[6] [Editor's note: Robert Kahn is credited with being the system designer of the ARPANET and the architect of the Internet. The following is an excerpt from the Supplemental Background information which he submitted with his testimony before the Congressional Subcommittee on Basic Research on March 31, 1998.] From the Internet: Some Background* By Robert Kahn In the early 1970s, DARPA was exploring radio and satellite-based packet networks along with the ARPANET. Each network had different communication speeds, interfaces, packet sizes and internal operations. After joining DARPA, I became the principal architect of the packet radio network, a high-speed forerunner to today's CDMA cellular technology. I also assumed management responsibility for creating a packet satellite network, which was ultimately deployed on Intelsat IV and linked several European sites with a kind of "ethernet in the sky." The challenge, back then, was to connect these three different packet networks into a seamless whole whereby any computer on one of the three networks could talk to any computer connected to any one of the three networks without necessarily knowing the location of the other sites or the underlying network connectivity. The Internet resulted from this effort to connect those three networks and their computers in such a way that other networks and computers could be easily connected in the future. At the time, there were no personal computers or workstations as we now know them. Local area networks (such as the ethernet and ring networks) were only in development within various research laboratories, but had not been deployed. By solving the network and computer connectivity problem in a generic way, we were able to ensure that new technological developments in the future could be accommodated. The key technical contribution which enabled this "network of networks" to be constructed was an architecture consisting of gateways (now called routers) which were placed between the networks, and a protocol, now known as TCP/IP, which was used by the computers and the routers. I collaborated with my colleague Vinton Cerf, then at Stanford University, on the development of this protocol which was presented publicly for the first time in September 1973 at a meeting in Sussex, England and published by the IEEE in May, 1974. Subsequently, I enlisted the help of BBN and University College London to work with Stanford in creating the initial implementations of the protocol (for different computers). With support from DARPA, BBN created the initial Internet gateway software for experimental use in the mid 1970s. Until the early 1980s, the Internet was used primarily for experimental purposes. During that period, the protocols were steadily refined and tested. Other networks were connected during that period including many of the early local area networks; a few European research networks were also connected. During this period, the overall management of the Internet was handled by DARPA in the person of either myself or Dr. Cerf, who was with DARPA during the period 1976-1982. Many of the basic issues under consideration in this hearing can be traced to decisions we made during that period. However, since there were few commercial organizations participating at the time, and very little international involvement, decisions we made were largely determined on the basis of logically defensible criteria and fairly complete knowledge of all the relevant matters; fortunately, we were also in charge of the overall research program and, as a result, there was remarkably little controversy about the Internet within the research community. One of the decisions we made during that period was to delegate responsibility for maintaining information about key Internet parameters to Jon Postel, currently a researcher at the University of Southern California (USC) Information Sciences Institute who had been carrying out similar functions for the ARPANET. While DARPA retained the ultimate authority for decisions about policy and procedures, increasingly Jon Postel assumed primary responsibility for these functions, with DARPA retaining an oversight responsibility in the event this was necessary to invoke. During that period, no occasion arose when there was a need to second guess his decisions (although we often would inquire as to how he came up with certain decisions). This function, performed by Jon Postel under USC's contract with DARPA, eventually became known as the Internet Assigned Numbers Authority (IANA) and included certain policy matters associated with domain names as well as IP addresses and protocol parameters. With DARPA's permission, Jon delegated certain clerical and operational functions to SRI International, while retaining other functions. Among the former were the maintenance of a database which mapped Internet names to Internet addresses and making this resource available on the Internet. Moving ahead toward the present, the ARPANET was phased out in 1990 and was effectively replaced by a higher-speed backbone known as NSFNET built by IBM, MCI and Merit under an award from the National Science Foundation (NSF). With encouragement and help from DARPA, NSF took over responsibility for maintaining most of the Internet management infrastructure from Defense, and recompeted the contract that the Defense Department had with SRI International. Network Solutions, Inc. (NSI) won the competition for providing the domain name registration services and has provided this service ever since, with a few exceptions, such as country codes. When the Internet naming service known as the Domain Name Service (DNS) was first proposed in the 1980s by Paul Mockapetris (also from USC/ISI along with Jon Postel) most of the then existing sites could be characterized as educational (EDU), US government (GOV & MIL) or other (this included network (NET), organization (ORG), some commercial sites that had first class research laboratories (COM) and a few special cases involving matters such as testing and multi-national experiments (ARPA and INT)). It was envisioned at the time that the overall database of names, which had previously been so small that it was trivial for a site to download the entire database from SRI daily, might become somewhat unwieldily if the number of hosts or networks increased significantly. Breaking the Internet names into categories such as EDU, COM, etc. would allow them to be managed separately and resolved into IP addresses separately, thus affording an opportunity for efficiency and increased autonomy in the operation of the Internet. In addition, two letter country codes were introduced as domain names that could be managed by individual countries according to policies developed by the countries themselves. It is not necessary that all countries participate, and indeed not all have in the past. The IANA made the determination of who in a given country would be responsible for that countries domain, but gave deference to the legitimate government of the country if it chose to weigh in. In the mid 1990s, the rapid commercial growth of the Internet was fueled in large measure by the success of the NSFNET, the introduction of many commercial Internet Service Providers, the Boucher bill which allowed NSF to open the NSFNET for commercial use (in addition to research and educational use), the continuing attraction of electronic mail and file transfer capabilities, and the subsequent introduction of the point-and-click browser for the World Wide Web. With competitive commercial service available for access to the Internet, NSF reduced its subsidy for the NSFNET and stopped subsidizing the services provided by NSI in order to put them on a pay-as-you-go-basis. NSI has continued to do an excellent job of providing such services for the Internet under a Cooperative Agreement with NSF that is currently due to expire later this year[1998]. However, with several million domain names in existence and the potential for many more in the future, the annual revenue derived from domain name registrations could easily exceed a hundred million dollars per year if the current level of fees were to be maintained. Although the fee for individual domain name registrations has been $50 per year (it has since been announced that the fees will be reduced somewhat), many individuals and organizations have expressed strong feelings that the existing fee structure and organizational arrangements are untenable in the long term and should be rectified. One proposed approach for domain name registration is to require the separation of service provider roles into registries and "registrars", although one party can provide both roles. In this approach, domain name registries would be placed on a not-for-profit basis, with the registrars offering competitive commercial services. I presume this need not imply that the organization running a registry must be non-profit, but only that the function must be based on cost recovery. In this model, NSI and/or other competent organizations could provide this function. Oversight would still have to be provided from some appropriately constituted body. It is still unclear how best to introduce competition in this approach. My view is that, in general, fewer separately managed gTLDs are better than more, but there is no obvious choice of the right number in a competitive environment unless, in principle, it can be arbitrarily large. Still, this general approach of increasing the number of gTLDs, at least as an interim approach, holds considerable appeal and almost all the parties are endorsing the principle but with considerable divergence of opinion about how to achieve it equitably and technically. Another solution is for the U.S. Government to recompete the function, as it did for the InterNIC, according to a set of agreed principles (hopefully with broad community consensus) with a goal of enabling this function to operate in a stable and reliable fashion without direct US government involvement in its operation. Others feel that this can be sorted out completely within the private sector. There would likely still be a need for an oversight role of some sort as there is for any critical societal function (even a competitive one) that cannot be allowed to fail. But even here, there is no consensus yet on what that oversight should be, who should provide it or even that it is needed. More time is needed to reach a consensus on how best to proceed here. ---- *SUPPLEMENTARY BACKGROUND INFORMATION from Testimony before the Subcommittee on Basic Research of the Committee on Science on the subject of Internet Domain Names by Dr. Robert E. Kahn, President and CEO Corporation for National Research Initiatives March 31, 1998. ---------------------------------------------------------------------- ---------------------------------------------------------------------- Reprinted from the Amateur Computerist Vol 10 No 1 Spring/Summer 2000. The whole issue or a subscription are available for free via email. Send a request to jrh@ais.org or see http://www.ais.org/~jrh/acn/ ----------------------------------------------------------------------