1. The ARPANET Roots and the Birth of IPv4 (1973–1981)
In the early 1970s, the Advanced Research Projects Agency Network (ARPANET) operated using the Network Control Program (NCP). As packet-switching networks expanded across satellite links and radio networks (PRNET), DARPA researchers Vinton Cerf and Robert Kahn recognized that NCP could not bridge heterogeneous network architectures.
In 1974, Cerf and Kahn published "A Protocol for Packet Network Intercommunication", introducing the Transmission Control Program. Later split into TCP for connection-oriented transport and IP for datagram routing, this architecture was codified in September 1981 with RFC 791, authored by RFC Editor Jon Postel. This marked the official birth of IPv4.
On January 1, 1983—widely known as Flag Day—ARPANET permanently disabled NCP and mandated TCP/IP, creating the operational foundation of the modern Internet.
2. Why Did IPv4 Use a 32-Bit Address Space?
RFC 791 allocated 32 bits for an IPv4 address, yielding 4,294,967,296 (232) unique numerical addresses. In 1981, when the entire connected research network comprised only a few hundred mainframe computers, 4.3 billion addresses seemed virtually inexhaustible.
As Vint Cerf famously reflected in later interviews:
"I thought it was an experiment, so 4.3 billion addresses seemed like enough to discover if the technology worked. Who could have known the experiment would never end?"
3. The Classful Era and the Inefficiency Crisis (1981–1993)
Initially, address allocation followed rigid class boundaries defined in RFC 790:
- Class A (0.0.0.0/8 to 127.0.0.0/8): 128 networks, each with 16,777,214 hosts. Reserved for massive organizations like IBM, MIT, General Electric, and the US DoD.
- Class B (128.0.0.0/16 to 191.255.0.0/16): 16,384 networks, each with 65,534 hosts. Heavily requested by universities and medium enterprises.
- Class C (192.0.0.0/24 to 223.255.255.0/24): 2,097,152 networks, each with only 254 hosts. Too small for most emerging businesses.
Because Class C was too small and Class A was out of reach, virtually any growing company requested a Class B block. An enterprise with 500 computers received 65,534 addresses, wasting more than 65,000 IPs. By 1992, the IETF realized that the Class B address pool would be depleted within a few short years, and global router BGP routing tables were growing exponentially.
4. Buying Time: The Inventions of CIDR, NAT, and RFC 1918 (1993–1996)
The Internet Engineering Task Force (IETF) rapidly developed emergency triage solutions that extended IPv4's lifespan by more than two decades:
- CIDR (RFC 1519 - 1993): Classless Inter-Domain Routing abolished fixed octet boundaries. Subnet masks could now split host and network bits at any position (e.g.
/23for 510 hosts,/22for 1,022 hosts), while supernetting compressed bloated routing tables. - Private Address Space (RFC 1597 / RFC 1918 - 1994/1996): Reserved non-routable blocks for internal company networks (
10.0.0.0/8,172.16.0.0/12, and192.168.0.0/16). - NAT (RFC 1631 - 1994): Network Address Translation permitted thousands of private hosts behind a gateway to share a single public IPv4 address using port translation (NAPT/PAT).
While NAT successfully preserved IPv4 address space, it violated the core architectural principle of end-to-end IP reachability, complicating VoIP, peer-to-peer networks, and online gaming.
5. What Happened to IPv5? The Mystery of Version 5
One of the most frequent networking trivia questions is: Why did the industry jump directly from IPv4 to IPv6? Where is IPv5?
The answer lies in the 4-bit "Version" nibble located in every Internet Protocol packet header (bits 0–3):
Version Bit Nibble:
0100 (Binary 4) = IPv4 (RFC 791)
0101 (Binary 5) = ST / ST-II (Internet Stream Protocol - RFC 1190 & RFC 1819)
0110 (Binary 6) = IPv6 (RFC 2460 & RFC 8200)Between 1979 and the 1990s, DARPA and the IETF developed the Internet Stream Protocol (ST / ST-II), an experimental protocol designed to carry real-time voice, video, and conferencing over distributed networks. ST used version number 5 in its packet header. Although ST was never commercialized as a public routing protocol, the version 5 codepoint had already been allocated by IANA. To prevent protocol collision in network analyzers and routers, the successor protocol was formally designated as IPv6.
6. The Standardization of IPv6 (1994–1998)
In 1993, the IETF launched the IP Next Generation (IPng) initiative (RFC 1550). After evaluating rival proposals including TUBA (TCP and UDP with Bigger Addresses) and CATNIP, the working group selected SIPP (Simple Internet Protocol Plus), authored by Steve Deering and Robert Hinden.
The foundational specifications were published in RFC 1883 (1995) and formalized in RFC 2460 in December 1998 (later elevated to an Internet Standard in RFC 8200 in 2017). IPv6 brought groundbreaking architectural enhancements:
- 128-Bit Address Space: 340 undecillion addresses (2128), enough to assign trillions of IPs to every grain of sand on Earth.
- Streamlined Fixed Header: Fixed 40-byte header with fewer fields, enabling line-rate hardware routing without checksum calculations at every hop.
- Stateless Address Autoconfiguration (SLAAC - RFC 4862): Devices automatically generate their own globally routable IP addresses using Router Advertisements and neighbor discovery, eliminating mandatory DHCP servers.
- Elimination of Broadcast: Broadcast storms were eradicated and replaced with targeted Multicast and Anycast addressing.
- End of Mandatory NAT: Restores true peer-to-peer end-to-end communication across the globe.
7. The Exhaustion Timeline: Running Out of IPv4
Despite CIDR and NAT, exponential adoption of smartphones, broadband, IoT devices, and cloud computing consumed the remaining IPv4 pool:
Timeline of IPv4 Exhaustion Milestones:
• Feb 3, 2011: IANA distributes the final five /8 blocks to the 5 RIRs.
• Apr 15, 2011: APNIC (Asia-Pacific) exhausts standard free pool.
• Sep 14, 2012: RIPE NCC (Europe & Middle East) exhausts free pool.
• Jun 10, 2014: LACNIC (Latin America & Caribbean) exhausts free pool.
• Sep 24, 2015: ARIN (North America) exhausts free pool.
• Nov 25, 2019: RIPE NCC allocates its very last remaining IPv4 block.With official registries exhausted, organizations requiring additional IPv4 addresses turned to the secondary broker transfer market, where IPv4 addresses now trade for upwards of $40 to $60 per IP address.
8. The Present & Future: Dual-Stack and IPv6-Only Deployments
On June 8, 2011, major internet companies participated in World IPv6 Day, followed by the permanent deployment on World IPv6 Launch (June 6, 2012).
Today, mobile operators (such as T-Mobile, Reliance Jio, and Verizon Wireless) run pure IPv6-only cellular backbones, translating legacy IPv4 traffic via 464XLAT (RFC 6877). Major hyperscalers (Google, Cloudflare, Meta, Amazon) report that more than 45–50% of their global web traffic travels natively over IPv6.
As modern hyperscale data centers transition to pure IPv6-only internal topologies to eliminate the compute and memory overhead of NAT tables, the multi-decade journey from ARPANET's experimental 32-bit protocol to the infinite 128-bit fabric of IPv6 represents one of engineering's greatest triumphs.