Interesting People mailing list archives
IP: INTERNET and RELATED HISTORY -- Just the facts Ma'm
From: David Farber <farber () cis upenn edu>
Date: Mon, 20 Mar 2000 16:48:38 -0500
>To: Recipients of CYHIST digests <CYHIST () MAELSTROM STJOHNS EDU>_ >Date: Fri, 17 Mar 2000 16:16:25 -0800From: "Willis H. Ware" <willis () RAND ORG>>Subject: INTERNET and RELATED HISTORY______________________________________________________________________ Community Memory: Discussion List on the History of Cyberspace ______________________________________________________________________ -- Folder: YES -- A few issues ago, Les Earnest commented that Paul Baran's work had been classified and therefore, was generally unknown to the world at large. Someone else believed that the motivation for his work was a requirement for a survivable command-control system; and lastly, someone asserted that the contract that supported him came from the DoD. I don't want to reopen the sporadic argument about who knew what and who talked to whom, or what influenced what; nor do I wish to engage in this Digest's prior discussion of whether mythology (associated with some field) is a legitimate topic for historians of that field. I'll address just the three points above. I was in the management structure of RAND's Computer Sciences Department at the time of Paul's work, and interacted with him continuously. I can do what Sgt. Joe Friday always demanded: (in paraphrase) "Give me the facts; just the facts." 1. First, contract support. In the period of interest (1960s) RAND's two major contracts were with the USAF and the AEC. However, all the work on distributed communications was done under Project RAND (USAF) funding. 2. Next, the question of classification; here's the publication chronology. a. The first - and unclassified -- publication to discuss the payoff from redundancy in a communications system was: May 27, 1960, Reliable Digital Communications Systems Using Unreliable Network Repeater Nodes, P. Baran, The RAND Corporation Paper P-1995. It suggested a fishnet-like structure for a network and introduced (in>1960) the idea of "hot potato routing" for fixed length message blocks(described in more detail in documents noted below. b. Another slightly later unclassified paper of relevance was: September 1962, On Distributed Communications Networks, P. Baran, The RAND Corporation. Paper P-2626. This paper spelled out the distributed communications system and specifically described the main characteristics commonly associated with packet switching; including packet housekeeping and separation of the physical address from the logical address to allow users moving in the network to maintain their addresses. Concurrently with the above a longer series of documents was in preparation and we intentionally withheld publication of any until all could be released concurrently in August 1964. c. The series, often referred to as the "dozen research memoranda" were: August 1964, On Distributed Communications Vol. I. Introduction to Distributed Communications Networks, P. Baran, RAND Corporation, RM-3420-PR. Vol. II. Digital Simulation of Hot-Potato Routing in a Broadband Distributed Communications Network; Sharla P. Boehm, and P. Baran, RM-3103-PR. Vol. III. Determination of Path Lengths in a Distributed Network, J. W. Smith, RM-3578-PR. Vol. IV. Priority, Precedence and Overload, P. Baran, RM-3638-PR. Vol. V. History, Alternative Approaches, and Comparisons, P. Baran, RM-3097-PR. (This is must reading for anyone who wants to understand what work was going on at the time. Each new idea is almost always based to some degree on previous work.) Vol. VI. Mini-Cost Microwave, P. Baran, RM-3762-PR. Vol. VII. Tentative Engineering Specifications and Preliminary Design for a High-Data-Rate Distributed Network Switching Node, P.> Baran, RM-3763-PR.Vol. VIII. The Multiplexing Station, P. Baran, RM-3764-PR. Vol. IX. Security, Secrecy and Tamper-Free Considerations, P. Baran, RM-3765-PR. Vol. X. Cost Estimate, P. Baran, RM-3766-PR. Vol. XI. Summary Overview, P. Baran, RM-3767-PR. NOTICE that there were only 11 -- not 12 -- of them; and moreover, everything cited above is now and always has been unclassified. d. There were two collateral documents: RM-4236-PR by P. Baran and Rein Turn: Cryptophone: An All-digital Telephone Privacy Subsystem for the Distributed Adaptive Message Block Network (U) (Confidential) and RM-3247-PR by Rein Turn, Computer Simulation of the Cryptophone (U) (Confidential) Since each of these could stand alone, we decided not to put them into the series. While classified at the time, the declassification interval has long since run out. e. Finally, numbers 12 and 13 appeared and were included in the series which implies that the titles started out with "On Distributed ...": RM-5067-PR by P. Baran, July 1966. Vol. XII On Distributed Communications: Weak Spots and Patches (For Official Use Only); and Vol XIII by P. Baran and R. Heirschfeldt, RM-5174-PR, On Distributed Communications: User Locations for the Digital Distributed Communications System (U). (Confidential). These last four RM's were either "for official use only" or CONFIDENTIAL because 1) we discussed specific vulnerabilities, 2) we dealt with cryptography and 3) we dealt with real world geometry and physical vulnerability. Depending upon how one counts, the "dozen series" had either 11, 13, or 15 documents in it. As I mentioned above, however, the declassification interval has run out. I cannot know of course who read what or who, out of our earshot, might have talked to whom. What I can certify is that ARPA was on the distribution list for all documents, and that the ARPA folks had security clearances in any event. Based on calendar dates that others have mentioned and provided in past postings to this DIGEST, Paul's unclassified>work -- the first 11 in the series -- predated Larry Roberts' tenure atARPA; but we do know from various records (such as trip reports) that the work had been discussed with Lick Licklider and Bob Taylor. The bottom line: by August 1964, full and extensive engineering details were in the public domain because we had an extensive distribution list for this work, and we know from the records that ARPA [DARPA] was included. All essential key points were discussed in unclassified volumes. At that time, RAND document distribution always included a lengthy list of deposit campus and urban libraries; so these documents would have also been in such places. Hence, my phrase: ...in the public domain. RAND's policy also was to distribute P's (papers prepared for publication) as well as RM's (Research Memoranda) to clients. The record is clear; Paul Baran's work on Distributed Communications, so far as all essential engineering and system details are concerned, was never classified and it was widely distributed and available. 3. Motivation -- and here semantics becomes important. The USAF and national policy at the time was mutual assured destruction which meant that if the USSR were to launch a nuclear strike, the US would have to be able to launch its nuclear forces in spite of damage inflicted by a USSR first strike. In the jargon of the time, the emphasis was on getting out "the go code" which was a fairly brief highly encrypted message specifying what nuclear attack plan was to be executed and its time table. It would be sent to all nuclear forces, land-air-sea based. The communications infrastructure to support such a last-ditch force deployment of course functioned as a command-control system of sorts but not in the usual sense of the phrase; for example, there was no return path for communications from the forces to the command structure; there was no intent to use the>arrangement to prosecute a lengthy war.I personally discussed such issues with the USAF in that interval and would ask: how do you expect to reconstitute any surviving forces for subsequent strikes? The attitude was (and with some justification since no one could really know what might survive): "...we'll make do with whatever we have left....we'll cobble things together." Given such a posture in the national policy and given the brief nature of the go code, Paul's first proposal to the USAF was for in fact a single teletypewriter channel using the AM network in a distributed communications arrangement. It could be thought of as a minimal command control system, but certainly not an all-up C-C system with which to manage a long-term war. The USAF asserted that it needed more bandwidth and the final proposal did in fact provide it. So the motivation for Paul's work was to provide a minimal but highly survivable one-way communications arrangement to get out the go-code; it was NOT motivated by a requirement for a survivable command-control system that could support the forces fully in both peace and in war. Later on, of course, packet technology became involved as a component of contemporary C-C systems, and along the way, a part of the ARPANET was folded into Defense Data Net, the MILNET, etc. Without knowing in detail how the nuclear forces were intended to be used in the 60s, it is easy to understand how things would get confused. The usage of "command control network" is not unique, is sometimes ambiguous, and often context dependent. If someone wants to pursue the diffusion of the ARPANET and its technology into the support-infrastructure of the military services, probably the best individual to write about it would be Steve Walker, who was in the Office of Secretary of Defense at the time. He was the founding president of Trusted Information Systems, is now a venture capitalist, and can be>reached at steve () stevewalker com
Current thread:
- IP: INTERNET and RELATED HISTORY -- Just the facts Ma'm David Farber (Mar 20)