A brief history of IPv4 address space exhaustion

One day we will run out of IPv4 addresses. Probably it's not the first time you heard that. In this post I review the history of IPv4 address depletion, which to some extent is the same as reviewing the history of the Internet

Posted by Diego Pino García on May 25, 2017

IPv4 address space exhaustion was a hot topic in the 90s, when everyone started to foresee that inevitable future. However, we’re still relying on IPv4 today. So, what has actually happened? Did anyone find a vast range of unused IPv4 addresses locked in a closet? What happened to IPv6?

Reviewing the history of IPv4 address depletion is also reviewing the history of the Internet. Many decisions about the Internet have been made with the goal of solving or mitigating this problem. In this post I start from the beginning up to today. It’s not intended to be an exhaustive guide, but a recap of the most important events.

8-bit Internet

The Internet has its origin in the ARPAnet. A research network funded by the Advanced Research Projects Agency of the United States of America Department of Defense.

ARPAnet came to live in 1969, connecting just 4 hosts. The network grew in size over the years connecting more and more hosts, mainly universities and research centers in the US. In 1981, there were a total of 213 hosts connected.

But back in the days of the ARPAnet, there was not TCP/IP. Its equivalent was NCP (Network Control Protocol). Addresses in NCP were 8-bit numbers. That means each host could be addressed by a simple number such as 10, 23 or 145. Although popular, the ARPAnet was not the single computer network that fostered during the 70s. There was a need of connecting these networks in an inter-network or internet.

Already in the early 70s, Robert Khan from Darpa and Vinton Cerf, developer of NCP, started to work in a new protocol that allowed communications across several heterogeneous networks. The proposed protocol was called Transmission Control Program, first published in 1974 (A Protocol for Packet Network Intercommunication). Implementations of the protocol went through 4 major versions. In version 3, the protocol split in two: Transmission Control Protocol & Internet Protocol. The first TCP/IP v4 draft was published in 1978 but 3 years more passed until the draft became a standard.

On 1st of January 1983, also known as flag day, the ARPAnet switched from NCP to TCP/IP.

4.3 billion addresses will be enough

One of the novelties that TCP/IP introduced was 32-bit addresses. Vinton Cerf has often taken blame for that decision. But 32-bit addresses seemed very reasonable back in the 70s. In those days, the world’s population was 4.5 billion people and the personal computing revolution hadn’t started yet. Upgrading address space to 16-bit seemed too little and something bigger than 32-bit (4.3 billion addresses) unreasonable and unjustified.

In 1981, another TCP/IP network was created: CSnet (Computer Science Network) funded by the National Science Foundation. In 1983, Darpa decided to split the ARPAnet in two: a public ARPAnet and MILnet. Finally in 1985, NSF founded another network, NSFnet (National Science Foundation Network).

NSFnet was the most popular TCP/IP network in the 80s and eventually became the primarily backbone of the Internet at that time. By the end of the decade, the Internet was composed by almost 1000 networks (RFC 1118 “The Hitchhikers Guide to the Internet”) and had 3 million users approximately. ARPAnet ceased its operations in 1990, while CSnet followed in 1991.

The first concerns about the scalability of the Internet appeared in the early 90s, even before the Web was invented. RFC 1287 (“Towards the Future Internet Architecture”) is the first RFC to discuss the IP address space exhaustion problem.

One of the first measures to simplify the management of the Internet was the creation of RIRs or Regional Internet Registries in 1992. Before that, the global IP address registry was managed by a single organization, the IANA (Internet Assigned Numbers Authority). Each region was allocated a range of IP addresses. The regions have evolved over time. Today there are 5 RIRs:

  • AFRINIC (Africa).
  • APNIC (Asia-Pacific).
  • ARIN (Canada, many Caribbean and North Atlantic islands, and the United States).
  • LACNIC (Latin America and the Caribbean)
  • RIPE NCC (Europe, Middle East, and Parts of Central Asia).

The glorious 90s: the Internet explodes

The World Wide Web debuted in the early 90s leading to an exponential growth of the Internet. But even before that, there were already concerns its scalability.

The IETF created the ROAD WG (Routing and Addressing Working Group) to come up with proposals which could help to solve this problem. Some of the proposed solutions were:

  • RFC 1519: “Classless Inter-Domain Routing” (September 1993).
  • RFC 1597: “Address Allocation for Private Internets” (March 1994).
  • RFC 1631: “The IP Network Address Translator (NAT)” (May 1994).

RFC 791 (“Internet Protocol”) defines an IP address as:

Addresses are fixed length of four octects (32 bits). An address begins with a network number, followed by local address (called the “rest” field)

It also defines 3 classes of network addresses:

There are three formats or classes of internet addresses: in class a, the high order bit is zero, the next 7 bits are the network, and the last 24 bits are the local address; in class b, the high order two bits are one-zero, the next 14 bits are the network and the last 16 bits are the local address; in class c, the high order three bits are one-one-zero, the next 21 bits are the network and the last 8 bits are the local address.

Summarizing:

Class Leading bits Start Address End Address Network field Rest field
A 0 0.0.0.0 127.255.255.255 8 bits 24 bits
B 10 128.0.0.0 191.255.255.255 16 bits 16 bits
C 110 192.0.0.0 223.255.255.255 24 bits 8 bits

This scheme is known as classful network.

Class Inter-Domain Routing defines a variable-length network field for IP addresses which doesn’t depend on its class. This scheme allows two things:

  • To divide a network address into subnetworks, which leads to a more efficient use of the address space.
  • To group networks into supernetworks, which reduces the number of entries in the routing tables.

This latter issue was the main motivation for the creation of CIDR. Before that, a routing table had to contain one entry per network. For instance:

Network address Gateway
193.1.255.0 1.2.3.4
193.1.254.0 1.2.3.4

Since 193.1.255.0 and 193.254.0 are contiguous networks, an equivalent table could be represented as:

Network address Gateway
193.1.254.0/23 1.2.3.4

Class Inter-Domain Routing also introduced a new IP address notation known as CIDR notation in which an address is represented as a pair {IPv4 address/bit-mask}. Bit mask is a number between 0 and 32 that represents the number of contiguous bits used as a network mask. Address 193.1.254.0/23 is equivalent to 193.1.254.0/255.255.254.0.

Class Inter-Domain Routing highly helped to reduce the size of routing tables as well as optimize IP address use and simplify IP address allocation.

Another standard that enormously helped to mitigate IPv4 address exhaustion was RFC 1597 (“Address Allocation for Private Internets”).

On its conception, the Internet was designed as an peer-to-peer network where every host was addressable from any other host. Hosts inside private networks that only needed to communicate with other hosts within the same network over TCP/IP were also addressable from the Internet. RFC 1597 explains:

With the proliferation of TCP/IP technology worldwide, including outside the Internet itself, an increasing number of non-connected enterprises use this technology and its addressing capabilities for sole intra-enterprise communications, without any intention to ever directly connect to other enterprises or the Internet itself. The current practice is to assign globally unique addresses to all hosts that use TCP/IP. There is a growing concern that the finite IP address space might become exhausted.

The standard proposed the reservation of 3 blocks, each per network class, for private addresses. Hosts using private addresses are not reachable from the Internet, but can communicate to other peers inside the same intranet.

Class Start Address End Address Total IP addresses
A 10.0.0.0 10.255.255.255 16,777,216
B 172.16.0.0 172.31.255.255 1,048,576
C 192.168.0.0 192.168.255.255 65,536

The last standard that highly reduced IP address exhaustion was RFC 1631 (“The IP Network Address Translator (NAT)”).

NAT maps one IP space realm to another IP space realm. That enables a host using a private address, thus non-routable in the Internet, to borrow the address of another host which indeed has a public address assigned. That allows the private host to be addressable in the Internet.

The original NAT proposal came with some experimental implementations that proved it successful and made it to be adopted very quickly. On the down side, NAT broke the original design of the Internet as a peer-to-peer network.

Lastly, I think is also worth to mention DHCP (RFC 1541, “Dynamic Host Configuration Protocol”). DHCP has its origins in BOOTP, which at the same time is an evolution of RARP (RFC 903 “A Reverse Address Resolution Protocol”, June 1984). DHCP didn’t come out as the result of ROAD WG deliverance, but when used by ISPs it has greatly helped to optimize public address usage.

The birth of IPv6

In addition to the mitigation efforts commented above, during the early 90s the IETF also started to evaluate whether to develop a new version of IP which could definitely solve the address space problem. With that goal in mind, the IETF created the IPng area (Internet Protocol Next Generation area).

In 1994, the IPng area came up with RFC 1752 (“The Recommendation for the IP Next Generation Protocol”), which encouraged the development of a new successor to IPv4.

The IPng area also created a new working group called ALE (Address Life Expectation). The goal of this group was to determine the expected lifetime for IPv4. If IPv4 address space was estimated to last many years more, the new version of the Internet Protocol could feature new functionalities. On the other hand, if lifetime was estimated to be short, only the address space exhaustion problem could be handled. The working group estimated IPv4 address space exhaustion would happen sometime between 2005 and 2011.

That very same year of 1994, the Internet Engineering Steering Group approved the IPv6 recommendation by the IPng area and drafted a proposed standard. In 1995, RFC 1883 (“Internet Protocol, Version (IPv6)”) was published. After several iterations over the proposed standard, an updated document was published in December 1998 (RFC 2460).

Note: Version 5 of the IP protocol was used during the development of an experimental protocol called Internet Stream Protocol, back in 1979. There was a second review of this protocol in 1995 (RFC 1819, “Internet Stream Protocol Version 2 (ST2)”), but the protocol never gained traction. To avoid potential confusions, the IETF preferred to skip this number and pick Version 6 as the successor of IPv4.

In the years that followed many hardware vendors and operating system developers began to add and implement support for IPv6 in their products. A first alpha version of IPv6 was implemented in the Linux kernel as early as 1996, although it remained in experimental status until 2005. In 1996 too, a testbed backbone for IPv6, called 6bone, was deployed. The original mission of this backbone was to establish a network to foster the development, testing, and deployment of IPv6. The backbone ceased its operations in 2006. Another important milestone occurred in 2008 when the IANA added AAAA records for the IPv6 addresses of 6 root name servers. That made possible to resolved domain names using only IPv6.

The decade after the publication of RFC 2460 served as a period of time for development, testing and refinement of IPv6, as well as adaptation of existing products. Originally, the IETF estimated that massive adoption of the new protocol would happen around 2005, although that never happened. Part of the reason was that many of the new functionalities that IPv6 featured, for instance IPSec, were back-ported one way or another to IPv4, so the actual need to make the transition was less urgent and NAT was working pretty well. The only feature which could not be back-ported was the increase of available address space…

The day the world run out of IPv4 addresses

On 31st January 2011, the IANA allocated their two remaining top-level address blocks to APNIC. APNIC run out of IPv4 public addresses some months later. RIPE followed the next year, as well as LACNIC in 2014 and ARIN in 2015. Today, the only RIR with IPv4 public addresses available is AFRINIC, but it won’t last long, only until 2018.

The event didn’t catch anyone off-guard, in fact the dates were in line to the estimation of the ALU working group. Perhaps witnessing the actual IPv4 address depletion in 2011, served as a wake-up a call to accelerate IPv6 adoption worldwide. Since 2010, adoption has been constantly increasing and in some cases doubling every year. Today, IPv6 traffic represents a total of 18% Internet worldwide traffic according to Google. But that’s the world’s average, truth is IPv6 adoption worldwide is uneven across countries. Belgium ranks first with 47% traffic as IPv6, while there are many other countries, such as Italy or Spain, where IPv6 roll-out hasn’t even started yet.

And this all for now. In a next article I will cover IPv6 adoption and what strategies are ISPs implementing to complete the transition from an already exhausted IPv4 address space to IPv6.


igalia networking