Classful vs classless addressing

The classful system: three fixed sizes

Before 1993, every IPv4 address belonged to a class determined by its first few bits, and each class had one fixed mask. Class A covered 1.0.0.0 through 126.255.255.255 with a /8 mask, 16,777,214 usable hosts per network. Class B covered 128.0.0.0 through 191.255.255.255 with a /16 mask, 65,534 usable hosts. Class C covered 192.0.0.0 through 223.255.255.255 with a /24 mask, 254 usable hosts.

Two more classes sat outside normal host allocation: Class D (224.0.0.0 to 239.255.255.255) for multicast, and Class E (240.0.0.0 to 255.255.255.255), reserved for experimental use. Neither is handed out to organisations as a network.

The class was baked into the address itself: a router could tell a network's size just by looking at its first octet, with no separate mask needed alongside it.

Why classful addressing broke down

The three fixed sizes did not match how organisations actually grew. A company with 2,000 devices needed more than a Class C's 254 hosts, so it was handed an entire Class B of 65,536 addresses, using only about 3 percent of it. Multiply that across the fast-growing internet of the early 1990s and the usable IPv4 address space was on track to run out years sooner than it needed to.

It also bloated routing tables: every class-sized allocation needed its own routing table entry, with no way to summarise several small allocations into one announcement.

What classless addressing (CIDR) changed

CIDR, introduced in 1993, detached the network boundary from the address's leading bits. Every address is instead written with an explicit prefix length, like 192.0.2.0/22, and that prefix can be anything from /0 to /32, not just /8, /16, or /24.

That flexibility means an organisation needing 2,000 addresses can get a /21 (2,046 usable hosts) instead of an entire Class B, and a provider can summarise sixteen contiguous /24s into a single /20 route. See what is CIDR for the full mechanics of prefix notation.

Is anything still classful today?

The address ranges themselves are unchanged; RFC 1918 still describes 10.0.0.0/8, 172.16.0.0/12, and 192.168.0.0/16 along their historical class boundaries, out of convention rather than because routers enforce classes. See private vs public IP addresses for what those ranges are used for.

The one place classful thinking survives is habit: people still call a /24 a class C network in conversation, and the old default masks (255.0.0.0, 255.255.0.0, 255.255.255.0) remain the fallback a device assumes if you configure an address with no mask at all. Every router and operating system, though, works purely in classless CIDR terms.