57 years of The Internet and TCP/IP

Photo by Scott Rodgerson on Unsplash

The internet started as a small research project and academic experiment in the 1960s by the US government’s Advanced Research Projects Agency (ARPA). It was initially named ARPA Network (ARPANET), and its primary immediate purpose was to share resources between universities. As a matter of fact, the first official live connection was between UCLA, Stanford, UC Santa Barbara and University of Utah in December 1969.

In 1970, Vinton Cerf and Robert Kahn, the foundational “creators” of ARPANET, developed the TCP/IP standard for it which allowed discrete networks to connect to it to birth the inter-network (internet). The rest is history.

⚡IPv4 Addressing

IP Addressing is a key component of TCP/IP and TCP/IP networking can not exist without it. IPv4, the first official TCP/IP addressing scheme, was introduced under RFC 791 in 1981. Theoretically, it’s a 32-bit system with ~4.3 billion (2³²) unique IPv4 addresses. It was assumed that this seemingly huge pool of addresses was inexhaustible because only a few 100 hosts were connected globally at the time.

Every host is required to have its own unique IP address in TCP/IP but to be able to surpass the LAN and connect to the internet, the IP has to be a publicly routable IPv4 address referred to as a “public IP.” The other type of IPv4 address is a “private IP” and it’s meant for LAN-only connectivity.

⚡The World Wide Web

In 1989, British scientist Tim Berners-Lee invented the World Wide Web, which truly democratized the internet and eventually led to a commercial explosion.

People often confuse the internet with the World Wide Web but they’re not the same. The internet is the underlying network infrastructure (the roads). The World Wide Web is the system of websites and hyperlinks (the cars) sitting on top of the internet.

By the early 1990s, courtesy Tim Berners-Lee’s World Wide Web, over 2 million hosts were already connected to the internet. This exponential growth coupled with the inefficient allocation of public IPs completely shattered the initial projection for the ~4.3 billion IPv4 address pool. If no remedial action was taken, the pool was going to run out within a short period and potentially collapse the internet or jeopardize further connectivity.

⚡IPv6 Addressing

IPv6, a 128-bit IP addressing scheme, was introduced as a long-term fix. In theory, it has an address pool of 340 undecillion (2¹²⁸), which essentially means that there’s enough unique addresses for every single grain of sand on earth, with still enough for thousands of more earths.

⚡NAT

Network Address Translation (NAT) was introduced in 1994 under RFC 1631 as a makeshift solution while IPv6 was in development. With NAT, an entire SOHO or enterprise LAN can connect to the internet with a single public IP. This means that 10 separate SOHO/enterprise LANs with 100 nodes each can connect to the internet by using a total of only 10 public IPs instead of 1000.

A CPE NAT gateway (typically a router) in each LAN would act as an IPv4 proxy by bearing a single public IP to concurrently coordinate internet access for every node through a process called Port Address Translation (PAT).

⚡CGNAT

Carrier-Grade NAT (CGNAT) was introduced in the early 2000s to basically propel traditional NAT to enterprise scale and take it up a notch. With it, an ISP NAT gateway acts as the last mile IPv4 proxy, with the CPE NAT gateway right behind it. This architecture makes it possible for several SOHO/enterprise LANs to connect to the internet with a single public IP.

Using the scenario above, CGNAT can enable 100 separate LANs with 100 nodes each to connect to the internet over 1 public IP. That translates to 10,000 nodes behind a single public IP.

⚡End Game: IPv6-only

CGNAT is the ultimate stopgap for the IPv4 dilemma. Eventually, IPv6 will completely take over from IPv4. There’s an ongoing push for full IPv6 adoption but the complete switch hasn’t materialized yet because there are a gazillion moving parts. Therefore, a sudden overhaul is literally impossible. It’s a gradual process that should be complete by 2050.

Until then, TCP/IP will continue to run a model referred to as Dual-Stack, where both IPv4 and IPv6 co-exist with priority given to IPv6 through the Happy Eyeballs algorithm (RFC 8035). This has been in effect since 2012.