PACKET SWITCHING: They Labored too

Nothing is invented and perfected at the same time…


A Latin Proverb

The inventions of mortal men are no less mortal….


Sir Walter Raleigh in History of the World

These ancient proverbs must be kept in mind when commenting on the article A
Paternity Dispute Divides Net Pioneers* by Katie Hafner: NY Times, 8 November,
2001.

It is generally not important who stakes the claim for an invention. Often,
the honor bestowed by peers in the scientific fraternity carries more weight.
But our society also has a bad habit of ignoring many important facts or events
when determining the real inventors. The purpose of this paper is to highlight
additional events that really shaped data communications during the period from
1955 to 1968.

Let
us consider John Brady’s book "The Communications Miracle: The
Telecommunications Pioneers from Morse to the Information Super Highway"
(Plenum Press, 1995). It does not mention the names of Arthur Collins, a real
genius behind the radio and Single-Side-Band (SSB) communications, and all the
names mentioned in the NY Times article. Due to his background, he acknowledges
only the British and Bell Laboratories Pioneers.

John Naughton, author of A Brief History of the Future: From Radio Days to
Internet Years in a Lifetime (Overlook Press, 2000) credits Dr Davies (who
presented a paper at the 1968 IFIP congress held at Edinburgh in 1968 dealing
with PS system) and Paul Baran (who wrote the famous 1964 paper envisioning a
highly distributed and survivable network based on packet switching) with having
jointly invented packet switching. However, Professor Leonard Kleinrock has now
staked his own claim on having invented packet switching (PS) (as based on his
1961 PhD dissertation at MIT and his book, Communication Nets: Stochastic
Message Flow and Delay (McGraw-Hill, 1964). I consider that an invention must
always describe a prototype of the equipment needed to achieve full resource
sharing in the PS node and transmission lines served by the node. If one
considers the above facts, Dr Donald Davies becomes the real winner considering
the papers also delivered at the 1968 IFIP Congress by his collaborators at
Great Briton’s NPL namely, PT Wilkinson, RA Scantlebury and KA Bartlett.

The real purpose of this paper is to consider additional historical facts
that led to the actual development of the first store-and-forward message
switching ( S/F MS) system that also included a PS application to utilize a fast
I/O channel connecting a S/F MS node to the reservation computer. Collins Radio
Company had already developed the first commercial modem during 1955 based on an
original patent awarded to their famous inventor, Melvin L Doelz. It employed
differentially coherent phase phase-shift keying (DCPSK) and delivered up to
4,800 bps. It was housed in a refrigerator-size cabinet. By 1958, it was widely
used to connect remote job entry (RJE) stations with a mainframe computer, and
to connect remote sensing devices in Canada to provide early warning to the US
authorities in case of any air-space violation by an enemy. Such two-node
networks (also called point-to-point connections) also employed proprietary
coded feedback communication logic to detect errors and sometimes forward error
correcting codes using both Hamming and cyclic redundancy codes.

Based on their success in the area of the so-called 2-node point-to-point
networks, Collins Radio Company was immediately approached by airlines,
railroads and banks to automate the manually operated torn-tape teletype (TTY)
machines (similar to the telex machines) employed to send short messages between
TTY machines connected by low-speed leased lines at enterprise locations.

To meet the new communications needs of enterprises, Collins Radio Company
devised a two-step plan. They first acquired a newly developed read-only memory
developed by Aeronutronic Division of Ford Motor Co. located across the street
from Collins Radio Company in Newport Beach, California. These ancient ROMs
employed tiny magnetic rings (each called a CORE and each storing either a 0 or
a 1). These ROMs were immediately converted to a first micro-programmable
communications processor. One could create any set of basic communication
instruction for the following purposes:

l Breaking the incoming message into
fixed sized blocks and reconstituting the output message

l Performing queuing/dequeuing from RAM
storage

l Converting the TTY keystrokes into
computer codes

l Performing disk read/writes

l Executing magnetic tape read/writes,
and to do a myriads of other useful instructions required for error detection

Once the communication processor was functioning, the first commercial S/F MS
system soon became a reality in 1963. I described the actual (not just based on
pure theory) performance of such a system in a paper delivered at the 1965 IFIP
Congress held in New York. Most airlines, railroads and banks acquired
standalone S/F MS nodes as message exchanges.

Their existing TTYs became their sources or destinations for messages, thus
preserving their existing investments. In 1968, Collins S/F MS became a FE for
American Airline’s Reservation system. Soon it became a standard for all
airlines due to its multi-processor reliability and additional availability
offered by the S/F MS capability. In contrast, the IBM FE would go down each
time the reservation computer went down. One of the requirements for a Collins
S/F MS/FE was also to provide a fully shared use of the fast I/O channel
connecting the FE and the reservations computer. We called it the core switch
for expediency. We never knew that it employed the packet switching technique
until I presented my own paper dealing with disc file performance at the 1968
IFIP Congress and actually heard papers of Davies and his collaborators.

Melvin L Doelz (the holder of original patent on the first modem) directed
both of the above efforts. Doelz also holds a majority of patents dealing with
Collins S/F MS and FE systems. Since S/F MS and PS nodes both employ a
store-and-forward technique and since a core switch was already implemented by
1968 as a simple subset of the MS functions, it should be imperative to
acknowledge the contributions of Melvin Doelz. It also proves that as systems
become highly complex (e.g. a PS or S/F MS node) it becomes very hard to
acknowledge only one person as the sole inventor.

A similar confusion is now being experienced when considering real inventors
of the Internet or the Wide World Web. The Internet has two important components–the
TCP/IP protocol which is just another incarnation of packet switching developed
at ARPA during the period 1963-1968; and Hyper Text Markup Language (HTML) as
invented by Tim Berners-Lee of CERN, Switzerland during 1989. It is interesting
to note that TCP/IP was around for over 21 years before the Internet revolution
really started. And that happened only after the availability of HTML. It is
really crazy when we don’t hear the name of Tim Berners-Lee in the same breath
when crediting the invention of Internet to Vinton Cerf, Lawrence Roberts and
Robert Kahn (all originally from ARPA) and Leonard Kleinrock of UCLA.

According to the two proverbs quoted at the beginning of the article, all the
names mentioned above deserve credit for the invention of very dynamic and
complex entities such as packet switching and Internet. We should also
acknowledge all those we have not heard of yet but who also labored and waited
in a similar vein to what Milton said in his famous epic, Paradise Lost.

Roshan Lal Sharma adjunct professor
Southern Methodist University

*The reproduction of the NYT article can be read at http://www.interesting-people.org/archives/interesting-people/200111/msg00121.html
or

https://mail.lab.net/lists/archive/ip-exploder/2001-November/001603.html