Return-Path: Date: Thu, 12 Mar 92 19:14:24 -0500 From: wls@psi.com (William Schrader) To: com-priv@psi.com Subject: my testimony, House Subcomm. on Science (long) Cc: wls@psi.com The House Subcommittee on Science held a three hour hearing on March 12 Hearing on the Management and Operation of the NSFNet by the National Science Foundation. There were two panels, with a total of seven panelists. Each provided written and oral testimony, answered questions and were given an opportunity to offer other comments verbally. Since this hearing is in line with the topic of this list, I am including the entire written testimony (10 typed pages) for your review. I hope that the other panelists will place their testimony here as well. Bill Schrader ---------------------------------------------------------------------- Testimony invited by the U.S. House of Representatives Committee on Science, Space, and Technology Subcommittee on Science Hearing on the Management and Operation of the NSFNet by the National Science Foundation 12 March 1992 By William L. Schrader President and CEO Performance Systems International, Inc. 11800 Sunrise Valley Dr., Suite 1100 Reston, VA 22091 703.620.6651, wls@psi.com Executive Summary: The Internet, the most innovative mass communications infrastructure of this century, was begun by DARPA and is now primarily managed by the NSF. This testimony briefly describes my knowledge of a series of actions, agreements, and events resulting in the inappropriate privatization of the NSFNet backbone. Given the private nature of the agreements, the delay in revealing the agreements, and the apparent conflicts of interest shared by the principals, I am providing this Subcommittee with my understanding of the dates, involved parties, public rationale, actual results, and my conclusions regarding this privatization of the NSF resource. These actions were unfortunate and unnecessary and should be repaired before the government invests even more heavily in attempting to achieve the goals of the NREN. Mr. Boucher, Committee Members and staff: My name is Bill Schrader, and like my carpenter father and his father before him, I am a builder. In 1989, a long time friend, Martin Schoffstall, and I saw an opportunity to build a company to sell computer networking services to colleges and businesses around the world. With our spouse's and children's full support, we borrowed against our houses, withdrew our savings, worked two jobs, raised money from our parents, family and close friends and founded Performance Systems International, Inc. (PSI). We now employ 40 people in offices in New York, Virginia and California, serve 1,500 organizations plus 3,000 individuals in forty states and ten countries. In terms of marketshare, we are about 15% of the world-wide network of networks called the Internet. We are profitable and pay taxes. Prior to PSI, I helped start NYSERNet, one of the first Regional Networks, with Dr. Richard Mandelbaum, Marty Schoffstall and others. I also helped start two supercomputer centers, one at Cornell University with Kenneth Wilson and one at Syracuse University under DARPA. My career for the past 12 years has included many areas covered by the HPCC. We operate our own public data internetwork, called PSINet, consisting of a nation-wide T1 network using leased telephone circuits and PSI owned equipment. PSINet is interconnected with all similar U.S. commercial networks through the Commercial Internet Exchange Association, (CIX) which we helped found to ensure an open and level playing field for this new industry. The CIX is headed by Mitchell Kapor as Chairman of the Board. We connect to the NSFNet backbone for research and academic activities, and to several international networks. Through these many connections, our customers communicate electronically with all of the 7,500 organizations on the global Internet and the 20,000,000 people who read electronic mail regularly. Our services range from unlimited electronic mail for an organization at $25/month (suitable for high schools and small businesses), to high performance full service supercomputer connections at over $50,000 per year. We are technology builders, having led the industry in the design of network management software, now licensed to DEC, SONY and over two hundred other computer and telephone systems manufacturers, resellers, and consumers. We are actively helping many small and large computer software, hardware and service companies begin to offer their products over the network. We are pleased to be competing in this industry and believe it is one area where American companies enjoy a lead and are well positioned to keep it. The information I am providing today covers three topics: 1 NSF Policy Decisions 2 The Economics of the NREN 3 Recommendations for Congressional action 1. NSF Policy Decisions The NSF initiated and operated the NSFNet program during a rapid evolution of the technology and the demand in the marketplace. Below is a brief review of some decisions made by NSF. Decision A NSF Signed Merit/IBM/MCI contract - 1987 % Rationale - Achieve a national backbone network rapidly by leveraging the NSF budget by inducing private contributions of: a) equipment, b) personnel, and c) bandwidth, and encourage the development of high speed networking technology. % Actual Events/Impact - One year was required to bring the network up reliably, because IBM computer equipment was used as routers, and new software had to be written. All the existing Regional Networks and other networks were using commercial grade, off-the-shelf routers available at that time. No IBM routers were ever marketed based on the design used in this project, and thus did not contribute to the nation's commercial position. MCI and IBM provided staff, but all network engineering, operations, and management work was performed by Merit staff (paid by NSF). Discounted bandwidth was paid for by NSF budget. % Conclusion - The IBM equipment contribution was valued at about $100,000 per site, but could have been supplied commercially for $25,000 per site. In contrast to openly bid DARPA R&D programs in advanced technology, this program produced no useful prototypes and was delivered by IBM which was not leading in the field nor using any innovative designs. Except for the R&D staff converting computers to routers, the staff contributions by MCI and IBM were in sales and marketing. NSF essentially funded an IBM R&D project which never produced a product, and was justified on the basis of cost sharing to leverage the NSF budget. The success of the NSFNet and of the industry is not to be confused with the success of any individual contractor or grantee. NSF leveraged their budget by a factor of two by obtaining cost sharing in equipment and staff overvalued and not particularly well suited for the task. It is clear to many that the same amount of NSF budget spent without such leveraging would have produced a better, more reliable network, sooner, and would not have unduly shifted NSF policy to favor a single entity. Decision B NSF requires international connections at backbone switches - 1989 % Rationale - To produce organized international connectivity which is cheaper and better than marketplace decisions made by others. % Actual Events/Impact - For all new connections using any NSF funds, all connections were terminated at Merit/ANS nodes, allowing ANS to own the connection to a foreign market. Free market decisions by non-NSF funded parties were delayed up to 6 months, while NSF insisted on seeking approval from foreign PTTs for these "non NSF connections". Connections from the UK, Germany, Singapore and Korea which involved no NSF funds were denied access to NSFNet for up to six months. According to NSF, this delay was caused by State Department regulations. % Conclusion - NSF has effectively given ANS indirect control over many international connections, while free market connections continue to be delayed. Decision C NSF approves ANS organization, (and privately authorizes ANS's exclusive use of NSFNet for commercial use) - 1990 % Rationale - ANS was formed by the Merit, IBM, MCI team to leverage MCI/IBM equipment, staff, and bandwidth and to position itself for competition in the future. Private ANS meetings with NSF and select Regional Networks, and public comments about a complex shared "infrastructure pool", introduced a packet charging concept (called settlements) which ANS would fund by charging commercial customers attaching through the Regionals. There was no public mention or debate of the exclusive commercial use by ANS, which was the key element of their earlier private agreements with the NSF. Yet, these private agreements between NSF and ANS drove these complex agreements with the Regionals. % Actual Events/Impact - This situation publicly positioned ANS as a not-for-profit, public spirited company willing to share its "profits" with Regional Networks which were willing to sign additional (complex) agreements. NSF approved this subcontracting arrangement without prior public notice, debate, or open bidding. Further, NSF helped provide visibility with press releases quoting Senator Gore and Dr. Wolff. Once approved, ANS took over the NSFNet leadership from Merit in the marketplace by hosting all negotiations and discussions with Regionals, and issuing policy/contract related statements which represented NSF backing. ANS began competing for commercial and non-commercial customers by telling prospective customers that they could "connect directly to the backbone" without using the Regional Networks, and that they should connect to ANS since "at any time, ANS could disconnect PSI or any of the Regionals which had not signed the ANS agreements". In 1991, ANS represented itself as the only network which could guarantee full commercial use of the NSFNet. This was true then and is true now. One example of this is Dialog, a large commercial supplier of electronic information to academic, government and commercial users. It appears that ANS first convinced Dialog that it should connect to NSFNet (ANSNet) for "commercial only" traffic. ANS then attempted to use Dialog to attract the Regional Networks to sign the complex ANS connection agreements, preventing those who did not sign from reaching Dialog. Few Regional Networks signed, and when Dialog discovered that it could access fewer than 5% of the Internet users it converted to a normal ANS customer, and agreed to comply with the NSFNet policy of supplying only research and education traffic. % Conclusion - NSF has thus positioned the ANS/Merit/NSF/IBM/MCI partnership to approach commercial, government, and academic customers with significant advantages no one else can offer, without disclosing this to the public or allowing anyone else to bid. ANS's handling of Dialog's attachment, and subsequent month long delay in disclosing Dialog's request to change, was seen by many as clear positioning for ANS's for profit subsidiary ANS CO&RE to gain marketshare. Decision D Upgrade T1 to T3, and Privatize the NSFNet - 1990 % Rationale - Push networking technology to avoid congestion on the T1 backbone. Leverage NSF funds by allowing some private use. % Actual Events/Impact - NSF negotiated the T3 upgrade arrangement with no apparent technical compliance specifications and no penalty clause for non-compliance. To date, less than half of the T3 nodes are operational beyond test mode, after 15 months of full payments, despite intermittent claims of full operational status by NSF and the contractor. ANS used IBM-provided T3 equipment which was not the same as that used on the T1, had no significant R&D preparation, and failed when deployed. As during the earlier T1 IBM router design, the commercial R&D on T3 routers had been underway for two years by other router vendors using their own limited funds (eg. Proteon, Cisco) and could have been used. The use of IBM computers produced a poor quality network, and damaged these leading commercial suppliers investment in R&D. In November of 1990, ANS's president claimed in a public talk at a Harvard workshop that "in essence, we have privatized the NSFNet". Although few understood, he meant that the NSF was now buying its NSFNet service as a portion of ANS's private network, rather than paying him to operate the NSF's network. While the NSF had sole use of the NSFNet T1, the NSFNet T3 was provided through a "cloud" and could also be used by ANS for their own customers. After the agreements which the NSF had signed creating ANS, and providing it with exclusive commercial access were released in December of 1991, it was clear that ANS's president was correct, the T3 had been privatized. This occurred without public discussion or disclosure, and was effectively hidden for a year. % Conclusion - NSF perceived a need to leverage its budget further, enlarged the scope of the contract from T1 to T3, upgraded the financial size from $4 million to $10 million per year, and privatized the original contract, using private agreements, without additional bidding, and without notice once it was completed. Privatizing a federal facility without notice, and at no cost to the recipient/contractor is improper and should be illegal for any agency of the government. Decision E NSF and other FNC agencies accept apparent conflicts of interest, and unclear boundaries, routinely % FNC/ANS - The Federal Networking Council (FNC) consisting of NREN agencies created an Advisory Council (FNCAC) charged with helping agencies understand how to best spend NREN funds. On the FNCAC are the venture funding directors of ANS (John Armstrong >from IBM, and Richard Liebhaber from MCI), and another member of the ANS board, Mr. Joe Billy Wyatt from Vanderbilt University. No private providers are represented. % ANS Formation - The Chairman of the Board of Merit, Inc., Dr. Douglas Van Houweling, negotiated the multimillion dollar government subcontract with his newly formed subcontractor, ANS, for which he serves as co-founder and Chairman of the Board. When ANS formed its for-profit subsidiary ANS CO&RE, he also became its Chairman of the Board. He remains on the Board of Merit, and as Vice Provost for the University of Michigan, oversees the President of Merit, Inc. He sits here today representing Merit, Inc. to Congress. % NSFNet Operations - For both operational and oversight questions, Merit retains the prime contract for the NSFNet. It subcontracts 100% of it's responsibility to ANS, which operates the T1 network as an NSF facility and the T3 network as ANSNet. ANS then sub-subcontracts some, or all of, the technical and operational aspects of the NSFNet and ANSNet back to Merit. This apparently includes ANS's non-commercial customers as well as its commercial ANS CO&RE customers. ANS staff use computers at Merit, Inc. and the University of Michigan for their work. ANS can subcontract to any new subcontractor without NSF's approval. When something fails to work on this network, finding the person (or organization) who has an "arms length relationship", that is, who will withhold payment for non-compliance, is not possible. % ANS/ANS CO&RE - An ANS salesperson who is selling a commercial or non-commercial connection to ANSNet does not know whether he/she is charging travel and labor hours to ANS or ANS CO&RE until after the sale is made. ANS CO&RE pays tax on profit, which is revenue in excess of cost. However, does ANS CO&RE's cost include the circuits paid for by MCI's donation to ANS (the not-for- profit), space rented by ANS for its offices, ANS CO&RE's prorata share of the depreciation of equipment donated by IBM to ANS, the travel costs by its Chairperson/Board member to testify before Congress when wearing at least three hats? Do funds received from government contracts pay for attorneys, accountants, and public relations firms to keep these involvements straight and attempt to present the correct image to the government and the marketplace? % Standards Process - ANS hired a number of well positioned people in the industry, including the head of the Internet's public open technical group, the Internet Engineering Task Force, as one of nine vice presidents. Dr. Phillip Gross continues to hold his position in IETF. This provides ANS (and ANS CO&RE) with advanced knowledge of industry technical developments as well as some influence in guiding the timing and structure of emerging standards. % Conclusion - Clear or apparent conflicts of interest situations occur in many aspects of ANS, Merit, NSF, IBM, and MCI. The entire group, which is now publicly self-characterized as a "partnership" has total control over the $50 million NSFNet backbone contract, now privatized, and well positioned to win any NREN contracts in the future. In my opinion, there is little question that the actions of the ANS, Merit, NSF, IBM, MCI "partnership" have: 1) interfered with international and interstate commerce; 2) used the structure of the arrangements to influence NSF to make extraordinary decisions, privately; 3) planned their actions together, in advance; and 4) provided unfair advantage to ANS for the duration of the contract and beyond. These arrangements do not provide a distinct separation between the role of the NSF oversight and the operation of the network by private parties. The lines between the government and the contractor (grantee) have confused the regulators with the regulated. Decision F NSF extends the Merit contract for 18 months, announces a rebid - 1991 % Rationale - More time is needed to plan the follow-on. NSF must provide the backbone because the Regional Networks do not want to take responsibility for buying their own connections, even with NSF funds. There is concern that two backbones cannot operate together smoothly with today's technology. % Actual Events/Impact - Extending the current contract provides up to $15 million more to ANS without competitive bidding, assuring ANS and Merit a steady revenue stream for 28 more months. ANS maintains its exclusive rights to sell direct backbone connections and guaranteed commercial use of NSFNet to all commercial and non commercial customers. When, and if, a follow-on contract is signed and implemented, the NSF rebid plan calls for the same bandwidth, and a reduction of payments for each successful bidder to $3 million (down from $10 million). % Conclusion - It seems that the NSF will save at least $4 million per year ($10 million current cost, $6 million for two suppliers after the rebid) if the rebid is completed before the Merit/ANS contract expires in November 1992. ANS continues to establish its own policies, representing them as NSF policy, such as settlements and infrastructure pools described earlier. There is a danger that this appears to be NSF policy which it is not. In my opinion, NSF hired a contractor and then allowed that contractor to unfairly influence its policy, funding, and technical decisions through the period of the contract and beyond. Many decisions were made, but some of the more serious policy and contracting decisions failed to recognize the larger market, were made privately without open discussion, did not allow full participation, and did not follow proper contracting procedures. 2 The Economics of the NREN Sound economic principles may have been considered in the design of the NREN legislation. However, the market has evolved more rapidly than planned. Basic market forces have shifted the geographic, economic, technological and political realities of the NREN to their natural equilibrium of ubiquitous networking. Therefore, rather than investing NREN's hundreds of millions of dollars in the technology, it is better to invest in connecting people to the network, making it easier to use, and teaching them how to participate. NREN is still built on the principle of "putting money into the backbone". It has been proven by the ANS contract that doing so produces little innovation and results in no self-sufficiency, since organizations will not pay for a free good. The correct method for government involvement in a burgeoning industry like this is to subsidize individual target organizations: colleges and public libraries at first to ensure public access, then high schools and elementary schools to ensure access for children. A program using "Yellow Stamps" was first proposed in 1989 by the National Research Council's report "Toward a National Research Network", and again in 1991 by Dr. Wolff. In this program, NSF would directly fund these organizations' networking projects, and could be traded in with any network supplier which had met the criteria established by NSF. This program would require serious work to handle thousands of organizations, but would create the most stable results and still allow the national network to reach gigabit speeds in the same time frame. The difference is whether the NSF feeds the market at the bottom, or attempts to lead the market with advanced technology. Since the market has been reliably ahead of the NSF's technical program for the entire five year term of the current contract, there is sufficient reason to believe NSF cannot lead the market and, therefore, should feed it from demand. If this program is created, I believe in five years we will see: % local control and interest, where users come to value the network % local leveraging of funds, 100 to 1 as seen in the university sector % local control to buy from provider(s) of choice, on local schedule % opportunity for all competitors to offer services, build economies of scale on their own initiative. % a natural, permanent aggregation of traffic from hundreds of thousands of small and large NREN target organizations % a natural convergence of commercial and NREN traffic on the highest speed "gigabit highways", taking advantage of fiber economies of scale % no single monopoly, but instead a working, integrated commercially built operation provided with the service distinctions required by each market segment % a smooth way for government funds to be ramped down after the five year program, for those organizations which participated in the early days % direct political recognition of NREN's value in each local area, by parents, teachers, local political leaders and taxpayers. 3 Recommendations for Congressional action Overall: Direct the NSF to A) remove the unfair advantages ANS has acquired, B) to consider industry and economic trends in future policy decisions, and C) to cease signing large contracts without administrative due process. Specific Actions: % Terminate the ANS contract on schedule. Congress should direct NSF to force ANS to either: A) remain under contract for NSFNet funds for the duration of the contract and extension, providing a service only to NSF, or B) sell commercial and academic access connections on the open market, but terminate the NSFNet contract on schedule in 1992. ANS cannot be allowed to continue both the contract and the private use, since it will then be able to continue to use its government subsidized backbone network to offer its service to commercial and academic customers at prices which do not reflect actual cost. % Open Commercial Access. Congress should direct NSF to either remove ANS's right to sell commercial access to NSFNet funded gateways or to allow other commercial carriers equal access at no cost. % Make NSF the NREN lead agency. NSF is best equipped to handle thousands of small proposals involving colleges, K-12, libraries, and similar NREN activities. % Require accountability for NREN funds by NSF. NREN funds have few controls now, being split over four agencies. There is no direction or control in the implementation strategies among the four agencies, each of whom can spend it any way they wish. At the very least, agencies should be directed to report how the money was spent. % Request that Dr. Wolff develop his "Yellow Stamp" program to distribute funds directly to institutions. America's libraries, colleges, secondary, and elementary educational institutions and non-profit research centers would then be able to use the funds solely to purchase internetworking services. % Maintain "research funding" but do not confuse operational network access with network research. It is clear that the commercial marketplace is properly motivated to invest its own resources without government R&D subsidy. The government should limit its R&D scope to innovative industrial/academic research in high speed communications technologies and applications at the frontier.