|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |
Date: Sun, 22 Sep 91 21:28:59 -0500 From: lhl@cs.wisc.edu (L.H. Landweber) To: com-priv@psi.com, ietf@isi.edu Given the recent discussion regarding the demise of CSNET, I thought I would take the opportunity afforded to make a few comments and provide a bit of history. Newspapers last week reported on passage of the NREN authorization bill and the hundreds of millions of dollars it will provide. It is worthwhile remembering that the importance of networking has not always been universally recognized. In the late 1970s there were no widespread academic networks. The groundwork for the present universal acceptance of the value of networking remained to be laid. At the time, the ARPANET and related networks were experimental and were only available to a small group of users who were working on relevant DARPA/DOD-funded research projects. . The idea for CSNET was conceived in early 1979 as a result of conversations between a number of people some of whom at the time were chairs of computer science departments. At the time, about ten to fifteen computer science departments were connected to the ARPANET and other departments were finding that they were increasingly at a disadvantage in a number of areas because they were not part of this communications environment. One possibility for these departments was to join the ARPANET; another was to establish a new network. Since many departments, including ours at Wisconsin, had been unsuccessful in the former, I called a meeting for May 1979 in Madison to discuss how and if a new network might be established. Present at that meeting were Kent Curtis, Director of the NSF Computer Science Section, Peter Denning from Purdue, Rich DeMillo and Phil Enslow from Georgia Tech, Richard Fateman and Larry Rowe from Berkeley, Randy Frank and Tony Hearn from Utah, Bob Ritchie from Washington and Ed Desautels and Larry Landweber from Wisconsin. Bob Kahn from DARPA also attended for part of the meeting. At the meeting, Bob Kahn described an expected effort to provide ARPANET access to many more departments. The participants decided that while an ARPANET expansion would be worthwhile, they would also pursue a parallel effort, formation of a new network, hopefully, having gateways to the ARPANET. In December 1979, after a great deal of discussion, a proposal, co-sponsored by 11 computer science departments, was submitted by Wisconsin to NSF. The primary focus was on use of X.25 public data networks and a variety of information services. The referee reports were mixed. Referees criticized the technical plan for proposing to "reinvent the wheel," suggested that the ARPANET could solve existing problems and made a variety of other negative comments. As a result, the project was put on hold. Kent Curtis suggested a small grant to explore alternatives and obtain more feedback from the community. Bob Ritchie prepared a survey to assess the networking needs of the CS community. In addition, a planning meeting was organized for June 1980 in Berkeley. This meeting was key to the future success of CSNET. At the meeting, Vint Cerf proposed cooperation between DARPA and the CSNET group. In particular, Cerf proposed basing the network on TCP/IP. This would make it easy to establish gateways to the Internet. Most important he offered his help in making the project a success. TCP/IP would be run over X.25 public data networks with IP gateways to other networks. In addition, Dave Farber gave a presentation of work at Delaware on MMDF. As a result, the Phonenet component of CSNET, based on MMDF, was added to the project A second CSNET proposal was submitted which, after receiving another set of mixed reviews, was funded by NSF in January 1981. Four components of the project were funded, Phonenet (Delaware), Nameserver (Wisconsin), IP/X.25 Interface (Purdue) and operation of a Phonenet Relay (RAND). Peter Denning, Dave Farber, Tony Hearn and Larry Landweber formed a Management Committee which was to work with Bill Kern of NSF to oversee the project. In June 1992, a Coordination and Information and Information Center, headed by Dick Edmiston, was established at BBN following an open solicitation. The CSNET grant was for $5 million for five years. A major condition of the grant was that the network become self sufficient by the end of the five years. This was achieved on schedule because of the willingness of university departments to pay dues and because a number of leading industrial research laboratories were willing to pay significantly higher dues than that of the university departments. In addition, all participants paid the direct cost of their network usage. CSNET also benefitted from a large number of volunteers from the CS community. At its peak, CSNET had over 170 academic, government and industrial members. For most of these CS departments and industrial labs, CSNET provided the only access to a national networking environment. CSNET pioneered in a number of areas. First, it demonstrated that users were willing to pay for network services. Prior to CSNET there was great debate in the community as to whether this was the case. Second, CSNET offered a variety of connection methods, with different service levels and costs. Third, CSNET, working with BBN, developed the first community supported NIC and NOC. Fourth, the three software packages developed, MMDF by Delaware, the Nameserver by Wisconsin and the IP/X.25 interface by Purdue all broke new ground. MMDF and the Nameserver are still in use today. The IP/X.25 interface developed the basic ideas for interfaces of this type. A key component of CSNET was its agreement for gatewaying to and from the DARPA Internet. This set CSNET apart from other networking activities. For domestic CSNET members such access was a natural outgrowth of the Berkeley meeting. Details of this agreement were later worked out by Bob Kahn and Larry Landweber; a key principle was that there was to be no charge to Internet users whose traffic traversed CSNET; similarly, CSNET users would not be charged when their traffic traversed the Internet. For international CSNET affiliates, an agreement was negotiated between Kahn and Landweber. in 1983. This agreement required that groups outside the U.S. control the source of traffic flowing to the Internet and the destination of traffic flowing from the Internet. This led to the rapid establishment of mail gateways to Germany, Korea, Japan, Finland and others based on use of MMDF. Both of these agreements broke new ground with respect to the principle of access and openness. At the beginning of CSNET, the founding group agreed that its ultimate success would be demonstrated when it was no longer needed. Today, the availability of NSFNET and the impending NREN have caused that time to arrive. I would like to finish by noting that CSNET would not have been funded had it not been for the efforts of Kent Curtis, head of the NSF CS Section. Curtis understood the importance of networking and was committed to finding a way to support CSNET. It is unfortunate that he passed away before seeing his vision come to fruition in the form of the NSFNET and the NREN. Larry Landweber PS: Many individuals were involved in the establishment of CSNET. Listed below are those from the contractor sites. There were also many volunteers who served on various committees. My apologies to anyone who has been inadvertently omitted. BBN: Laura Breeden, Dick Edmiston, Frank Heart, Dan Long, Craig Partridge DARPA: Vint Cerf, Bob Kahn Delaware: Dave Crocker, Dave Farber, Brendan Reilly NSF: Rick Adrion, Kent Curtis, Bill Kern Purdue: Doug Comer, Peter Denning RAND: Tony Hearn, Mike O'Brien UCAR: Leonard Romney Wisconsin: Larry Landweber, Mike Litzkow, Marvin Solomon PPS: I have many of the original documents and memos which I can make available to those who are doing research on the history of networking.
|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |