The Internet Is a Network of Networks, Not a Protocol
Historical Context of the Internet & the Web
The Internet is a physical and operational system composed of interconnected networks. It exists only where independent networks are physically linked, operationally coordinated, and commercially sustained. Protocols operate inside that system. They do not create it.
TCP, IP, BGP, DNS, HTTP, HTML, SSL, and the World Wide Web define how data behaves once a network exists. They do not build networks, lay fiber, negotiate interconnection, guarantee performance, or enforce service levels. Confusing protocols with the Internet itself is the foundational error behind most Internet origin myths.
The Standards Said So Themselves
The title of this page is not a slogan. It is the standards’ own language. RFC 1122, Requirements for Internet Hosts, edited by Robert Braden and published in October 1989, states it in seven words: “The Internet is a network of networks.” RFC 793, the Transmission Control Protocol specification published by Jon Postel in September 1981, opens by defining TCP as “a highly reliable host-to-host protocol between hosts in packet-switched computer communication networks, and in interconnected systems of such networks.” TCP lives in the hosts at the edges. The networks between the hosts are presupposed. TCP presupposes the Internet. The Internet does not presuppose TCP.
The official registry agreed. RFC 1200, the IAB Official Protocol Standards document of April 1991, graded IP as Required and TCP as only Recommended. Adoption itself was a policy decision, not a law of nature: the Internet Society’s own history records that in 1985 Dennis Jennings, leading the NSFNET program at the National Science Foundation, worked with the community toward the decision “that TCP/IP would be mandatory for the NSFNET program.” The full protocol record is at Internet Protocols and OSI.
The Web Made Information Visual, Not Global
The World Wide Web is a software layer, not an Internet. Hypertext predates the Web by decades. Robert Cailliau proposed the CERN hypertext system in 1987. Tim Berners-Lee implemented an early version in 1990 on a NeXT workstation. Nicola Pellow made it usable beyond CERN with the Line Mode Browser. Mosaic made it visually appealing in 1993.
None of that created global reach. Matthew Gray’s survey at MIT counted 23,500 web sites in June 1995 and roughly 100,000 by January 1996. The Network Wizards DNS survey counted 9.5 million Internet hosts in January 1996. Worldwide users numbered in the low tens of millions, and every contemporaneous count concentrated them in the United States. Large portions of the world were effectively unreachable. A website could be written, standards-compliant, and operational, and still be inaccessible to most of the planet.
The limitation was not software alone. It was infrastructure, and it was policy. The NSFNET Acceptable Use Policy banned commercial traffic on the government backbone, and ecommerce on the Internet was banned until April 30, 1995, the day the NSFNET backbone was decommissioned. Performance was unpredictable. Security was minimal. Commerce at scale was impossible. Software existed. A global Internet did not.
Protocols Describe Behavior. Infrastructure Determines Reality.
TCP/IP does not lay fiber.
HTTP does not cross oceans.
HTML does not negotiate peering.
Browsers do not guarantee delivery.
The Internet requires:
- Physical transport (terrestrial + submarine)
- Direct Tier 1 interconnection
- Operational control
- Redundancy and disaster recovery
- Enforceable service levels
- Capital willing to take global risk
None of that is provided by protocols. None of it is provided by standards bodies. None of it is provided by browser authors. The Internet exists only where networks are unified into a single operational system.
1996: When the Internet Became Real
By 1996, TCP/IP was over 20 years old, and the World Wide Web was six years old. Yet the Internet still did not exist as a global system.
Regional networks were loosely peered, with no guarantees and no accountability. ISPs were territorial. Telcos were monopolistic. Governments restricted interconnection. No one had built a single end-to-end global network.
That changed in 1996, and the change carries dates. On August 29, 1996, Autonomous System Number 6553 was issued to Digital Island. In September 1996, while cofounding Digital Island, Mark Nichols proposed Merchant Transport and Tier-0 networking: a private, quality-of-service-engineered media layer for commerce, publicly accessible and eCommerce native. The Cisco Systems Remote Data Services Agreement, effective November 1, 1996, put Cisco Systems’ commerce traffic on that network under contract, with secure transaction response bounded below 300 milliseconds. The proof-of-concept network activated in the fourth quarter of 1996, and the production IPLC network went live in the first quarter of 1997, placed within one network hop of 99 percent of all Internet-accessible users. Global reach stopped being aspirational and became contractual. Infrastructure caught up to software.
This transition did not come from academia. It did not come from protocol authors. It did not come from browser teams. It came from execution.
The Line History Keeps Getting Wrong
Protocols made the Internet possible.
Infrastructure made it real.
The Internet did not emerge when code was written. It emerged when someone unified the world’s networks into one operational system. That unification required agreements, circuits, data centers, staffing, operational processes, and risk-taking at global scale.
That is the line history refuses to draw. Protocol authors enabled possibility. Execution created reality. Without the physical and operational network, TCP/IP and the Web remained ideas rather than a functioning worldwide utility.
Digital Island bridged that gap. By deploying infrastructure, acquiring customers, and enforcing service levels, Digital Island made the Internet contractually and commercially viable across continents. Digital Island’s network delivered performance, security, and uptime guarantees previously unavailable at global scale. The dated record is in the evidence vault and the Digital Island S-1.
Without this work, a browser could display content, but it could not reach the majority of the world. eCommerce could be designed, but transactions could not complete reliably. Financial systems, media distribution, and global communications remained fragmented and regional.
Execution also created accountability. Investors, customers, and regulators could now contract for real outcomes. The Internet became not just a research experiment, but a utility with enforceable service expectations, marking a fundamental shift in how the world could operate digitally.
The Modern Internet is eCommerce native and publicly accessible. It lives in the media layer, not the host layer. TCP was temporary. OSI never materialized. The Internet is a network of networks, not a protocol.
Sources
1. RFC 793, Transmission Control Protocol, September 1981
2. RFC 1122, Requirements for Internet Hosts, R. Braden, Editor, October 1989
3. RFC 1200, IAB Official Protocol Standards, April 1991
4. Internet Society, Brief History of the Internet
5. Matthew Gray, MIT, Measuring the Growth of the Web, 1993 to 1996
6. Matthew Gray, MIT, Internet Growth Summary, Network Wizards host counts