Newsgroups: alt.amateur-comp From: ronda@ais.org (Ronda Hauben) Subject: Summary of Cook Report on Commercialization of the Net Message-ID: Date: Fri, 08 Jan 93 21:18:34 EST Organization: UMCC Lines: 293 [ Article crossposted from umich.interesting.people ] [ Author was Bernard.A.Galler@um.cc.umich.edu ] [ Posted on 1 Jan 1993 12:43:34 -0500 ] >Posted-Date: Thu, 31 Dec 92 19:28:34 EST >Date: Thu, 31 Dec 92 19:28:34 EST >From: cook@tmn.com (Gordon Cook) >To: psi!com-priv@uunet.uu.net >Subject: COOK Report v. 1 no 10 & 11: Summary History of NSFnet Privatizat A special 35,000 word issue of the COOK Report on Internet -> NREN based on documents obtained from the National Science Foundation under the Freedom of Information Act will be published tomorrow. The issue also contains extensive comments by NSFnet Director Steve Wolff (circa 2,000 words) and many other reviewers. The EXECUTIVE SUMMARY from COOK Report Vol 1 Issues 10 & 11 FOLLOWS: (2,800 words) NSFnet "Privatization" and the Public Interest: Can Misguided Policy Be Corrected? An Historical Review of Network Policy Making - 1989 to the Present With Pointers for Corrective Action Executive Summary: Part One: MERIT, IBM and MCI Present ANS as an Offer the NSF Cant Refuse With planning for an NREN going forward in Congress and competition there between the Department of Energy and the National Science Foundation for lead agency, the NSF had ambitious plans for continued growth of the NSFnet backbone. Unfortunately for it had little money with which to fund an upgrade to T-3 speed. At some point in early 1990 with IBM in the lead, MERIT came up with a plan to create ANS as a non profit operator for a new high speed privatized commercialized backbone. The NSF was asked to accept the privatization of the backbone by means of a realignment of the cooperative agreement with MERIT. Short of terminating the cooperative agreement and the politically unthinkable course of immediately putting the backbone up for rebid, it had no other recourse but to accept the terms offered. The design was immediately and grievously flawed for not only was there no plan for the privatization; there were no criteria by which to measure its success or failure. Furthermore the design held unacknowledged economic implications -- for what was being privatized was the only part of the network that had no customers. Part Two: Fallout of Privatization without a Plan: Contradictions Behind ANS Yield a Zero Sum Solution Very quickly after ANS's creation it became apparent that ANS had one major aim in mind. To raise the speed of the backbone to multi hundred megabit speeds within two years. As ANS would explain to NSF, despite the fact that its sponsors had committed very large sums to bring their networks to SONET capability sooner than otherwise planned and produce gigabit switches as soon as possible on behalf of American competitiveness, ANS would nevertheless need large additional sums of money. It would obtain this income largely from the one part of the American Internet that did have customers, the mid-level networks. Although ANS listed 9 sources of revenue, it would obtain the money mostly in two ways. First it would seek to detach from the mid-levels institutions that it thought it could better serve by direct attachments to its privatized backbone. Secondly, now that it was about to be granted the right to sell openly commercial use of the T-3 backbone it was building, it would develop plans to charge the mid-levels attachment fees to that backbone, as well as additional fees for sending commercial traffic across it. ANS had spent much of the fall working with representatives from some mid-levels to determine the new criteria by which it expected to run the network if the NSF did not choose to rebid the backbone in 1992. Although its plans for "Capital Infusion into the Backbone via Mid-level Network Fees and Settlements for Commercial Traffic," did not explicitly say so, if the mid- levels had to raise costs to their end users in order to pay the fees ANS imposed, so much the better because such higher costs would narrow the difference between what it charged for its direct connections and the generally less expensive fees charged by the mid-level networks. What was not clear was the extent to which the NSF would be bound to bail mid-levels out should ANS be successful in taking away sizable numbers of their customers. Under these zero sum conditions, a win for ANS was a loss for the mid-levels. In allowing ANS to sell direct access its own network (ANSnet) that used the same physical facilities as NSFnet, the NSF spoke of two virtual and presumably distinct networks. It properly insisted that commercial traffic placed by ANS on its network not diminish bandwidth needed by its own customers. However ANS's January 15 1991 Proposal to the NSF made clear that once dumped into the network packets from its commercial customers would be indistinguishable from those of the NSF's customers. Furthermore if the network became congested it would be almost impossible to ascertain whether the congestion resulted from academic or commercial use. In ANS's own words: "there is no agreed to model for measuring congestion within an IP based network At present, even the model of congestion determination described in the NSFnet solicitation of 1987 is not currently measurable by any tools that have been developed to date nor does there seem to be any work in the open literature suggesting that this is feasible." ANS went on to say how it would estimate congestion. One critic of the document said that it seemed to him that just as long as the counts of packets carried by the network kept going up ANS would be able to say that it had satisfied its part of the bargain. The NSF was buying services from a new entity with an agenda more complex that just one of helping research and education. The NSF had assigned most of the control over the network to an entity that appeared determined to sell speed in the hope that the Fortune 1000 would climb aboard, and economies of scale would be achieved, which, it was said, would benefit the research and education sector. Part Three: Competing Designs for Commercialization ANS CO+RE versus CIX In early 1991 while ANS prepared to charge mid-level networks connection and settlements fees, two of its commercial competitors and one mid-level formed the Commercial Internet Exchange and pledged to interconnect freely, and without settlements charges, all networks who wished to join and pay an annual $10,000 membership fee. On May 24 1991 the NSF gave its blessing to a profit making spin off of ANS that would sell explicitly commercial use of the network backbone. While this was a privilege denied its commercial competitors, it was generally justified by assertions that ANS's sponsors, IBM and MCI, had spent far more on providing the network than they had received from the US government. ($60 million total expenditure compared to $18 million in Federal funds was an estimate given by Steve Wolff the NSFnet Director in July of 1991.) However a degree of doubt was cast upon this assertion in late 1992 when ANS's first Form 990 became available and showed expenditures for the backbone that roughly equaled income from Federal grants. In his review of the Editor's study, NSFnet Director Steve Wolff pointed out that in asserting that he could find no evidence of cost sharing, the editor had ignored contributions in kind from MCI and IBM as well as ANS's "inheritance" from its sponsors. However a phone call to the IRS resulted in the statement that a properly and completely filled out Form 990 must list both contributions in kind and any "inheritance," used to perform the entity's program service activity. While ANS made it clear that it was now interested in offering turnkey outsourcing for Fortune 1000 T-1 networks, critics suggested that its marketing task would be a difficult one because of security concerns raised by offering services based on Layer 3 sharing in an environment accessible to hackers. By late 1992 it became apparent that offering services from a Layer 2 based backbone would be much more desirable from a corporate security point of view. Such a positioning would allow the commercial service provider to sell both a Virtual Private Network (VPN), and access to the Internet -- the best of both worlds. In the summer of 1992 PSI would move to a frame relay based backbone and gain VPN capability, a move that would apparently be denied to ANS as long as it remained dependent on IBM routers. ANS's strategy to expand dramatically the size of its network through corporate sales that would feed money into an infrastructure pool that would in turn benefit its research and education customers seemed to be facing some major handicaps. Meanwhile ANS had been trying to get many mid-level customers to buy direct connections to it. Not surprisingly this alienated to a greater or lesser degree most of the mid-levels. In August of 1991, faced with the not easily understood cost impacts of the ANS Plan for Commercial Services, the mid-levels at a meeting of FARnet indicated that they wanted the NSF to stay in the backbone provisioning business in order to avoid what they feared would be tremendous fiscal uncertainty if ANS were left as the sole backbone proprietor when the cooperative agreement expired in November of 1992. At the end of 1991 the impact of ANS's commercial backbone policy became clearer when Dialog was announced as ANS's first commercial customer and those mid-levels that had not yet signed the ANS connectivity agreement were blocked from connection. This made the impact of ANS's exclusive right to move commercial traffic across the former NSFnet backbone apparent to all. The resulting angry dispute made the New York Times and in January of 1992 Dialog was allowed to become an R&E customer of ANS so that it could be reached by the entire network. Part Four: Executive Summary Ignoring CIX and Courting Mid-Levels: St Louis in April 92 and the Infrastructure Pool In 1992, faced with pressure from the network community to give up its insistence on settlements and join the CIX, ANS agreed to connect to the CIX so that networks which were already CIX members could use its backbone to send data to the CIX. It also decided to try to repair its relations with the mid- levels at a two day meeting in mid April attended by representatives from 20 mid-levels. There ANS made an honest effort to deal with mid-level mistrust head on. Glenn Ricart remarked that "ANS is unsure if it is in the wholesale or retail business -- being in both businesses simultaneously is confusing." As an unfortunate result "we are all fighting over the 2% of the customers already connected rather than the 98% that are unconnected." An unresolved issue, he concluded was "what is ANS going to be good at." By meetings end ANS had promised "to clearly state its mission and goal (and then act accordingly) to help dispel the fog." By November 1992 ANS CO+RE had been in existence almost a year an a half. While ANS had said that about half the mid- levels had signed connectivity or gateway agreements, it had never released their identities. These mid-levels would be eligible to share in funds from ANS commercial usage fees gathered in an Infrastructure Pool. In an interview with Communications Week, Al Weis had almost a year earlier estimated that the pool would be worth $250,000 by the end of 1992. Questions were asked about the status of the Pool and the Resource Allocation Committee (RAC) designated to administer it. The RAC was summoned for the first time via a telephone conference call on October 23, 1992. Maintaining that the Pool and RAC were part of the profit making side of its operations, ANS took a position that it need not answer such questions. Eventually however information about the RAC leaked out via a network discussion list and a December 4th article in Communications Daily. The amount of money in the pool ($62,000) also leaked. ANS had been asked by the attendees at the April 1992 St Louis meeting to hold a follow up meeting as soon as possible. None has apparently so far been held. (Some involved have said that they considered the RAC conference call to have served the function of the follow up meeting.) Nevertheless from communications with some of the mid-level directors who are members of the RAC, it is clear that at the end of 1992 ANS has managed to win some friends among the mid-levels. However some mid-levels seem to have a foot in both the ANS and CIX camps. And still others seemed determined to either join the CIX or try to remain independent of both camps. Part Five: Who's in Control? Measuring Network Usage and Contribution to Infrastructure From FOIA evidence released by the NSF it seems that the NSF has had difficulty in ensuring responsiveness from MERIT and ANS. In both September 1990 and May of 1991 the NSF had asserted that it would develop with MERIT a technical means of verification that "the attachment and service sponsored by the NSF under MERIT's Cooperative Agreement with the NSF is not diminished." In early November 1990 the author publicly asked the NSF to disclose the means of verification agreed upon. The NSF was silent. Finally in November 24, 1992 it received from MERIT a four page letter containing the compliance plan. As with the first plan proposed by ANS in January 1991, it made it very plain that, once in the network, NSF and ANS traffic were indistinguishable. It proposed that some benchmark measurements be used to determine when links were congested and needed upgrading. However in any given instance it would be nearly impossible to tell whether network congestion would be due to supercomputer center traffic or traffic from Perot Data Systems, a large ANS commercial customer, to use an arbitrary example. Finally when in November 1992 the author asked what infrastructure ANS had contributed to the network that had not been paid for by the National Science Foundation, an ANS vice president sent Steve Wolff a lengthy listing of ANS contributions. However, in the opinion of the author they were almost entirely contributions that were absolutely necessary to deliver the minimally acceptable backbone service for which the NSF had been paying in the first place. Part Six: Where Do We Go from Here? It has been too easy to confuse the network goal of technology development and transfer with that of access for all Americans to what should become the most powerful infrastructure of the information age. The current arrangement places the network in the hands of ANS, an entity that has shown itself to be far more interested in technology development and transfer than in the cost of access for the ordinary user. With its recently announced intention of support a scaled down backbone that connects on the four NSF supercomputer centers, the NSF appears to be heading in the right direction. However, given the contradictions that ANS faces between its profit making, commercial, proprietary side and its non profit research and education side, it seems unlikely that ANS would be able to use further stewardship of the backbone in such a way that it would unambiguously benefit the research and education community that the NSF is intended to serve. Therefore the NSF would be well advised to evaluate with great caution any ANS responses to its backbone solicitation. The NSF should also take a very wary approach to cost sharing suggestions and should not under any circumstances permit the successful bidder to resell commercially AUP free backbone bandwidth. It is questionable whether the US government should be involved in the building of a production backbone. What the new administration must make clear is that it will do whatever is necessary to ensure affordable access to the network. It must also help to educate Congress concerning the policy implications behind the network by supporting Senator Breaux's call for a Presidential Commission on Telecommunications Policy. --------- PS. I may in a few days make available by email request 20 kilobytes or so of directed excerpts from the FOIA materials. Directed? I meant to type direct excerpts.