Editors : Scott Bradner • sob@har vard .edukc c laf fy • kc@caida .org
kc claffyand Sascha D. MeinrathCooperative Association for InternetData Analysis
Scott O. BradnerHarvard University
The (un)Economic Internet?
The Internet Economics track will address how economic and policy issues relate
to the emergence of the Internet as critical infrastructure. Here, the authors
provide a historical overview of internetworking, identifying key transitions that
have contributed to the Internet’s development and penetration. Its core
architecture wasn’t designed to serve as critical communications infrastructure
for society; rather, the infrastructure developed far beyond the expectations of
the original funding agencies, architects, developers, and early users. The
incongruence between the Internet’s underlying architecture and society’s current
use and expectations of it means we can no longer study Internet technology in
isolation from the political and economic context in which it is deployed.
T his article kicks off IC’s new serieson policy, regulatory, and business-model issues relating to the Internet
and its economic viability. These articleswill explore a range of topics shapingboth today’s Internet and the discourse inlegislatures and deliberative bodies at thelocal, state, national, and internationallevels in pursuit of enlightened steward-ship of the Internet in the future.
Mindful of Internet connectivity’sfundamental import for advanced aswell as emerging economies and its day-to-day irrelevance for the unconnectedvast majority of human beings, piecesfor this series will cover technology aswell as political, economic, social, and
historical issues relevant to IC’s interna-tional readership. In this inaugural arti-cle, we provide a historical overview ofinternetworking and identify topics thatneed further exploration — topics we par-ticularly encourage authors to cover infuture articles in this series.
A History of Internet(un)EconomicsThe modern Internet began as a relative-ly restricted US government-fundedresearch network. One of the most revo-lutionary incarnations of this network,the early ARPANET, was limited inscope — at its peak, it provided dataconnectivity for roughly 100 universi-
ties and government research sites. In thedecades since, a few key transitions have beencritical in radically transforming this communi-cations medium. One of the most important ofthese critical junctures occurred in 1983, whenthe ARPANET switched from the Network Con-trol Program (NCP) to the (now ubiquitous)Transmission Control Protocol and Internet Pro-tocol (TCP/IP). This switch helped change theARPANET’s basic architectural concept from asingle specialized infrastructure built and oper-ated by a single organization to the “network ofnetworks” we know today. Dave Clark discussesthis architectural shift in his 1988 ComputerCommunications Review paper, “The Design Phi-losophy of the DARPA Internet Protocols.”1 Hewrote that the top-level goal for TCP/IP was “todevelop an effective technique for multiplexedutilization of existing interconnected networks.”
During this same period, network developerschose to support data connectivity across multi-ple diverse networks using gateways (now calledrouters) as the network-interconnection points.Preceding communications networks, such as thetelephone system, used circuit switching, allocat-ing an exclusive path or circuit with a predefinedcapacity across the network for the duration of itsuse, regardless of whether it efficiently used thecircuit capacity. Breaking with traditional circuit-switching network design, early internetworkingfocused on packet switching as the core transportmechanism, facilitating far more economically aswell as technically efficient multiplexing of exist-ing networking resources. In packet-switchingnetworks, nonexclusive access to circuits is nor-mative (although companies still sometimes buydedicated lines to run the packet traffic over);thus, no specific capacity is granted for specificapplications or users. Instead, data is commingledwith packet delivery occurring on a “best effort”basis. Each carrier is expected to do its best toensure that packets get delivered to their desig-nated recipients, but no guarantee exists that aparticular user will be able to achieve any partic-ular end-to-end capacity. In packet-switching net-works, capacity is more probability-based thanstatically guaranteed. Internet data transport’sbest-effort nature has caused growing tension inregulatory and traditional telephony circles. Like-wise, as the Internet becomes an increasingly crit-ical communications infrastructure for business,education, democratic discourse, and civil socie-
ty in general, the need to systematically analyzecore functionality and potential problem areasbecomes progressively more important.
Early developers couldn’t have foreseen thelevel to which the Internet and private networksusing Internet technologies have displaced othertelecommunications infrastructures. It wasn’t untilthe mid 1990s that visionaries such as Hans-Werner Braun started warning protocol develop-ers that they needed to view the future Internet asa global telecommunications system that wouldsupport essentially all computer-mediated commu-nications. This view was eerily prescient, yet coreInternet protocols haven’t evolved to meet increas-ing demands and are essentially the same as theywere in the late 1980s.
A growing number of researchers are convincedthat without significant improvements andupgrades, the Internet might be facing serious chal-lenges that could undermine its future viability.Features such as network-based security, detailedaccounting, and reliable quality-of-service (QoS)control mechanisms are all under exploration tohelp alleviate perceived problems. In response tothese concerns, the International Telecommunica-tion Union-Telecommunication StandardizationSector (ITU-T) Next Generation Networks studygroup (NGN; www.itu.int/ITU-T/ngn/) is workingto define a very different set of protocols thatwould include these and other features.
Security: Not the Network’s JobVarious people have offered explanations regard-ing the lack of security protocols in the Internet’sinitial design. Clark’s seminal paper doesn’t men-tion security, nor does the protocol specificationfor IP.2 Because the network itself doesn’t con-tain security support, the onus has fallen to thosewho manage individual computers connected tothe Internet, to network operators to protectInternet-connected hosts and servers, and to ISPoperators to protect their routers and other in-frastructure services. Services such as user orend-system authentication, data-integrity veri-fication, and encryption weren’t built into thecore Internet protocols, so they’re now layeredon an infrastructure that isn’t intrinsicallysecure. Currently, few existing studies examinethe potential economic rationale for this currentand continuing state of affairs and the ramifica-tions for the infrastructure’s efficiency, perform-ance, and sustainability.
54 www.computer.org/internet/ IEEE INTERNET COMPUTING
Internet Economics Track
QoS: Too Easy to Go WithoutThe original IP packet header included a type ofservice field to be used as “an indication of theabstract parameters of the quality of service de-sired.”2 This field, later updated by DifferentiatedServices,3 can define priority or special handling ofsome traffic in some enterprise networks and with-in some ISP networks, but it’s never seen signifi-cant deployment as a way to provide QoS acrossthe public Internet. Thus, the QoS a user gets fromthe Internet is typically the result of ISP design andprovisioning decisions rather than any differentialhandling of different traffic types. Thus far, “throw-ing bandwidth at the problem” has proven to be afar more cost-effective method for achieving goodquality than introducing QoS controls.4
Yet, what happens if conditions change so thatoverprovisioning is no longer a panacea? The day-to-day quality most users experience from theirbroadband Internet service is good enough, forexample, to enable voice-over-IP (VoIP) servicessuch as Skype and Vonage, which compete favor-ably with plain old telephone services. However,the projected explosive growth of video and otherhigh-bandwidth applications might increase con-gestion on parts of the current infrastructure to thepoint that special QoS mechanisms could berequired to maintain usable performance of eventhe most basic services.
Accounting: A Missing GoalIn their first paper on TCP/IP, Vint Cerf and RobertKahn felt that accounting would be required toenable proper payments to Internet transportproviders.5 More than a decade later, Clark echoedthis requirement in his Design Philosophy paper.In his listing of second-level goals affecting theTCP/IP protocol suite’s design, the seventh andfinal goal was that “the resources used in the Inter-net architecture must be accountable.”1 As withsecurity, however, no evidence exists that account-ing was ever an operational goal for DARPA indeveloping and running the ARPANET, nor is thereany indication that accounting was a goal for theUS National Science Foundation (NSF) in the fol-low-on NSFnet. Indeed, if a government agency ispaying in bulk for the entire system, accountingitself is a technical as well as economic inefficien-cy. Consequently, today’s Internet has no built-inaccounting mechanisms, making it fundamental-ly different from previous circuit-switched net-works and creating substantial debate as to how to
fairly meter and charge for broadband infrastruc-ture and usage.
The End-to-End Model’s ImpactThe Internet’s architecture and initial deploymentused an end-to-end (e2e) model of connectivity.Jerome Saltzer, David P. Reed, and Clark first dis-cussed elements of this model in their 1981 paper,“End to End Arguments in System Design.”6 Thegeneral rationale behind the e2e model is that thenetwork doesn’t have to know the applicationsrunning on it because it’s simply a neutral trans-port medium. This neutral traffic handling hasenabled explosive innovation in edge services andapplications over the past several decades. Forexample, an application developer doesn’t need toget permission from ISPs, or pay them anythingother than normal service fees, to deploy a newapplication. Likewise, network operators don’tknow what applications are running on their net-works, nor can they participate in the value chainfor these applications.
Clark once said in an Internet Research TaskForce (IRTF) presentation that the Internet “did notknow how to route money.” He held that there wasno efficient way for an independent service providerto cost/profit share with an ISP so that the ISP wouldprovide better service to users who weren’t directcustomers. The Internet economic model has alwaysbeen “sender keeps all” — an ISP serving a particularcustomer keeps all the revenue from that customerwithout regard to where his or her traffic is going.In many countries, no regulations covering peeringrelationships among providers exist, thus leavingISPs on their own to decide whether to peer. Typi-cally, especially in the commercial sector, these deci-sions are based primarily on immediate businessinterests with little public policy input.
Telephone RegulationMany parts of the world have well-developed tele-phone networks. However, this robustness oftencomes at a cost to the networks’ users. Regulationsrequiring that telephone carriers ensure the relia-bility and price controls that they themselvesdemand in order to guarantee a rate of return ontheir investment boost service prices. A less regu-lated and price-controlled future for telephone car-riers seems inevitable. It remains to be seen if theywill be as willing to put significant resources intoreliable infrastructures and the personnel neededto run them if competition sets the prices rather
MAY • JUNE 2007 55
The (un)Economic Internet?
than regulation. Likewise, the intersections amongregulatory structures, pricing, service quality, andinterconnectivity with other data communicationsservices are still wide open for exploration.
Internet (non)RegulationAlthough open-access regulations on PSTN trunkswere essential to the development of the Internet,regulation of Internet service itself has remainedlargely laissez-faire. For example, until recently, USISPs didn’t have to register with the governmentbefore offering services, and governments typical-ly haven’t regulated either ISPs’ service offeringsor their service quality. Yet, government attitudestoward the Internet are beginning to change. Thefirst major US regulation covering ISPs — theCommunications Aid to Law Enforcement Act(CALEA) — goes into effect in May 2007 and
requires ISPs to register with the government andbe able to track users. Already, numerous regula-tors have begun investigating the viability of man-dating that ISPs install QoS mechanisms becausethey believe that this is required to ensure that theInternet can reliably help emergency workersrespond to natural or man-made disasters. Unlessthe network research community fundamentallychanges our approach, future regulations will beconsidered, ratified, and implemented with littlepeer-reviewed empirical research documentingtheir likely technical and economic effects.
Internet MeasurementBecause no systemic measurement activities existfor collecting rigorous empirical Internet data, inmany ways, we don’t really know what the Inter-net actually is. Thus, we don’t know the totalamounts and patterns of data traffic, the Internet’sgrowth rate, the extent and locations of conges-tion, patterns and distribution of ISP interconnec-tivity, and many other things that are critical ifwe’re to understand what actually works in the
Internet. These data are hidden because ISPs con-sider such information proprietary and worry thatcompetitors could use it to steal customers or oth-erwise harm their business. The information mightnot even be collected at all because no economicincentive exists to do so, nor do any regulationsrequire its collection.
The Changing ISP CommunityThe original Internet was provided for “free” bygovernments and government-supported researchinstitutes. In the US, direct federal governmentsupport for the backbone and attached regionalnetworks ended in the mid 1990s, although taxincentives continued to promote private as well aspublic infrastructure development. However, com-plete private ownership of the entire US Internetinfrastructure hasn’t yet occurred. Today, manystates and consortiums continue to run their ownnetworks, usually restricting who can use them insome way — most often to educational andresearch constituencies — and connecting theseresearch networks into the global Internet.
Historically, most telephone carriers weren’tinterested in offering Internet service to individ-ual homes or to the business community. Evenwhen a telephone carrier did offer such services, itwas usually through a separate division that com-pany management often viewed as outside itsbasic mission. Instead, commercial ISPs often pro-vided Internet service by leasing telephone carrierfacilities or by setting up dial-up modem banks tointerconnect with the plain old telephone system.
After Internet infrastructure commercializationbegan, the Internet service provision businessmodel was predicated on making a profit bycharging customers more than it cost an ISP to runthe service. This business model is problematicgiven that Internet connectivity is a commodityservice, with most customers caring more aboutlow prices than claims of better quality oradvanced services. Thus, competition, along withundefined accounting mechanisms for the newtechnology, drove prices below sustainable levelsfor most ISPs. The resulting massive provider con-solidation is still in play, but customers are nomore willing to pay high prices for Internet serv-ice in the new environment. A survey quoted in a2002 US Federal Communications Commission(FCC) report determined that only 12 percent ofcustomers would be willing to spend US$40 permonth for broadband Internet service.7
56 www.computer.org/internet/ IEEE INTERNET COMPUTING
Internet Economics Track
Today’s Internet has no built-inaccounting mechanisms, making itfundamentally different fromprevious circuit-switched networks.
Meanwhile, telephone carriers began to offerbroadband Internet service directly over their ownfacilities, particularly in higher-income, urban res-idential markets, directly competing with commer-cial ISPs who had been offering service viaoverlays on the telephone carriers’ facilities. Par-alleling telephone carriers’ entry into the broad-band market, cable TV companies also beganproviding broadband Internet service over theirown facilities. Today, most residential customersget Internet access service from telephone carriersor cable TV companies, in which the Internet busi-ness is only part of their service offerings.Although standardized, “cookie cutter” servicepackages have hampered what customers can dowith their network services, no one has yet stud-ied how the shift of broadband service provisionfrom ISPs to phone and cable TV companiesimpacts Internet service quality and dynamism.
The (un)Economic InternetAll these factors form the background to the cur-rent debates on the Internet’s future, often lumpedunder the heading of “network neutrality” — a dis-cussion with far wider and deeper implicationsthan that label conveys. The key question at theroot of the debate is whether viable economicmodels exist for Internet service provision, giventhe high cost of deploying physical infrastructureand operating the network, coupled with ISPs’current inability to participate in the much moreprofitable application value chain. Further compli-cating analyses of these factors are the internallyconflicted regulatory agencies, tasked with ensur-ing both that the general public’s best interests arekept foremost and that the “free market” beallowed to innovate and police itself.
Many first-generation ISPs went out of busi-ness because they couldn’t find a successful busi-ness model given constraints from both theIndependent Local Exchange Carriers (ILECs) andtheir own customer base. The current generationof telephone-carrier-based ISPs is asking regula-tors for the ability to charge differentially basedon the applications used and content consumed.These companies claim that they won’t be able toafford to deploy the necessary infrastructureupgrades without this type of discriminatory pric-ing. Their opponents worry that letting ISPs decidewhich applications can use their facilities and atwhat cost would destroy the very environment thatenabled the creation of today’s Internet.
Meanwhile, a growing number of communitieshave decided that they aren’t well served by exist-ing ISPs (generally meaning the telephone carri-ers) and have decided to build their own Internetinfrastructures. This is similar to what the academ-ic community undertook immediately after NSFretired its (NSFnet) backbone, and to what manystate education networks — such as California’sCorporation for Education Network Initiatives(CENIC), Florida’s Lambda Rail, and New Mexico’sLambdaRail — are doing. There is a growing,though far from universal, view that basic Inter-net connectivity is a fundamental civil societyrequirement (much like roads, schools, and so on)and that governments should thus ensure univer-sal access to this valuable resource.
Another scenario that will deeply alter the eco-nomics is commercial ISPs’ leasing of government-funded infrastructure. These public–privatepartnerships are currently being developed inthousands of communities around the globe.Objective empirical analyses of the various busi-ness models for providing Internet infrastructureaccess, including empirical validation of inputs,outputs, and interacting technological factors, isone of the least understood yet vital aspects of thisemerging critical infrastructure.
T he Internet Economics track in this magazinewill focus on the ongoing debates surrounding
issues of economics and policy, and how they'reinfluenced by, and should influence, science andengineering research. We are heading into anoth-er decade of tremendous innovations, not only inwireless connectivity and high-bandwidth appli-cations and services that use it but in the businessmodels that will lead to their success or failure.Gaining a better understanding of the tussles(known outside our field as “economics and poli-tics”) among providers, users, and regulators ofInternet access, services, and applications will helpensure enlightened progress on security, scalabili-ty, sustainability, and stewardship of the globalInternet in the 21st century and beyond.
References
1. D. Clark, “The Design Philosophy of the DARPA Internet
Protocols,” Proc. Symp. Comm. Architectures and Proto-
cols, 1988, pp. 106–114.
2. J. Postel, Internet Protocol, RFC 791, Sept. 1981; www.ietf.
org/rfc/rfc791.txt.
MAY • JUNE 2007 57
The (un)Economic Internet?
3. T. Li and Y. Rekhter, “A Provider Architecture for Differen-
tiated Services and Traffic Engineering (PASTE),” IETF RFC
2430, Oct. 1998; www.ietf.org/rfc/rfc2430.txt.
4. P.C. Fishburn and A. Odlyzko, “The Economics of the Inter-
net: Utility, Utilization, Pricing, and Quality of Service,”
Proc. 1st Int’l Conf. Information and Computation
Economies (ICE 98), ACM Press, 1998, pp. 128–139.
5. V. Cerf and R. Kahn, “A Protocol for Packet Network Inter-
