N-1-1-040.52 What's Important in Coordinating Internet Activities Internationally, Steven N. Goldstein*, sgoldste@cise.cise.nsf.gov The opportunity to write comes on the heels of the November 13-15, 1991 meetings of the Coordinating Committee for Intercontinental Research Networking (CCIRN), generally pronounced "kern", and its engineering advisory body, the Intercontinental Engineering Planning Group (IEPG), in Santa Fe, New Mexico, USA. These were followed the next week by the meeting of the Internet Engineering Task Force (IETF), including a session of its Operational Requirements Area Directorate (ORAD), also in Santa Fe. In addition, the Federation of American Research Networks (FARNET) met jointly with the ORAD. All of these groups are concerned with a central issue: maintaining stability in the Internet and encouraging network interconnection architectures and engineering practices which in one way or another fit people's views of "optimality". I was not able to attend the IETF/ORAD or FARNET meetings, and the minutes and business of the CCIRN and IEPG meetings have not yet been finalized. So, I will not attempt to report on those events per se, but rather, I will present a general discussion of the ideas that face us all in this arena. I recently read an article about bulletin boards which referred to Fidonet in terms like "chaotic, self-organizing beast", and, by juxtaposing the Internet implied similar qualities for our collective body. There may be a ring of truth to that, especially by limited analogy with fractals in Chaos Theory: LANs connect to MANs and WANs, and, now WANs are connecting into super-WANs, and the trend may take even higher steps of organizing. Yet, this is being done without any central authority. In the U.S., many campuses have several levels of LANs which may ultimately connect to the regional WAN. The regional WAN connects to the NSFNET WAN. Some regional WANs also connect directly to other regional WANs. Also, portions of a campus may connect to the ESnet or the NASA Science Internet WAN, or to the Terrestrial Wideband Net. And, the NSFNET, ESnet, NSI and TWB WANs connect to each other at two Federal Internet eXchanges, FIXs. But, similar things are happening among commercial nets in the U.S., and the CIX (for Commercial Internet eXchange) Association has formed CIXs. There have been proposals to link FIXs and CIXs into National EXchanges (NEXs). There is talk of one or more CIXs in Europe. Japanese research and academic networks are talking of a JIX. A proposed European Backbone, Ebone, would create a supra-national network infrastructure to which national and [intracontinental] international European nets would connect at main nodes, again WAN-to-super-WAN. If one were to visualize each network as a chain link with shape somehow indicative of topology and link thickness and size representing network size (number of connections, traffic levels, capacities, etc.), the result would be a three-dimensional mail (fabric). Some network researchers, engineers and operators assert that the fluttering of the wings of a butterfly in some distant link can cause huge storms throughout the net. Less whimsically stated, some of my colleagues maintain that the present state of Internet technology is not up to the task of protecting the stability of their networks from poor engineering choices in neighboring networks. And, as the technology catches on throughout the world, new links continue to materialize, as do new connections among them. New tools are being developed to try to cope with this, especially hierarchical routing protocols (e.g., the Border Gateway Protocol - BGP) and the ability to interject policy into routing decisions (policy-based routing). So, it is a race of sorts between the proliferation of scale and complexity and tools designed to cope with them to preserve stability and performance. This, then, is the context as I see it. And the question(s), as yet unsolved: "Can we, collectively, create a forum for exchanging information and evaluating proposed linkages before the fact in order to preserve stability and performance in the Internet?" And, relatedly, "Is it possible to have a shared sense of optimality against which alternative solutions emanating in the forum can be evaluated?" Finally, "Under what sets of circumstances might we expect individual network administrations to behave according to the best judgments of other network administrations represented in the forum?" The CCIRN and the IEPG and the IETF/ORAD and FARNET did not achieve closure on these issues this time around, and they may not do so in the next few rounds. Yet other bodies may have to join the forum as the Internet becomes increasingly populated with commercial interests. Yet, the quest must not be given up, because we all live together in the same flat address space, and in one way or another we will share similar fates if instabilities occur. *Program Director, Interagency & International Networking Coordination, Division of Networking and Communications Research & Infrastructure, National Science Foundation.