IP address

• 1. Introduction

  ›

  An IP address is a numerical label assigned to every device that connects to a computer network using the Internet Protocol. Without it, there would be no way to identify your laptop, your phone, or any other machine on the network, and no way to find a path to send information across the internet. Two examples of what these labels look like are familiar to anyone who has poked around a router settings page: sequences of numbers separated by dots. But behind that simple-looking format lies a decades-long struggle with a surprisingly finite resource: the address space itself. When the first standalone specification for the Internet Protocol, known as IPv4, went into operational use in 1983, few could have anticipated that its supply of addresses would one day run critically short. That shortage, and the engineering effort to solve it, is the thread that runs through the story of how the internet came to look the way it does today.

• 2. Two Functions, One Label

  ›

  The header of every IP packet carries two pieces of information: the address of the sending host and the address of the destination host. Those two addresses do a great deal of work. The Internet Protocol's designers articulated the role of an address with a crisp three-part description: a name indicates what we seek, an address indicates where it is, and a route indicates how to get there. So the IP address sits at the middle of that hierarchy. It is not the name of a device in a human sense, and it is not a full set of routing instructions. It is the location marker, the coordinate that lets a packet find its way.

  In practice, identifying a host and providing its location in the network are two faces of the same act. A single device can hold more than one unicast address, and a network interface is what gets the address rather than the physical machine as a whole. That distinction matters in environments where one machine hosts multiple network interfaces, each needing its own identity on the network.

• 3. IPv4 and the Exhaustion Problem

  ›

  IPv4 defines an address as a 32-bit number, a constraint that limits the total address space to roughly four billion unique values. Some of those values are reserved for special purposes: approximately 18 million are set aside for private networks, and approximately 270 million are reserved for multicast addressing, leaving the usable pool of public addresses considerably smaller than that four-billion ceiling. Early network designers did not set out with a plan to partition the address space carefully. In the earliest stage of the Internet Protocol, the network number was always the highest-order octet, a scheme that allowed for only 256 networks before running into a wall.

  In 1981 the addressing specification was revised with the introduction of classful network architecture, dividing addresses into three classes for universal unicast use: A, B, and C. Each class used successively more octets for the network identifier, trading off the number of possible networks against the number of hosts each could contain. Class A had 128 possible networks, each capable of holding more than 16 million hosts; Class C had more than 2 million networks, but each held only 256 addresses.

  Classful design served the internet during its startup years, but by the early 1990s rapid growth had made the system unworkable. In 1993 the class system was replaced with Classless Inter-Domain Routing, or CIDR, which uses variable-length subnet masking to allocate and route based on arbitrary-length prefixes. CIDR slowed the rate at which address blocks were consumed, but it could not stop the clock. By the 2010s, IPv4 address exhaustion had become a lived reality.

• 4. The IPv6 Answer

  ›

  By the early 1990s the Internet Engineering Task Force had begun exploring ways to expand the internet's addressing capacity. The result of that effort was named Internet Protocol Version 6, or IPv6, formally designated in 1995. The gap in the version numbering between 4 and 6 has a specific explanation: version 5 had already been assigned to the experimental Internet Stream Protocol in 1979, a project that never carried the IPv5 label publicly. Versions 1 through 9 were all defined at various points; only 4 and 6 ever reached widespread deployment.

  IPv6 expands the address from 32 bits to 128 bits, a change that provides approximately 3.403 undecillion addresses. The designers intended more than just a larger number. IPv6 allows more efficient aggregation of routing prefixes, which produces slower growth of routing tables inside routers. The smallest possible individual allocation in IPv6 is a subnet capable of hosting 2 to the 64th power devices, which equals the square of the size of the entire IPv4 internet.

  IPv6 technology moved through testing until the mid-2000s, when commercial production deployment began. All modern desktop and enterprise server operating systems now include native IPv6 support, though deployment lags in residential routers, voice-over-IP equipment, and some networking hardware. Both versions remain in simultaneous use, and the generic term IP address still refers by default to IPv4, given its longer history.

• 5. Private Networks and Address Translation

  ›

  Early internet design assumed global end-to-end connectivity, meaning every device would carry a globally unique IP address. That model held as long as the internet remained a small academic and government network. As private industrial and corporate networks grew, it became clear that machines communicating only within a factory or an office building had no need of globally routable addresses. Three non-overlapping ranges of IPv4 addresses are now reserved for private use. These addresses are not routed on the public internet, so any organization can use them freely without coordinating with an IP address registry.

  Private networks typically reach the public internet through a mechanism called network address translation, or NAT. A NAT device maps different IP addresses on the private network to different TCP or UDP port numbers on the public network. The internal computers appear to share a single public address. In residential settings, this function is handled by the home router, which holds one public-facing address from the ISP while assigning private addresses to every device inside the home. Most home routers automatically use a default address range running from a standard private start address through a fixed endpoint within that same block.

  In IPv6 the private-address concept carries a different name: unique local addresses, assigned the routing prefix. These addresses include a 40-bit pseudorandom number designed to minimize the chance of collisions if two sites merge or packets are misdirected. An earlier IPv6 scheme called site-local addresses was abandoned because the definition of a site remained unclear and the policy created routing ambiguities.

• 6. Assignment: Static, Dynamic, and Sticky

  ›

  Dynamic Host Configuration Protocol, widely known as DHCP, is the technology most commonly used to assign IP addresses automatically. When a device joins a network, DHCP hands it an address along with a lease. The lease carries an expiration period; if the device does not renew before the lease expires, the address can be reassigned to another device. Some DHCP servers attempt to return the same address to the same device each time, using the device's hardware MAC address as the key. A network administrator can configure a DHCP server to enforce that pairing deliberately.

  Before DHCP, an earlier protocol called Bootstrap Protocol served a similar purpose. Dialup connections and some broadband networks use dynamic address features of the Point-to-Point Protocol instead. Routers and mail servers, by contrast, are almost always given static addresses, configured manually and held indefinitely, because other systems need to reach them at a known, unchanging location.

  Sticky is an informal industry term for a dynamic address that changes rarely. An ISP serving a home network may configure its DHCP service to maximize the chance that the same address is returned to the same router every time. That stability is a practical convenience, but it carries no guarantee. A sticky address can change; a static address is only changed on purpose. When no static or dynamic configuration is available, an operating system can assign a link-local address through stateless address autoconfiguration. Microsoft's implementation of this fallback, called Automatic Private IP Addressing, first appeared in Windows 98 and later became a formal IETF standard in May 2005.

• 7. Governance and Legal Status

  ›

  The global supply of IP addresses is managed by the Internet Assigned Numbers Authority, known as IANA, together with five regional Internet registries. IANA distributes blocks of addresses to those regional bodies, which in turn pass them to local Internet registries, including ISPs and large institutions. The structure creates a clear chain of custody for every address block in use.

  The legal status of an IP address has become a live question. In March 2024, the Supreme Court of Canada ruled that IP addresses are protected private information under the Canadian Charter of Rights and Freedoms, requiring police to obtain a warrant before accessing them. The European Commission treats IP addresses as personal data under the General Data Protection Regulation. In the United States, the California Consumer Privacy Act protects IP addresses only when they can be linked to a particular consumer or household; an address that cannot be tied to an individual is not protected under that law. The divergence across jurisdictions reflects ongoing disagreement about whether a number that identifies a network connection is, in any meaningful sense, a number that identifies a person.