IP Address Reference Guide

An IP address (Internet Protocol address) is a numerical label assigned to every device that participates in a computer network using the Internet Protocol for communication. In IPv4 — the version that has run the internet since the early 1980s — an IP address is a 32-bit unsigned integer. It uniquely identifies a network interface at the network layer (Layer 3 in the OSI model), allowing packets to be routed from a source to a destination across any number of intermediate networks.

Because 32-bit binary strings like 11000000101010000000000000000001 are impractical for humans, IPv4 addresses are written in dotted-decimal notation: the 32 bits are split into four 8-bit groups (octets), each converted to decimal and separated by dots. The example above becomes 192.168.0.1.

An IP address serves two functions: identifying the host (or more precisely, the network interface) and providing location information that routing protocols use to forward packets toward the destination. These two functions are sometimes in tension — a motivation for the design of IPv6, which separates them more cleanly through its address architecture.

Every IPv4 address is exactly 32 bits — four octets of 8 bits each, giving a theoretical address space of 2³² = 4,294,967,296 addresses. In practice, large portions of this space are reserved, private, or otherwise not publicly routable, leaving considerably fewer usable public addresses.

Network portion vs host portion

Every IP address is divided into two logical parts by a subnet mask (or CIDR prefix length):

• Network portion — the leading bits, identical for all hosts on the same subnet. Determined by the subnet mask.
• Host portion — the trailing bits, unique per device within the subnet.

For example, in the address 192.168.10.45/24, the first 24 bits (192.168.10) are the network, and the last 8 bits (.45) are the host identifier within that network.

Why 0 and 255 are special within a subnet

Within any subnet, two addresses are not assignable to hosts by convention:

• Network address — the address with all host bits set to 0 (e.g., 192.168.10.0 in a /24). Identifies the subnet itself and is used in routing table entries.
• Broadcast address — the address with all host bits set to 1 (e.g., 192.168.10.255 in a /24). Packets sent to this address are delivered to every host on the subnet.

A /24 therefore has 256 total addresses, 254 usable host addresses, one network address, and one broadcast address. This logic applies at any prefix length: a /30 has 4 addresses, 2 usable — the minimal building block for a point-to-point link.

Binary representation

Understanding the binary form is essential for subnetting. Each octet ranges from 0 to 255 (2⁸ − 1). The subnet mask is a contiguous run of 1-bits followed by 0-bits. A /24 mask in binary is:

The bitwise AND of an IP address with its subnet mask always produces the network address. This is the fundamental operation that routers and hosts use to determine whether a destination is on-link (same subnet) or off-link (requires routing through a gateway).

Before CIDR was standardized in 1993, the internet used classful routing, where an address’s class was determined by its leading bits. The class implied a fixed network/host boundary without needing a separate mask.

Class A networks handed enormous blocks (16 million+ addresses) to a small number of organizations — early internet players like MIT, Apple, and the US Department of Defense each received Class A allocations. This proved profoundly wasteful. Class B blocks went to universities and corporations. Class C allocations (254 hosts each) were assigned to smaller organizations but became inconvenient when an organization needed more than 254 hosts and had to manage multiple non-contiguous Class C blocks.

Class D (224–239) is reserved for IP multicast. Multicast allows a single packet to be delivered efficiently to a group of receivers — used by protocols like OSPF (which uses 224.0.0.5 and 224.0.0.6), PIM, and streaming applications.

Class E (240–255) was designated “reserved for future use” by DARPA. It has never been allocated for general internet use. Some proposals to repurpose this space as public unicast addresses have been floated during the IPv4 exhaustion crisis, but none have achieved broad adoption because many network stacks and firewalls still drop Class E packets.

Classful routing is largely obsolete — replaced by CIDR in RFC 1519 (1993) and now RFC 4632. However, the class boundaries still appear in vendor documentation, firewall rule names, legacy tools, and conversations, so understanding them remains useful.

CIDR (Classless Inter-Domain Routing, RFC 4632) replaced classful routing by allowing the network/host boundary to fall at any bit position, not just at octet boundaries. A CIDR address is written as an IP address followed by a slash and a prefix length: 10.0.0.0/8.

Prefix length and subnet mask

The prefix length (the number after the slash) is the count of leading 1-bits in the subnet mask. Common masks:

Subnetting vs supernetting

Subnetting splits a larger block into smaller ones by lengthening the prefix. Taking a /24 and creating two /25s divides the address space into two equal halves of 128 addresses each — useful for separating VLANs or departments.

Supernetting (also called route aggregation or summarization) combines multiple contiguous blocks into a single shorter-prefix entry. Combining eight /24 networks into one /21 reduces the number of routing table entries that upstream routers must maintain. This was the primary motivation for CIDR: the 1990s internet routing table was growing explosively because every organization had separately announced Class C allocations. CIDR aggregation dramatically slowed that growth.

RFC 1918 (Address Allocation for Private Internets, 1996) defines three IPv4 address blocks reserved for use within private networks. These addresses are non-routable on the public internet — routers operated by ISPs and internet exchanges are configured to drop packets sourced from or destined to these ranges. They exist to allow organizations to build internal networks without consuming globally unique address space.

The mechanism that makes private addressing practical at internet scale is NAT (Network Address Translation). A NAT device (typically a home router or corporate firewall) maintains a translation table that maps internal private addresses to a single public IP for outbound connections, and demultiplexes return traffic back to the correct internal host. Without NAT, the IPv4 address pool would have been exhausted far sooner.

Important: RFC 1918 addresses can be reused across any number of separate private networks, because they are never advertised publicly. This is why your home router at 192.168.1.1 and every other home router at the same address don’t conflict — they live in isolated NAT bubbles.

As IPv4 addresses grew scarce, ISPs needed a way to share a single public IP among many customers without running NAT on customer-owned equipment. The solution, standardized in RFC 6598 (2012), is CGNAT (Carrier-Grade NAT), also called Large-Scale NAT (LSN) or shared address space.

RFC 6598 allocates 100.64.0.0/10 (covering 100.64.0.0 through 100.127.255.255, approximately 4 million addresses) specifically for use between an ISP’s CGNAT device and customer premises equipment. These addresses are:

• Not routable on the public internet (ISPs must not forward them externally)
• Not part of RFC 1918 (deliberately a separate range to avoid conflicts with customer-side NAT)
• Not visible to the customer’s devices — the customer’s router receives a 100.64.x.x address from the ISP, but typically re-NATs it again before handing out RFC 1918 addresses to home devices

Jamaica context. Flow and Digicel both deploy CGNAT on their residential broadband networks. If checkmiip.com shows a public IP that traces back to a Flow or Digicel ASN, that is the shared public IP your CGNAT device is using — not an address uniquely assigned to your household. If your router’s WAN interface shows a 100.64.x.x address (or a 10.x.x.x address assigned by the ISP), you are behind CGNAT.

The following table covers all IPv4 ranges that are not available for general public unicast allocation. Sources: IANA Special-Purpose Address Registry, individual RFCs.

A public IP address is any IPv4 address not included in a special-use or private range — one that is globally unique and reachable from anywhere on the internet. The IANA (Internet Assigned Numbers Authority) is responsible for the global allocation of IP address space. IANA does not assign addresses directly to end users; instead, it allocates large blocks to five Regional Internet Registries (RIRs):

• ARIN — North America
• LACNIC — Latin America and the Caribbean (including Jamaica)
• RIPE NCC — Europe, Middle East, and Central Asia
• APNIC — Asia-Pacific
• AFRINIC — Africa

RIRs in turn allocate address space to ISPs and large organizations within their regions, which assign smaller prefixes to end customers. This hierarchical model is what makes BGP route aggregation possible — each RIR’s allocations form contiguous supernets that can be announced with minimal routing table entries.

IPv4 exhaustion

The inevitable consequence of a 32-bit address space serving billions of devices:

• 3 February 2011 — IANA allocated its last five /8 blocks to the five RIRs and declared its free pool exhausted.
• September 2012 — RIPE NCC moved to its final /8 allocation policy (one /22 per new member).
• 2014–2015 — ARIN exhausted its free pool and moved to a waiting list.
• June 2020 — LACNIC exhausted Phase 4 of its exhaustion policy, making new allocations from LACNIC essentially unavailable for most organizations.

New IPv4 addresses are still available on the secondary market (brokers buy and sell address blocks), but prices have risen sharply — from under $10 per address in 2015 to over $50 per address by the mid-2020s.

For Jamaica and the Caribbean, this means ISPs like Flow and Digicel face real pressure on address supply. CGNAT is the primary mitigation for residential customers. IPv6 deployment — still patchy across Caribbean ISPs as of 2026 — is the only long-term solution.

IPv6 (RFC 8200) addresses the exhaustion problem by expanding the address space from 32 bits to 128 bits — approximately 3.4 × 10³⁸ addresses, or roughly 340 undecillion. To put that in perspective, there are enough IPv6 addresses to assign a /48 subnet (65,536 /64 subnets, each with 18 quintillion addresses) to every person on earth and still not run out.

IPv6 addresses are written in eight groups of four hexadecimal digits, separated by colons: 2001:0db8:85a3:0000:0000:8a2e:0370:7334. Leading zeros within a group can be omitted, and one contiguous run of all-zero groups can be compressed to ::: 2001:db8:85a3::8a2e:370:7334.

Key IPv6 special ranges

Unique local addresses (fc00::/7) behave like RFC 1918 private addresses — they are not routed on the public internet and can be used freely within an organization. The distinction from global unicast is that ULA addresses are randomly generated to be unique enough to avoid collisions if two private networks are ever merged.

Link-local addresses (fe80::/10) are automatically configured on every IPv6-enabled interface and are used for neighbor discovery and router advertisement — the IPv6 equivalents of ARP and DHCP discovery. They are scoped to a single network segment and are never routed.

Transition mechanisms. Because the IPv4 and IPv6 internets coexisted for years (and still do), several transition technologies were developed: dual-stack (running both protocols simultaneously), 6to4 (now deprecated — the 192.88.99.0/24 relay range in the special-use table above), Teredo, and NAT64/DNS64 which allows IPv6-only clients to reach IPv4-only servers.