|< Day Day Up >|| |
By mid-1990 it had become evident that the Internet was outgrowing its research and education roots. TCP/IP was becoming the long-awaited open platform for global data networking. Regional networks and the NSFNET backbone operator were beginning to eye a potential commercial market for Internet access. By 1990 some of the mid-level networks had formed commercial Internet service providers (ISPs) to operate alongside their subsidized regional networks. [1 ]
In May 1991, after much public and private discussion of commercialization and privatization, the National Science Foundation (NSF) permitted commercial traffic to cross the NSFNET, provided that certain stipulations regarding cost recovery, surplus revenues, and quality of service were followed. [2 ]But the commingling of commercial and noncommercial traffic and providers posed serious policy dilemmas for NSF.
In 1993, NSF responded to the problem by moving to an entirely new architecture for the Internet. [3 ]The agency withdrew from backbone support altogether and tried to facilitate a commercial marketplace for Internet access composed of multiple, competing backbone providers. Commercial ISPs would be interconnected at five NSF-designated and partially supported network access points (NAPs). [4 ]The new architecture went into effect in 1995, and NSFNET was decommissioned in April of that year. The transition was so successful that no one noticed; indeed, within a few years of the changeover the NAPs' role as interconnection points had been minimized because most ISPs began to rely on private contracting with backbone and transit providers for most of their interconnection.
The supply of Internet connectivity by a commercial market was a major stimulus to the demand for domain names. For the new ISPs, domain name registration and IP address assignment became required inputs into their service provision and thus acquired commercial value. Ordinary households and businesses joining the Internet brought a consumer mentality rather than a technical-engineering perspective to the selection and utilization of domain names.
The most dramatic stimulus to the demand for domain names came from the emergence of the World Wide Web between 1990 and 1995. The World Wide Web was a client-server software application that made the Internet easier to navigate and more fun to use by linking and displaying documents (or other objects stored on networked computers) by means of a graphical user interface. The software code for Web servers and the first portable browser were created by European physicists at CERN in 1990 (Cailliau 1995). The Web was popularized by the public release of a graphical browser called Mosaic in early 1993 by the National Center for Supercomputer Applications in the United States. [5 ]
Only a year after its release, in January 1994, there were 20 million WWW users, 95 percent of them using Mosaic. The World Wide Web's hypertext transfer protocol (HTTP) had become the second most popular protocol on the Net, measured in terms of packet and byte counts on the NSFNET backbone. [6 ]After one more year, in early 1995, the World Wide Web passed the venerable file transfer protocol (ftp) as the application generating the most traffic on NSFNET. With the founding of Netscape in 1994, browsers and Web software became a commercial industry. Netscape released its first Navigator browser at the beginning of 1995 and quickly displaced Mosaic. Microsoft rushed Internet Explorer to market at the end of the year. With user-friendly, point-and-click navigation available, the Internet attracted a much broader base of users, including household consumers and small businesses. The Internet had suddenly become a mass medium for communication and commerce.
A quantum change now took place in the status of domain names. The Web had its own addressing standard, known as Uniform Resource Locators (URLs). URLs were designed to work like a networked extension of the familiar computer file name. Web documents or other resources were given names within a hierarchical directory structure, with directories separated by slashes. In order to take advantage of the global connectivity available over the Internet, URLs used a domain name as the top-level directory. The basic syntax of a URL could be represented thus:
http://<domain name>/<directory or resource name>/<directory or resource name>/etc . . .
The hierarchy to the right of the domain name could be as shallow or as deep as the person in charge of the Web site wanted. URLs were never intended to be visible to end users. The Web's inventors thought they would hide behind hyperlinks.
By using domain names as the starting point of URLs, the Web altered their function in profound and unanticipated ways. As the term resource locator suggests, Web addresses were names for resources, which meant any kind of object that might be placed on the Web: documents, images, downloadable files, services, mailboxes, and so on. Domain names, in contrast, had been originally intended to name host computers, machines on the Net. And URLs were not just addresses but locators of content. A user only needed to type a name into the URL window of a browser and (if it was a valid address) HTTP would fetch the corresponding resource and display it in the browser. A URL included 'explicit instructions on how to access the resource on the Internet' (Berners-Lee 1994). Domain names, in contrast, were originally conceived as locators of IP addresses or other resource records of interest to the network, not of things that people would be interested in seeing.
As the Web made it easy to create and publish documents or other resources on the Internet, the number of Web pages began to grow even faster than the number of users. It did not take users long to discover that shorter, shallower URLs were easier to use, remember, and advertise than longer ones. The shortest URL of all, of course, was a straight, unadorned domain name. Thus, if one wanted to post a distinct set of resources on the Web or create an identity for an organization, product, or idea, it made sense to register a separate domain name for it rather than creating a new directory under a single domain name. For example, a car manufacturer like General Motors with many different brand or product names such as Buick or Oldsmobile eventually learned to just register buick.com and use that as the URL rather than gm.com/cars/buick/, even if all the information resided on a single computer. The DNS protocol made it fairly easy to point multiple domain names at the same computer, so there was not much waste of physical resources. Domain names began to refer to content resources rather than just network resources.
As more and more users began to type domain names into their browsers' URL windows, yet another fateful transformation of domain names' function occurred. Many novice users did not understand the hierarchical structure of the domain name system (DNS) and simply typed in the name of something they wanted. The Internet would interpret this simple name as an invalid domain and return an error message. As a user-friendly improvement in Web browser software, the browser manufacturers began to use .com as the default value for a name typed in with no top-level extension. If the user typed 'cars' into the URL window, for example, instead of returning an error message the browser would automatically append .com to the end and www. to the beginning, and display to the user the Web site at www.cars. com. In doing so, the browser manufacturers reinforced the naive end user's tendency to treat domain names as a kind of directory of the Internet. This practice also massively increased the economic value of domain names registered under the .com top-level domain. For millions of impatient or naive users wary of search engines and other more complicated location methods, the default values turned the DNS into a search engine exclusively devoted to words registered under the .com domain.
Although it would take several years for the full economic effects to be felt, the 'Webification' of domain names was the critical step in the endowment of the name space with economic value. It massively increased the demand for domain name registrations and gave common, famous, or generic terms under the .com space the commercially valuable property of being able to effortlessly deliver thousands if not millions of Web site 'hits.'
A serendipitous intersection of technologies produced human and market factors that transformed one technology's function. The transformation of domain names was driven by rational economic concerns about visibility in an emerging global marketplace. In the early days of the Web, a simple, intuitive name in the .com space might generate millions of viewers with very little investment. If someone else controlled 'your' name in that space, your reputation or customer base might suffer. Thus, for economic and legal reasons, DNS policy has ever since been fixed upon the use of domain names as locators of Web sites. The forms of regulation and administration being imposed on DNS by ICANN are largely based on the assumption that DNS is used exclusively for that purpose.
Technologists who object that 'DNS was never designed to be used this way' are correct in a narrow sense but miss the larger point. Many technologies end up being used in ways that their designers never intended or visualized. These unanticipated uses in turn can generate inflection points in a technology's evolution by provoking new forms of economic activity and new forms of regulation. This in turn can reward certain technological capabilities and effectively foreclose others.
No. of Second-Level Registrations
Sources: For 1994-Network Wizards Internet Domain Survey, <http://www.isc.org/ds/WWW-9501/second-levels.html>; for 1996-Registration Services Performance Measures for February 1996, <http://www.networksolutions.com/en_US/legal/internic/coop-stats/feb96.html>.
The cooperative agreement between the National Science Foundation and Network Solutions (NSI) for registration services was concluded just months before the Web's sudden transformation of domain names.
Neither party to the transaction had any idea of what was in store for them. Post-Web, the new registry was faced with a huge increase in the volume of registrations, and almost all of the increase was concentrated in the .com top-level domain. Registration applications handled by Network Solutions went from 300 per month in 1992 to 1,500 per month in mid-1994, then to over 30,000 per month in late 1995. The statistics in table 6.1 show the growth in the total number of names registered by the Inter-NIC (see section 5.5.2) from July 1994 to February 1996, as well as the dominant role of .com registrations in accounting for that growth.
NSF had no charter to support commercial registrations. A report by Jon Postel to the Internet Architecture Board in October 1994 observed that 'NSF is getting tired of paying for 2,000 .com registrations per month, each one taking about four minutes of someone's time, allowing for little to no screening of the requests.' [7 ]NSF held consultative discussions on charging for domain names, and an expert advisory panel brought in to evaluate the performance of the InterNIC contractors concluded in a December 1994 report that Network Solutions should 'begin charging for .COM domain name registrations, and later charge for name registrations in all domains.' [8 ]Shortly after the decision to charge was made, a multibillion-dollar Washington-area defense contractor, Science Applications International Corporation (SAIC), purchased Network Solutions. The transaction was concluded in March 1995. In a letter from NSF to Network Solutions dated September 13, 1995, Amendment 4 to the cooperative agreement officially authorized the registry to charge fees for domain name registrations in .com, .net, and .org. [9 ]Initial registration of a name would cost US$100 and last for two years; annual renewal after the two-year period would cost US$50. Thirty percent of the registration fee would go into an 'intellectual infrastructure fund' at the disposal of the NSF. NSF would continue to pay for .edu registrations and on an interim basis for .gov. The charges went into effect September 14, 1995.
The Web and the commercialization of Internet access stimulated the development of a domain name market in other countries as well. In England, commercial ISPs arose in the early 1990s and formed an organization known as the London Internet Exchange (LINX). In order to meet the ISPs' demand for domain names for their customers, a registry was operated by a voluntary 'naming committee' that included the designated technical contact for the .uk top-level domain, Dr. William Black, and volunteers from various LINX members. No charge was levied by the naming committee for domain name registration. By 1996, however, the need for a more professional, well-defined, and open service for registrations in .uk prompted the creation of Nominet UK, a new nonprofit corporation. Nominet was organized as a wholesaler of domain names to ISPs, with the initial fee set at 60 pence for each two-year period. In Germany the .de registry was run by universities until late in 1996, when a consortium of ISPs formed DENIC. Asian and developing country registries, however, tended to remain closely tied to their roots in universities and government science and technology ministries for a longer period of time. As of March 1997 registries serving 67 of the world's country code top-level domains (ccTLDs) charged fees for domain name registration. [10 ]None of them, however, were operating at anything near the scale of the InterNIC.
The InterNIC's response to the Web explosion turned the generic top-level domain name space into a common pool resource. The massive increase in the volume of registrations undermined the feasibility of any administrative rationing rules except first-come/first-served.
The complete opening up of domain name registrations was not quite a deliberate policy decision, but it was the only option, given the pressures of growth and the prior commitment to accommodating the widest diffusion of the Internet. The costs created by registrations quickly exceeded the staff and budget constraints of the InterNIC cooperative agreement. Network Solutions was using funds from other projects to cover the cost of the staff, office space, phone systems, and computers needed to keep up. The InterNIC could not keep up with growth if it attempted to review and police registration applications. As a former worker put it, 'In growth from 400 requests per day total-including new, modify and delete for domains, contact updates, host updates-to 25,000 and up per day, much of the focus became turn-around time and protection of data being updated via authentication. Policing was not given much priority under the technology and funding constraints. The whole registration process/budget was not designed for vanity-tagging the Internet.' [11 ]
As the process of registering a domain name was accelerated and automated, the InterNIC abandoned previous attempts to enforce a rule of one domain name per person. [12 ]It also gave up any attempt to maintain distinctions between the types of registrants that were allowed to use names in .com, .net, and .org. [13 ]It was not possible for InterNIC to decide whether a particular applicant had a 'right' to the specific name he was trying to register. Any such reviews, which would require manual handling of applications, would have slowed the execution of domain name registrations to a trickle, creating a bigger and bigger backlog.
Thus, the InterNIC's part of the name space became a common pool resource. Individuals appropriated units of the resource (they registered second-level names) using the rule of capture. There were almost no economic or legal constraints on appropriation. A user could register any name she wanted and take any number of names. Until September 1995 names could be registered at no charge (although users who relied on Internet service providers might incur registration-related service charges). Even after fees were imposed, it took months for Network Solutions to begin billing and much longer to collect effectively. [14 ]Once it was implemented, the fee for registration was trivial compared to the perceived economic value of many names. And the prices were the same for all names, regardless of the variations in their value. The source of the economic value, of course, was a name's ability to deliver the browser-using public to a particular Web site.
The Web set in motion a positive feedback loop that led to the overwhelming dominance of the domain name market by .com, .net, and .org for the rest of the decade. The initial flood of registrations under .com encouraged browser programs to make it the default. The browser defaults vested .com registrations with a special value. That value encouraged individuals to appropriate names in .com, leading to even faster growth in that domain. The large number of registrations in .com reinforced the expectations of the user public that most of the content on the Web would be registered under .com, making it more likely that users would look for sites there. That further accelerated the demand for .com registrations, continuing the cycle. The .net and .org domains became second-best options for those who could not get .com or a way of protecting the exclusivity of a .com registration through multiple registrations. As the number of registrations exploded, it became less and less feasible to discriminate among applications.
The global dominance of .com was further reinforced by the more restrictive approaches to registration taken in most other countries. While the InterNIC strained and struggled to accommodate demand, registries in many other countries imposed rigid rules on who could get a domain name and how many they could get. Whereas .com opened up the second level of the hierarchy to any taker, many country codes created naming conventions at the second level that users were forced to fit into. Japan and France, for example, heavily restricted eligibility for domain names and tried to fit all registrations into predetermined hierarchies. [15 ]The .com, .net, and .org domains also were more attractive to businesses seeking a global audience because of their generic character. Thus, by July 1995 there were two and a half times as many host computers under the Inter-NIC domains (3.92 million) as there were in the seven largest country domains combined (1.52 million in Great Britain, Japan, Germany, Australia, Canada, Netherlands and France). [16 ]The U.K. registry, the second largest in the world at the end of 1996, was fielding 3,000 to 4,000 registrations a month; Network Solutions was registering 75,000 to 85,000.
As late as 1999 more French organizations were registered under .com than under .fr. As of late 2000 barely 200,000 domains were registered under .jp (Japan), fewer than the .com domain in January 1996.
By allowing the market to evolve spontaneously as a common pool, the Americans created rights conflicts, but they also created an entirely new industry and cultivated among U.S. businesses the technical and management skills needed to achieve global leadership in it. Moreover, precisely because the uncontrolled appropriation activity pulled the government and the industry into new kinds of property rights conflicts, American stakeholders, for better or worse, would take the lead in defining the terms of the institutional innovations that would be required to resolve them.
An inescapable feature of common pool resources is that as demand intensifies, appropriators are more likely to come into conflict with each other. As this happens, the conflicting parties may begin to articulate property claims and seek to have exclusive ownership rights created or upheld by legal and political institutions. Property rights conflicts over domain names began to achieve public visibility in 1994, fairly soon after the Web's transformation of the .com, .net, and .org domains. The conflicts became widespread in 1995 and 1996.
There were two catalysts of rights conflicts. One was the perceived clash between trademark protection, a preexisting form of property rights in names, and second-level domain name registrations. The other was a conflict over the right to top-level domain name assignments, which were valued because they might bring with them the right to sell registrations to second-level domain names. Both conflicts led inexorably to contests over control of the root, for whoever set policy at the root level would significantly affect events at the lower levels. The rest of the chapter follows the rights conflicts up the domain name hierarchy, starting with second-level domains and moving up to the root level.
[1 ]Nysernet formed PSINet, other regionals formed Cerfnet, and UUNet. Later, these three joined with Sprint to form the Commercial Internet eXchange (CIX) in 1991.
[2 ]Email, Steve Wolff, director, NSF CISE, to Eric Aupperle, president, Merit Network, May 24, 1991, <http://www.merit.edu/merit/archive/nsfnet/nsf.agreements/commercial.traffic>.
[3 ]Project Solicitation 93-52, Network Access Point Manager, Routing Arbiter, Regional Network Providers, and Very High Speed Backbone Network Services Provider for NSFNET and the NREN Program.
[4 ]NSF also supported a 'routing arbiter' to provide a database and other information needed by the NAPs to exchange traffic in an orderly fashion, as well as some transitional funding for the regional networks.
[5 ]Mosaic was the outgrowth of a program written by the Software Development Group at the NCSA called Collage, designed to enable researchers to collaborate over networks. As the project neared completion, programmers at the group got wind of the World Wide Web project and quickly realized that Web compatibility could turn the Collage project into something much broader than a collaboration tool. See <http://www.webhistory.org/historyday/abstracts.html>.
[6 ]NSFNET backbone statistics are archived at the Georgia Tech Graphics, Visualization and Usability Center, <http://www.cc.gatech.edu/gvu/stats/NSF/merit.html>.
[7 ]Minutes, October 13, 1994, IAB Meeting, <ftp://ftp.iab.org/in-notes/IAB/ IABmins/IABmins.941013>.
[8 ]'It is clear from the materials presented by NSI that a primary culprit in the RS work load is the .COM domain. . . . At present, the management of .COM is paid for by the NSF, and hence increasing demand for .COM registrations will require increasing support from the NSF. The panel recommends that NSI begin charging for .COM domain name registrations, and later charge for name registrations in all domains.' InterNIC Midterm Evaluation and Recommendations: A Panel Report to the National Science Foundation, December 1994, <http://www. networksolutions.com/en_US/legal/internic/midterm/index.html>.
[10 ]Global Domain Names Status Report, NetNames, London, March 1997.
[11 ]Email, August 17, 1998, from former NSI employee to newslist.
[12 ]In July 1993, Brian Reid, who at the time worked for Digital Equipment Corporation, received an email from InterNIC stating, 'We try to register only one name per 'organization.' DEC.COM has been around since day-one. Do you intend to replace DEC.COM with DIGITAL.COM?' <email@example.com> to Brian Reid, July 16, 1993; on file with author.
[13 ]According to Kim Hubbard, former director of ARIN, an early employee of Network Solutions, and a worker at the InterNIC from 1991 to 1996, it was Jon Postel who decided that it was a 'waste of time' to attempt to segregate .com, .net, and .org registrations.
[14 ]Network Solutions did not threaten to deactivate domain names for nonpayment of the fees until June 1996. A copy of a June 17 news release is archived at <http://www.iiia.org/lists/newdom/1996q2/0295.html>.
[15 ]The French ccTLD, for example, in 1996 required users to fit into one of eight categories: .asso, .barreau, .cci, .cesi, .dxxx, .gouv, .presse, and .tm.
[16 ]Internet Domain Survey, July 1995, <http://www.isc.org/ds/WWW-9507/distbynum.html>.
For U.K. statistics, see <http://www.nominet.org.uk/news/stats/stats.1996.html>; for U.S. statistics see <http://www.networksolutions.com/en_US/legal/internic/coop-stats/ >.
|< Day Day Up >|| |