The Backbone: Conclusion

And so we reach the conclusion of “The Backbone,” my story of the origins of the Internet1. We have seen the basic arc of the Internet’s development from the 1960s to the 1990s – nurtured in its youth by the government, given room to grow to fruition by the unravelling of the power of the Bell system, and finally emerging into public view in a frenzy of growth which smothered all potential competitors. Over the course of those decades, users repeatedly co-opted systems built in order to expand and share access to machine resources (time-sharing operating systems, ARPANET, and NSFNET), to use them instead for interpersonal communication – message boards and electronic mail.

In 1995, the National Science Foundation (NSF) successfully extricated itself from the network operations business while preserving a single, unitary Internet, composed of many heterogeneous but interlaced parts – networks owned by a variety of corporations; and websites and other services provided by an even wider variety of participants: hobbyists, local governments, small businesses, and more.

The Internet’s self-image coalesced in these years of its adolescence. It was distributed, decentralized, and decentralizing. Its most vocal proponents argued that its technological structure, which privileged the edges over the center, would replicate itself in new political structures, eroding the foundations of incumbent institutional power and enabling direct, disintermediated communication and market transactions between individuals. Louis Rossetto, editor of Wired, the magazine of the technologically-enlightened, put it this way:

This new world is characterized by a new global economy which is inherently anti-hierarchical and decentralist, and disrespects national boundaries or the control of politicians and bureaucrats or power mongerers of any; and by a global, networked consciousness that is creating a new kind of democracy for achieving social consensus that is turning the bankrupt electoral politics we are witnessing this year into a dead end. …a global hive mind that is arriving at a new, spontaneous order2.

This set of libertarian ideas found expression in statements repeated and replicated so often that they became a kind of scripture of the Internet: the book of David Clark (“We reject: kings, presidents, and voting. We believe in: rough consensus and running code.”); the book of John Perry Barlow (“Governments of the Industrial World, you weary giants of flesh and steel, I come from Cyberspace, the new home of Mind. ….You are not welcome among us. You have no sovereignty where we gather.”); the book of John Gilmore (“The Net interprets censorship as damage and routes around it.”).

These ideas took root, as I have said, when the Internet remained yet in its adolescent state – simultaneously a single, interconnected system, yet robustly heterogeneous in its structure at every level. Over the following fifteen years, its heterogeneity would diminish, with power over the network and its applications consolidating in a few key corporations3.

Boom: Networks Consolidate

As the popularity of the Internet exploded in the second half of the 1990s, a river of capital flowed into Silicon Valley in search of the huge returns promised by the unheard of yearly growth of digital traffic. The theory of the time held that, because of the global reach of the Internet and the power of network effects (e.g., Metcalfe’s Law), the first mover to occupy any given sector of online business would dominate it. According to this doctrine, short-term losses – even on per-unit sales basis – were irrelevant, even desirable. Only growth mattered, because growth could always be turned into profits later, once all potential competitors had dwindled into irrelevance4. This encouraged a gold rush mentality among investors, what we would now call FOMO, and attracted large amounts of money to such questionable enterprises as Priceline, Boo, and eToys, despite their evident lack of profitability.

Underneath this highly visible froth of increased application diversity, however, the deep current of network infrastructure flowed in the opposite direction, towards consolidation. The boom in web properties in the second half of the 1990s found its echo in a boom in fiber optic network construction. The telecom carriers, new and old, all wanted a piece of the exponential growth in data traffic promised by the rocket-like rise of the World Wide Web. The incumbent telecommunications carriers, freed from their silos by the 1996 Telecom Act, did not, as might have been hoped, use the opportunity to unleash their claws in all-out competitive battle, cutting profits to the bone to the benefit of their users. Markets may thrive on competition, but firms much prefer monopoly. And so the RBOCs and long-distance carriers, split apart in 1984, re-assembled themselves into giants with tremendous market power. Southwestern Bell absorbed Ameritech, Pacific Telesis, BellSouth, and finally AT&T, and then took the name of that former parent company. Bell Atlantic and NYNEX merged, then acquired GTE5, taking on at the same time a new moniker, Verizon. Of the former RBOCs, only US West remained an independent company.

While Southwestern Bell and Bell Atlantc re-assembled the scattered parts of AT&T into a pair of Frankenstein’s monsters, another would-be giant was busy absorbing the various internet service providers that had flourished in the first half of the 1990s. In 1983, Bernie Ebbers, who had made his first fortune as the owner of a dozen hotels, co-founded Long Distance Discount Services (LDDS) to compete in the market newly opened by the AT&T break-up, targeting long-distance service for small and medium-sized businesses. Over the ensuing decade, LDDS acquired a variety of other competitors, achieving fourth place in size among long-distance carriers behind AT&T, MCI, and Sprint. In the first days of the Internet boom, it took on a new name – WorldCom – thus announcing its newly hubristic ambitions. A new, much more extravagant buying spree ensued, using WorldCom’s its bubble-inflated stock to acquire one major network after another. In 1996, it bought Metropolitan Fiber Systems (MFS), which had itself just acquired major Internet provider UUNET. The acquisition of CompuServe’s network infrastructure from H&R block followed in 1997, along with the acquisition of the former NSFNET backbone operator, ANS, from AOL. The biggest purchase of all came in 1998, when WorldCom merged with MCI. After the market crash – and bankruptcy and scandal and prison6 – Verizon absorbed the wreckage of Bernie Ebbers’ conglomerate in 2006.

Finally, the cable television providers, though more historically localized in ownership than the rest of the telecom business, underwent a similar trend towards consolidation, with acquisitions across previously siloed markets making possible behemoths like Time Warner and Comcast; the latter became simultaneously the largest internet service provider and the largest pay-TV company in the U.S., to say nothing of its media holdings.

And so, by the mid-2000s, the diverse structure of peer networks which had characterized the early Internet in the U.S. had merged into a handful of major providers. At the retail level, as broadband networking replaced dial-up, most consumers had access to only one or two relevant ISPs – their local telephone and cable company – and nationally two companies dominated each of those sectors; Verizon and AT&T on the one hand, and Comcast and Time Warner on the other.

Bust: Applications Consolidate

Many of the first-generation dot-coms might have survived had they been allowed to grow gradually. But the gold rush theory was antithetical to anything but hypergrowth. So, when the crash came, most were caught with huge expenses and excess capacity due to overinvestment, and they quickly collapsed. A vigorous winnowing took place, with a great deal of chaff sifted out and only a few grains of wheat left behind.

Over the following decade, a new much more stable order emerged. Five giant companies came to dominate the application layer of the Internet. Two of them were dot-com era survivors. Google, founded in 1998, had raised the bar on what it meant to be a search engine by extracting ranking information from the structure of the Web and its skein of links, rather than merely from the content of individual pages. Through a series of further innovations and acquisitions, it leveraged its dominance in search into a powerful position in mobile computing, email, streaming video, and, of course, advertising. Amazon, founded as a book retailer in 1994, developed a world-beating logistical operation which it used to undercut competitors on price and delivery speed, then expanded into virtually every retail sector, and finally created a set of tools for hosting third-party applications that defined the new, and very lucrative, business of cloud computing.

Two of the other giants had come of age during the personal computing era, a decade before the commercial Internet. Microsoft took an early lead in the so-called “browser wars” of the mid-90s, but more important long-term was its continued dominance of business software, even as those businesses began to move their operations on-line. Apple Computer, made largely irrelevant in the 1990s by the dominance of Windows, seemed destined to a gradual decline into senescence. But it was rejuvenated by its successes with the iPod and iTunes, and then developed the most profitable mobile computing device to this day, the iPhone.

The final dominant power, Facebook, was the only one to come out of the second wave boom in Internet investment in the second half of the 2000s. It grew rapidly across college campuses before colonizing the wider world, and becoming the primary way that many millions of people keep in touch with people outside their immediate family and close friends. It has since acquired other communications businesses, and become a major platform for business marketing, real-time chat, and, of course, lots and lots of ads. Facebook’s rise marked the return of private digital spaces, comparable to the so-called “walled gardens” of the pre-Internet days such as CompuServe and Prodigy. Facebook has an entire ecosystem of user, community, and business pages that are only visible and accessible to a logged-in Facebook user. This is a sharp contrast to Facebook’s immediate predecessor, MySpace, which operated more like a customized version of a 90s-era web-hosting provider like GeoCities – it provided users with the ability to customize their own page on the public web, with very limited privacy options7.

These corporations did not monopolize access to applications in the same way that Verizon, AT&T and the like monopolize the delivery of packets. A huge variety of other companies of all sizes continued to thrive on the Web and in other parts of the Internet, and some of them (Netflix, for example), consume a far larger share of Internet bandwidth than the big five. But they do control much of the basic software infrastructure used by other companies of all sizes, and mediate much of the day-to-day interactions that people have online. This constitutes a major shift in the large-scale structure of the Internet.

Take email, for example. In the 1990s, this foundational communications platform of the Internet was almost entirely decentralized. Email messages nearly always departed from and arrived at a server owned by the sender or recipient’s university or employer, and then were downloaded onto the recipient’s local machine. Nowadays, however, most email is hosted on the servers of third-party providers, with Google by far the largest such provider. Likewise, finding information on the early Internet depended on a distributed system of Gopher servers, owned and operated by individuals or organizations, most often universities. Now, the vast majority of search traffic flows through Google – or Amazon for product searches. A final example is the rise of the nebulous “cloud.” It used to be the case that most corporate providers of Internet services, and even many hobbyists and small businesses, hosted their software on machines that they owned and operated. But in recent years the business-critical applications of even big players like Netflix can be found running in data centers owned and operated by Amazon, Microsoft, and Google.

The rise of app stores provide another example of how power has accrued to the likes of Google and Apple. In the early years of the Internet, new applications (like the World Wide Web) spread whenever one person or group of people wrote some software, and got others to download it and start using it. However, mobile computing vendors, wanting to protect their platforms from malicious software while taking a cut of all sales, required users to go through curated stores to find new applications. Without the say-so of the platform owner, it was no longer possible to spread your software (except to the few users sophisticated enough to bypass the app stores), and the rules governing access to those stores are extensive – Apple’s run to twenty-four printed pages, and include such details as forbidding “Overtly sexual or pornographic material” and “Apps that alter or disable the functions of standard switches, such as the Volume Up/Down and Ring/Silent switches” and requiring that “Your app should work on its own without requiring installation of another app to function.” All of these restrictions are sensible enough in the context of a corporation trying to create a user-friendly platform from which it can extract large amounts of money. But they break the symmetry of the Internet’s original topology, which devolved all semantic control of message transmission and interpretation to general-purpose computers on the edge of the network.

The forces driving the consolidation of power in a few application providers were many, but one interesting facet is the role of bad actors in pushing other users towards the likes of Google, Apple, and Facebook. The Internet was born in academia, among people with a set of shared cultural values (including the mission to expand human knowledge) and a baseline level of mutual trust. The applications they built reflected this assumption of mutual trust – for example, the fact that one can send an unlimited amount of email, free of charge, to any address. This all worked well enough until millions of new users flooded onto the Interenet, with none of those shared academic values. Email inboxes became buried under an avalanche of spam; Usenet choked on a firehose of advertising and pornographic binary files; FTP download sites for finding the latest free software became minefields of malware. Managed environments controlled by trustworthy companies employing high-quality spam filters became a welcome refuge from the relentless malice and avarice of one’s fellow man.

The De-Americanization of the Internet

This series of articles on “the backbone” has almost exclusively focused its attention on events, people, and institutions in the United States. That is a somewhat excusable way to tell the history of the Internet up through the 1990s. The Internet became international in scope very early on (even its ARPANET predecessor had international links), but for most of its history the U.S. took the lead in setting its policies and protocols and establishing its organizational structures. More generally, from the birth of the electronic computer up to the 1990s, the bulk of innovation in computing and networking came from American organizations (France and the U.K. have entered this story from time to time, and the creation of the World Wide Web by a Brit and Belgian in Switzerland is, of course, a striking exception to this general trend).

However, after 2000 this approach, convenient as it is for an American and anglophone such as myself, becomes increasingly hard to justify. American hegemony over the Internet has receded in a variety of ways. The most basic change one could point to is the loss of the United States’ position as the leader in Internet access – although still high in rankings relative to most of the globe, it fell well back from the top spot by 2010, with South Korea, the Scandanavian countries and the Netherlands all exceeding the U.S. in broadband household penetration and speed8. Another symbolic milestone came in 2009, when the body that allocates Internet domain names, the Internet Corporation for Assigned Names and Numbers (ICANN), agreed to leave the oversight of the U.S. Department of Commerce and become an independent international organization – a promise that was fulfilled in 2016.

More profoundly, after 2000 other polities, from China and Russia to the European Union, began formulating their own legal and technical frameworks to mold the Internet to local mores and the realities of geopolitical power9. Major landmarks in this process include the Russian law passed in 2014 which requires all data on Russian citizens held by online services to be stored in-country, and the 2016 passage of the EU’s General Data Protection Regulation10. Authoritarian countries also established patrolled digital borders to prevent information inconvient to the regime from entering the country – most famous of these being the “Great Firewall of China” first established in 2006. Even as it walled itself off from the rest of the world, however, China built up its own native Interent services, some of which, such as Bytedance and Tencent, have now acquired a global reach. The rise of these competing forces – whether corporation, state, or a hybrid of the two – fragmented power over the Internet without decentralizing it. At least, not in the sense that Rossetto had imagined.

In the 1970s, the end of the post-war economic boom and the rise of stagflation led to the era of deregulation, a widespread effort to unleash market forces on infrastructural sectors previously sheltered from them in the name of stability and equal access. In the face of the ongoing centralization of power within the Internet and decreasing American control over its structure, the U.S. may be entering a similar Kuhnian crisis with respect to the current laissez-faire Internet policy regime. The past sixth months alone have witnessed the U.S. government’s demand that TikTok be divested from Chinese ownership, calls from the Justice Department for Section 230 of the Communications Decency Act to be heavily revised and restricted, and a pending lawsuit against Google, the first American antitrust action against a major Internet application provider. Whether these events prove transitory or of lasting significance, the backbone as it was in 1995 is no more. Its adolescence having long since faded into nostalgic reminiscence, the Internet may now find itself forced to endure a mid-life crisis.


  1. Of course there is much more that one could say about the history of the Internet than I have said – many networks overlooked, many technical developments ignored, many personalities and events passed over, many analytical perspectives unexamined. This is simply the story that I have chosen to tell.

  2. Quoted in David Hudson, Rewired: A Brief (and Opinionated) Net History (1997), 236.

  3. As we saw with Usenet and FidoNet, even the most anarchic and unplanned networks have a strong tendency to develop a hierarchical structure. Though the technical specifications of the network are may be entirely democratic, differences in resources between nodes and the efficiency to be gainted in limiting the number of hops required to reach any destination create a gravitational tendency for flat mesh networks with many peer-to-peer connections to collapse into hierarchical stars with central nodes.

  4. Only one survivor of the dot-com era has succeeded in executing this strategy over the long term – Amazon.com.

  5. GTE had by that time acquired BBN, which had acquired many of the NSFNET regional networks. As part of the condition of the GTE acquisition, Verizon spun off BBN’s networks as Genuity, which in turn was swallowed and swallowed again, and now belongs to Lumen Technologies – the direct ancestor of that last independent RBOC, US West.

  6. Apologies to Frank Capra, et. al.

  7. Until 2006, only teenagers under 16 could mark their profiles private, such that only their friends could see them. The rest were visible and searchable on the open web.

  8. See OECD, “Households with broadband access”; Akami, “The State of The Internet – 4th Quarter, 2010 Report”. Both accessed on November 10, 2020.

  9. The early stages of this process are documented in Goldsmith and Wu, Who Controls the Internet? Illusions of a Borderless World (2006).

  10. GDPR, in particular, has had global, not just regional, effects, forcing compliance from American giants like Google. California also modeled its own Consumer Privacy Act on its European predecessor.

1 thought on “The Backbone: Conclusion”

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s