White Paper on Next Generation Internet R&D Programs of the Federal Government Submitted by Charles N. Brownstein, Cross-Industry Working Team A. Introduction This paper offers industry perspectives on program planning and implementation, and suggests several areas for research for the proposed Next Generation Internet (NGI) program of the federal government. The idea of coordinated federal interagency program development and joint project support, active collaboration and co-investment by government, industry and the academic community is sound. However, the present program plan is deficient in the following ways: 1) there appears to be too much emphasis on incremental advances to present day Internet technology and applications; 2) the planned funding levels cannot adequately support both the advances in technology and applications that can become the foundations for the next generation Internet and also fund a higher speed production network for the university and other education community; either goal will suffer under inadequate funding, and at best, the program will result in real and perceived subsidies to the non-research missions of the participating agencies. Explicit attention to defining the purpose of the program, and to adequately funding the components, is needed. 3) some of the hardest technical and policy issues facing the Internet involve privacy, security and property rights management. These are so tied to the general public benefit and welfare that government leadership is critical and should be a leading component of the NGI program. B. Approaches and Principles The NGI program can be most successful if it can stay consistent with contemporary technology and business environments. In particular, commercial, private sector Internet data transport facilities and technologies are the infrastructure upon which even advanced R&D should rely. The rapidly expanding installed base of Internet technologies should form the core of future generations. Thus the proposed program and its projects need to promote and support: a) overall network integrity; b) interoperability and herterogenaety; c) open technology standards. The present private investment in the private sector for Internet services and applications with near term payoff is large relative to the proposed federal investment. As a result, the federal program needs to focus, in the main, on longer term payoffs. Similarly, the pace of applied research, development and product introduction in industry is fast relative to the pace of fundamental academic research, and industry has access to the university research community in many more ways than ever before. To attract industry investment, the federal programs need to be of reasonably long-term duration. (This should not be confused with the time duration of awards: a mix of relatively short term and long term commitments may be needed as appropriate to specific program substance). The program needs to employ open competition to encourage and select superior proposals. Only open competition can mirror the winnowing processes of selection and the discipline provided by the commercial marketplace. Open competitions with technical excellence as the primary criteria (rather than block grants to existing institutional clients, desired social benefit, or local and regional economic development) should be used as the primary means of award selection to interest industry partners. The program needs to devote the majority of its proposed activites (and the resources actually obtained from the Congressional Appropriations process) toward fundamental advances in technology and its application. Such advances can include barrier breaking or deadlock breaking projects which create existence proofs about technology capability or utility, even if they use known existing technology. Such advances may require general support for common use facilities- but the costs of any such facilities will be of such magnitude that they should be shared substantially by the institutions and organizations which use them. There is no sound rationale and little industry interest in using federal-industry partnerships for incremental research or any form of product development. The program needs to focus beyond the simple, albeit powerful, technical drivers of "brute force" speed improvements in data transmission and computing. This approach is being exploited within present private sector investment and other R&D efforts. Rather, the program needs to adopt "broad systems views" which combine technologies, applications, organizational and business models, skilled people, social and economic settings and social requirements (safety, privacy, security, reliability and property protections and economic viability). The program needs to assume that its research should identify new limits and open new problem areas, and that risk acceptance and tolerance of failure are positive features which add value to the massive private investment which supports the research. It should place high priority on areas of research which may produce benefice for all and which do not create marketplace advantages for any single marketplace participant. The program needs to employ a continuous diffusion model (demonstrated in the HPCC initiative) in order to sustain the pace of technology transfer and utilization. Features of this process are as follows: 1) The results of publicly funded parts of the program should be available for use by the interested research and commercial community as they are produced, during the course of the program to maximize opportunities for evaluation and permit rapid technology transfer; there is a special opportunity, given the present interest and need, for the rapid diffusion of technology which enhances privacy, security, reliability and the general safety of the network and its users. 2) The research performers and managers should be able to learn from the experiences of participants so as to incorporate shared knowledge into their individual research efforts; program coordination and information sharing should be required in the NGI program; 3) The program needs to plan for technology transition. This can be done in several ways. First is the training of skilled personnel from the beginning. Human resource development and mobility are the proven efficient mechanism for technology transfer. There can be strong benefits to the economy and to society from investment in graduate student support at reassert universities, as well as to carefully focused post graduate and career retraining for qualified personnel. Second is attention to prototyping end- user software for all major platforms to minimize barriers to diffusion and promote the use of results among all segments of the potential market. Third is to emphasize "middleware" approaches to further reduce platform dependencies. Fourth is to emphasize "rapid prototypng" and open beta testing strategies to continually diffuse the results in ways that feed back to the program. Fifth is to employ the Internet itself as the major diffusion and feedback mechanism. C) Suggested Areas of Research Several areas of which can have high payoff are as follows: - models and mechanisms for flexible and dynamic configuration of networks and applications. This assumes that open standards prevail and capable devices are everywhere. The idea is to get safe, personalized environments on demand as users need full nomadic support with interoperability, process control and accounting, easy state configuration, and means of recovery from failures and interruptions via explicit flow segmentation and segment accounting. - models and mechanisms for "forensic quality" auditing of digital objects and processes in the networked environment for both privacy and security- development of appropriate primitives, mechanisms and tools, notions of classes of audit capability, authorization, and service level guarantees, experimental safety enhancing technology; - models and methods for policy management-how to describe, manage and enforce user defined policies within and across multiple administrative domains; - architectural models for active networks--how to build networks with various internal operational and service components (and the end-systems that connect to them so that the overall system becomes more (rather than less) robust as the network scales to large size and becomes more heterogeneous; - mobility architectures for nomadic usage of a wide variety of untethered computing and communications devices capable of interacting directly and via heterogeneous (radio based and wireline networks); - ways to deal with capability heterogeneity- the huge variations in the link speeds from kilobit to gigabits, including large asymmetry in two directions (e.g., cable down, POTS up); - theoretical and experimental research on the policy and economics of future infrastructure class Internt- how real end user policies can be scalably reflected in the fully distributed federated Internet and how accounting and cost recovery models can encourage development of new capabilities while remaining attractive to consumers of the services; - experimental consumer network services, particularly to the home, regarding making networks easier and more robust to deploy and access; - interoperable traffic analysis and management tools- which make sense across industry segments, and traffic models and analysis tools for multimedia (to replace present POTs based tools; - traffic capacity and congestion measurement models and effective traffic monitoring methods with capabilities to "look into" the public networks to retrieve shared capacity and congestion information on route availability; - studies of integrated services for real time and asynchronous data flows. Finally, it is appropriate that the NGI program support facilities for the research it funds. This should not be confused with nor encumbered by the provision of facilities for the general use of the research and education communities for purposes other than NGI program activities. Of expenditures for facilities, a high priority should be placed on creating "community accessible" high performance application testbeds. These should serve as the facility base for transient projects. To the extent reasonable, such facilities should employ multiple equipment vendors with inter-operable, standards-based solutions, rather than relying on a single vendor or network provider. Such "testbed class" experimental networks should selected by open competition. All should be the product of closely coordinated interagency collaboration, funding, and management, and a single point of contact should be available to industry and the academic community for interaction. Charles N. Brownstein Executive Director Cross-Industry Working Team (XIWT) Corporation for National Research Initiatives 1895 Preston White Drive Suite 100 Reston, VA 22191-5434 Internet mail: cbrownst@cnri.reston.va.us Web: Tel: (703) 620-8990 Fax: (703) 620-0913