Network Working Group V. Cerf
Request for Comments: 1217 CSCR
1 April 1991
Memo from the Consortium for Slow Commotion Research (CSCR)
Status of this Memo
This RFC is in response to RFC 1216, "Gigabit Network Economics and
Paradigm Shifts". Distribution of this memo is unlimited.
To: Poorer Richard and Professor Kynikos
Subject: ULSNET BAA
From: Vint Cerf/CSCR
Date: 4/1/91
The Consortium for Slow Commotion Research (CSCR) [1] is pleased to
respond to your research program announcement (RFC 1216) on Ultra
Low-Speed Networking (ULSNET). CSCR proposes to carry out a major
research and development program on low-speed, low-efficiency
networks over a period of several eons. Several designs are
suggested below for your consideration.
Military requirements place a high premium on ultra-robust systems
capable of supporting communication in extremely hostile
environments. A major contributing factor in the survivability of
systems is a high degree of redundancy. CSCR believes that the
system designs offered below exhibit extraordinary redundancy
features which should be of great interest to DARPA and the
Department of Defense.
This system uses a highly redundant optical communication technique
to achieve ultra-low, ultra-robust transmission. The basic unit is
the M1A1 tank. Each tank is labelled with the number 0 or 1 painted
four feet high on the tank turret in yellow, day-glo luminescent
paint. Several detection methods are under consideration:
(a) A tree or sand-dune mounted forward observer (FO) radios
to a reach echelon main frame computer the binary values
Cerf [Page 1]
RFC 1217 ULSNET BAA April 1991
of tanks moving in a serial column. The mainframe decodes
the binary values and voice-synthesizes the alphameric
ASCII-encoded messages which is then radioed back to the
FO. The FO then dispatches a runner to his unit HQ with
the message. The system design includes two redundant,
emergency back-up forward observers in different trees
with a third in reserve in a foxhole.
(b) Wide-area communication by means of overhead
reconnaissance satellites which detect the binary signals
from the M1A1 mobile system and download this
information for processing in special U.S. facilities in the
Washington, D.C. area. A Convection Machine [2] system
will be used to perform a codebook table look-up to decode
the binary message. The decoded message will be relayed
by morse-code over a packet meteor burst communications
channel to the appropriate Division headquarters.
(c) An important improvement in the sensitivity of this system
can be obtained by means of a coherent detection strategy.
Using long baseline interferometry, phase differences
among the advancing tank column elements will be used to
signal a secondary message to select among a set of
codebooks in the Convenction Machine. The phase analysis
will be carried out using Landsat imagery enhanced by
suitable processing at the Jet Propulsion Laboratory. The
Landsat images (of the moving tanks) will be correlated
with SPOT Image images to obtain the phase-encoded
information. The resulting data will be faxed to
Washington, D.C., for use in the Convection Machine
decoding step. The remainder of this process is as for (b)
above.
(d) It is proposed to use SIMNET to simulate this system.
Using the 16" guns of the Battleship Missouri, a pulse-code modulated
message will be transmitted via the Pacific Ocean to the Ames
Research Center in California. Using a combination of fixed and
towed acoustic hydrophone arrays, the PCM signal will be detected,
recorded, enhanced and analyzed both at fixed installations and
aboard undersea vessels which have been suitably equipped. An
alternative acoustic source is to use M1A1 main battle tanks firing
150 mm H.E. ordnance. It is proposed to conduct tests of this method
in the Persian Gulf during the summer of 1991.
Cerf [Page 2]
RFC 1217 ULSNET BAA April 1991
The ULS system proposed in (2) above has the weakness that it is
readily jammed by simple depth charge explosions or other sources of
acoustic noise (e.g., Analog Equipment Corporation DUCK-TALK voice
synthesizers linked with 3,000 AMP amplifiers). An alternative is to
make use of the ultimate in jam resistance: neutrino transmission.
For all practical purposes, almost nothing (including several light-
years of lead) will stop a neutrino. There is, however, a slight
cross-section which can be exploited provided that a cubic mile of
sea water is available for observing occasional neutrino-chlorine
interactions which produce a detectable photon burst. Thus, we have
the basis for a highly effective, extremely low speed communication
system for communicating with submarines.
There are a few details to be worked out:
(a) the only accelerator available to us to generate neutrino
bursts is located at Batavia National Laboratory (BNL).
(b) the BNL facility can only send neutrino bursts in one
direction (through the center of the Earth) to a site near
Tierra del Fuego, Chile. Consequently, all submarines must
be scheduled to pass near Tierra del Fuego on a regular
basis to coincide with the PCM neutrino signalling from
the BNL source.
(c) the maximum rate of neutrino burst transmission is
approximately once every 20 seconds. This high rate can be
reduced considerably if the pwer source for the accelerator
is limited to a rate sustainable by discharging a large
capacitor which is trickle charged by a 2 square foot solar
panel mounted to face north.
(a) Anti-Huffman Coding. The most frequent symbol is
assigned the longest code, with code lengths reducing with
symbol probability.
(b) Minimum likelihood decoding. The least likely
interpretation of the detected symbol is selected to
maximize the probability of decoding error.
(c) Firefly cryptography. A random signal (mason jar full of
fireflies) is used to encipher the transmitted signal by
optical combining. At the receiving site, another jar of
fireflies is used to decipher the message. Since the
Cerf [Page 3]
RFC 1217 ULSNET BAA April 1991
correlation between the transmitting and receiving firefly
jars is essentially nil, the probability of successful
decipherment is quite low, yielding a very low effective
transmission rate.
(d) Recursive Self-encapsulation. Since it is self-evident that
layered communication is a GOOD THING, more layers
must be better. It is proposed to recursively encapsulate
each of the 7 layers of OSI, yielding a 49 layer
communications model. The redundancy and
retransmission and flow control achieved by this means
should produce an extremely low bandwidth system if,
indeed, any information can be transmitted at all. It is
proposed that the top level application layer utilize ASN.1
encoded in a 32 bit per character set.
(e) Scaling. The initial M1A1 tank basis for the land mobile
communication system can be improved. It is proposed to
reduce the effective data rate further by replacing the
tanks with shuttle launch vehicles. The only slower method
of signalling might be the use of cars on any freeway in the
Los Angeles area.
(f) Network Management. It is proposed to adopt the Slow
Network Management Protocol (SNMP) as a standard for
ULSNET. All standard Management Information Base
variables will be specified in Serbo-Croatian and all
computations carried-out in reverse-Polish.
(g) Routing. Two alternatives are proposed:
(1) Mashed Potato Routing
(2) Airline Baggage Routing [due to S. Cargo]
The former is a scheme whereby any incoming packets are
stored for long periods of time before forwarding. If space
for storage becomes a problem, packets are compressed by
removing bits at random. Packets are then returned to the
sender. In the latter scheme, packets are mislabelled at the
initial switch and randomly labelled as they are moved
through the network. A special check is made before
forwarding to avoid routing to the actual intended
destination.
CSCR looks forward to a protracted and fruitless discussion with you
on this subject as soon as we can figure out how to transmit the
proposal.
Cerf [Page 4]
RFC 1217 ULSNET BAA April 1991
NOTES
[1] The Consortium was formed 3/27/91 and includes David Clark,
John Wroclawski, and Karen Sollins/MIT, Debbie Deutsch/BBN,
Bob Braden/ISI, Vint Cerf/CNRI and several others whose names
have faded into an Alzheimerian oblivion...
[2] Convection Machine is a trademark of Thoughtless Machines, Inc.,
a joint-venture of Hot-Air Associates and Air Heads International
using vaporware from the Neural Network Corporation.
Security Considerations
Security issues are not discussed in this memo.
Author's Address
Vint Cerf
Corporation for National Research Initiatives
1895 Preston White Drive, Suite 100
Reston, VA 22091
Phone: (703) 620-8990
EMail: CERF@NRI.RESTON.VA.US
Cerf [Page 5]