the control passing
strategy, reminiscent of the Newhall and Farmer work [Loomis, 1973].
Instead of using a bipolar violation to indicate passing of control, the
DCS system eventually used a bit pattern to distinguish a special control
token that could be passed among machines. A station with a packet to send
would wait for the token to arrive, change 1 bit making it a connector
instead of a token, then transmit its packet and regenerate the token. With
one bit of storage in each node, the packet was -- in effect - pulled off
the ring at the source ring interface (RI) as it was being sent.
Instead of depending on a centralized loop supervisor to detect a lost
token, however, DCS proposed a distributed method in which any machine
might time out and try to recreate the token, and send a packet with it
around the net. If two machines simultaneously tried to create a token and
start sending, they would both be taking off each other's packets; further
timeouts might take place, but the system would eventually stabilize.
This application has led to a more formal treatment of the problem, cast in
terms of mutual exclusion and access to a resource: given a set of machines
configured in a circular topology, what form of distributed algorithm can
they use to mutually agree that one and only one machine should be given
access to the resource (i.e., that either the highest or lowest numbered
host should regenerate the token) [Le Lann, 1977, esp. section 4; Chang &
Roberts, 1979].
The DCS design also included at the hardware level the use of 16-bit
process addresses, rather than specific host addresses; the intention was
to allow position independent addressing, and even let processes migrate
from host to host. Addresses were to be recognized by an associative name
table in the ring interface at the destination host.
The reality of this system was a bit less than its image. DCS really only
supported three directly connected hosts, although various papers had
illustrations showing many more [Mockapetris & Farber, 1977]. Furthermore,
no process ever migrated from one machine to another. The 16-bit address
was not treated as a flat process name space, but was hierarchically split
to include a 4-bit host field, combined with a process ID. It was possible
to start two equivalent processes on two machines, give them the same
process number in their respective hosts, and then address a packet to both
of them by using a "wild card" specification in the host portion of the
address. But those processes were never able to move to a different
machine. "
John Shoch
[With some apprehension, re-reading the section on distributed ring control
forces me to comment on the Ethernet-vs-Token Ring battles that emerged
later.
I don't think DCS made this argument, but IBM tried to claim that TR beat
Ethernet because a TR could have a "guaranteed, bounded access time, for
real-time applications." We knew that was BS, and pointed out, "Yes, if a
ring breaks or is just unplugged, you lose control. And to regenerate the
control token, you may have to repeatedly resolve, oh, a collision! Sigh.]
On Thu, Jul 23, 2020 at 6:13 AM <internet-history-request at elists.isoc.org>
wrote:
> Send Internet-history mailing list submissions to
> internet-history at elists.isoc.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
> https://elists.isoc.org/mailman/listinfo/internet-history
> or, via email, send a message with subject or body 'help' to
> internet-history-request at elists.isoc.org
>
> You can reach the person managing the list at
> internet-history-owner at elists.isoc.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Internet-history digest..."
>
>
> Today's Topics:
>
> 1. Dave Farbers DCS (farber)
> 2. Arpanet physical connectors (Stephane Bortzmeyer)
> 3. Re: Dave Farber's DCS (Dan York)
> 4. Re: Arpanet physical connectors (Carsten Bormann)
> 5. Re: Arpanet physical connectors (Andrew G. Malis)
> 6. Re: Arpanet physical connectors (Alex McKenzie)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Thu, 23 Jul 2020 11:01:03 +0900
> From: farber <dave at farber.net>
> To: Karl <karl at cavebear.com>
> Cc: internet-history at elists.isoc.org
> Subject: [ih] Dave Farbers DCS
> Message-ID: <14EB750C-C3CD-47FB-A620-8646359194AF at farber.net>
> Content-Type: text/plain; charset=utf-8
>
> Your note in internet history motivated to look into my files and retrieve
> papers on DCS. It was a seminal research effort which developed many of the
> ideas that evolved into the basis of modern computing. It also created a
> great set of grad students (somewhat hard to handle but that was great). If
> I don?t find good papers I will create a video chat on DCS with some of the
> students that were involved.
>
> Stay tuned
>
> Dave
>
> David J. Farber
> Distinguished Professor and
> Co-Director of the
> Cyber Civilization Research Center
> Keio University Tokyo Japan
> https://www.ccrc.keio.ac.jp
>
>
>
>
>
> ------------------------------
>
> Message: 2
> Date: Thu, 23 Jul 2020 14:02:04 +0200
> From: Stephane Bortzmeyer <bortzmeyer at nic.fr>
> To: internet-history at elists.isoc.org
> Subject: [ih] Arpanet physical connectors
> Message-ID: <20200723120204.GA16851 at nic.fr>
> Content-Type: text/plain; charset=us-ascii
>
> https://twitter.com/nielstenoever/status/1286254151874293760
>
> Dear Internet History nerds, what did the connectors and cables of the
> ARPAnet look like?
>
> Were these serial cables? Seems likely because the RS-232 standard dates
> back to 1960.
>
>
> ------------------------------
>
> Message: 3
> Date: Thu, 23 Jul 2020 12:14:37 +0000
> From: Dan York <york at isoc.org>
> To: Toerless Eckert <tte at cs.fau.de>
> Cc: Karl Auerbach <karl at cavebear.com>,
> "internet-history at elists.isoc.org" <
> internet-history at elists.isoc.org>
> Subject: Re: [ih] Dave Farber's DCS
> Message-ID: <61093DF5-4550-4AA7-95D7-EAC4C4BBECE1 at isoc.org>
> Content-Type: text/plain; charset="utf-8"
>
>
> > On Jul 22, 2020, at 8:48 PM, Toerless Eckert via Internet-history <
> internet-history at elists.isoc.org> wrote:
> >
> > Would also be interested if there was some more info on DCS.
>
>
> To that end, if anyone has any pointers to more information about DCS, I
> would be glad to add some info to relevant pages of Wikipedia so that
> others might find out and learn more about it.[1] There is not currently a
> page about ?Distributed Computing System (DCS)? specifically, and the only
> brief mention I found was in the second paragraph under Career on Dave
> Farber?s page: https://en.wikipedia.org/wiki/David_J._Farber At the very
> least, it might be useful to expand a little bit more on that page.
>
> If anyone has pointers to papers or articles, that would be helpful. For
> any text added to Wikipedia, there should be citations to ?reliable
> sources? - https://en.wikipedia.org/wiki/Wikipedia:Reliable_sources
>
> Regards,
> Dan
>
> [1] During the pandemic, I?ve found myself diving deeper into Wikipedia
> editing as part of my own way of dealing with all that is going on around
> us.
>
> ------------------------------
>
> Message: 4
> Date: Thu, 23 Jul 2020 14:47:28 +0200
> From: Carsten Bormann <cabo at tzi.org>
> To: Stephane Bortzmeyer <bortzmeyer at nic.fr>
> Cc: internet-history at elists.isoc.org
> Subject: Re: [ih] Arpanet physical connectors
> Message-ID: <1BED3CA8-6642-41DB-BE39-84D24ECA65EA at tzi.org>
> Content-Type: text/plain; charset=utf-8
>
> On 2020-07-23, at 14:02, Stephane Bortzmeyer via Internet-history <
> internet-history at elists.isoc.org> wrote:
> >
> > Dear Internet History nerds, what did the connectors and cables of the
> ARPAnet look like?
>
> I wasn?t there, but BBN 1822 says (local host):
>
> "The Host cable supplied with the 516 IMP and the Pluribus IMP is 30 feet
> long and contains 12 RG 174/U coaxial conductors with grounded shields.
> Host personnel must provide an appropriate connector for the Host end of
> the cable."
>
> "The Host cable supplied with the 316 IMP is 30 feet long and contains 32
> twisted pairs. The cable is terminated at the IMP end with a paddle card
> which plugs directly into the 316 Host interface. Each pair of the cable
> consists of a colored wire and a black wire numbered with the pin number of
> the paddle card to which the colored wire is connected. All black wires
> connect to the paddle card signal ground. Host personnel must provide an
> appropriate connector for the Host end of the cable."
>
> The interface had a four-way handshake per bit (3 lines + a separate end
> of message signal), and was based on 0V/5V signals (*). Distant host was
> differential -6V/6V.
>
> Fortunately, at the time, most people working in this space had the EE
> degree to be able to work with this :-)
>
> Gr??e, Carsten
>
> (*) "All Host-IMP logic signals (Data, Ready-For-Next-Bit,
> There's-Your-Bit, Last Bit) are unbalanced, source-terminated lines with a
> nominal characteristic impedance of 68 ohms. The line is terminated at the
> driving end with the characteristic impedance. The receiver is ideally an
> open circuit; in practice,
> TTL gate. In this scheme a voltage step of half the nominal level is
> propagated from source to receiver. At the receiver, it is reflected by
> the high impedance termination, resulting in a full level step at the
> receiver and another half level step propagating back to the source, where
> it is absorbed by the termination.?
>
> "The IMP will properly receive 5-volt logic signals; however, signals from
> the IMP may go to 6 volts. Therefore, the Host must provide a voltage
> divider, if these signals are to be received by normal 5-volt logic, to
> prevent destruction of the receiving circuit."
>
> ------------------------------
>
> Message: 5
> Date: Thu, 23 Jul 2020 09:06:06 -0400
> From: "Andrew G. Malis" <agmalis at gmail.com>
> To: Stephane Bortzmeyer <bortzmeyer at nic.fr>
> Cc: "Nelson H. F. Beebe via Internet-history"
> <internet-history at elists.isoc.org>
> Subject: Re: [ih] Arpanet physical connectors
> Message-ID:
> <CAA=
> duU1Wb+4h3jVo3urax3sHBXZEkUSAxx6ow6y34io0MhcbKg at mail.gmail.com>
> Content-Type: text/plain; charset="UTF-8"
>
> Stephane,
>
> See http://www.bitsavers.org/pdf/bbn/imp/BBN1822_Jan1976.pdf ,
> particularly
> section 4 and Appendices B-D for a complete description.
>
> I've also been searching for an image of the actual connector, but have
> been unable to find it.
>
> Cheers,
> Andy
>
>
> On Thu, Jul 23, 2020 at 8:02 AM Stephane Bortzmeyer via Internet-history <
> internet-history at elists.isoc.org> wrote:
>
> > https://twitter.com/nielstenoever/status/1286254151874293760
> >
> > Dear Internet History nerds, what did the connectors and cables of the
> > ARPAnet look like?
> >
> > Were these serial cables? Seems likely because the RS-232 standard dates
> > back to 1960.
> > --
> > Internet-history mailing list
> > Internet-history at elists.isoc.org
> > https://elists.isoc.org/mailman/listinfo/internet-history
> >
>
>
> ------------------------------
>
> Message: 6
> Date: Thu, 23 Jul 2020 13:12:45 +0000 (UTC)
> From: Alex McKenzie <amckenzie3 at yahoo.com>
> To: Stephane Bortzmeyer <bortzmeyer at nic.fr>, "Andrew G. Malis"
> <agmalis at gmail.com>
> Cc: "Nelson H. F. Beebe via Internet-history"
> <internet-history at elists.isoc.org>
> Subject: Re: [ih] Arpanet physical connectors
> Message-ID: <1497811355.6012344.1595509965222 at mail.yahoo.com>
> Content-Type: text/plain; charset=UTF-8
>
> As mentioned in another post, BBN supplied the cable and specified the
> signals, but it was up to each Host to supply its own connector.
> Cheers,Alex
>
> On Thursday, July 23, 2020, 9:06:31 AM EDT, Andrew G. Malis via
> Internet-history <internet-history at elists.isoc.org> wrote:
>
> Stephane,
>
> See http://www.bitsavers.org/pdf/bbn/imp/BBN1822_Jan1976.pdf ,
> particularly
> section 4 and Appendices B-D for a complete description.
>
> I've also been searching for an image of the actual connector, but have
> been unable to find it.
>
> Cheers,
> Andy
>
>
> On Thu, Jul 23, 2020 at 8:02 AM Stephane Bortzmeyer via Internet-history <
> internet-history at elists.isoc.org> wrote:
>
> > https://twitter.com/nielstenoever/status/1286254151874293760
> >
> > Dear Internet History nerds, what did the connectors and cables of the
> > ARPAnet look like?
> >
> > Were these serial cables? Seems likely because the RS-232 standard dates
> > back to 1960.
> > --
> > Internet-history mailing list
> > Internet-history at elists.isoc.org
> > https://elists.isoc.org/mailman/listinfo/internet-history
> >
> --
> Internet-history mailing list
> Internet-history at elists.isoc.org
> https://elists.isoc.org/mailman/listinfo/internet-history
>
>
> ------------------------------
>
> Subject: Digest Footer
>
> Internet-history mailing list
> Internet-history at elists.isoc.org
> https://elists.isoc.org/mailman/listinfo/internet-history
>
>
> ------------------------------
>
> End of Internet-history Digest, Vol 10, Issue 34
> ************************************************
>
More information about the Internet-history
mailing list