Les Classes d'Adresses IP Expliquées

The IP address classification system was established in 1981 as part of the original Internet Protocol specification (RFC 791). It defined five address classes — A through E — based on the leading bits of the first octet. This classful architecture was designed to simplify routing by providing a fixed boundary between the network and host portions of an address, eliminating the need to transmit subnet mask information alongside routing updates.

Each class was intended for a different scale of network. Class A accommodated a small number of very large networks, Class B served medium-sized organizations, and Class C supported the vast majority of smaller networks. While this system worked in the early internet era, the rigid boundaries between classes eventually led to significant address waste and contributed to the accelerated depletion of available IPv4 space.

Class B addresses begin with binary bits 10 in the first octet, covering the range from 128.0.0.0 to 191.255.255.255. With a 16-bit network prefix and 16 bits for hosts, each Class B network supports up to 65,534 usable host addresses. This middle-ground allocation was designed for medium to large organizations such as universities, large corporations, and government agencies.

A total of 16,384 Class B networks are available. During the 1990s internet boom, Class B allocations were heavily requested because Class C (254 hosts) was too small for many organizations while Class A was far too large. This led to rapid depletion of the Class B space and was one of the key factors driving the adoption of CIDR, which allowed organizations to receive precisely the amount of address space they needed rather than being forced into class-sized allocations.

Class D addresses (224.0.0.0 to 239.255.255.255) are reserved exclusively for IP multicast. Unlike unicast addresses that identify a single destination, multicast addresses allow a single packet to be delivered to multiple recipients simultaneously. This is used for applications like video streaming, online gaming, and network protocols such as OSPF (224.0.0.5) and RIP (224.0.0.9). Class D addresses are not assigned to individual hosts or organizations.

Class E addresses (240.0.0.0 to 255.255.255.255) were designated as experimental and reserved for future use. Despite representing approximately 268 million addresses, they have never been made available for general allocation. The address 255.255.255.255 serves as the limited broadcast address. Various proposals have suggested releasing the Class E space to help alleviate IPv4 exhaustion, but compatibility concerns with legacy network equipment have prevented this from happening.

The classful addressing system suffered from a fundamental inefficiency: the gap between class sizes was too large. An organization needing 300 addresses couldn't use a Class C (254 hosts) and would be assigned a Class B (65,534 hosts), wasting over 65,000 addresses. This "all or nothing" approach led to rapid exhaustion of the Class B space and enormous waste across the internet.

Classless Inter-Domain Routing (CIDR), standardized in 1993 through RFC 1517–1520, eliminated the rigid class boundaries entirely. Under CIDR, any prefix length from /0 to /32 can be used, allowing allocations like /22 (1,024 addresses) or /20 (4,096 addresses) that don't correspond to any traditional class. This flexibility dramatically improved address utilization and slowed the rate of IPv4 exhaustion.

Today, classful networking is considered a historical concept taught for educational purposes. All modern routing protocols (BGP, OSPF, EIGRP) are classless and support VLSM natively. However, understanding the original class system remains valuable because many network conventions — such as the default subnet masks, private address ranges, and the /24 as a minimum routable prefix — trace their origins directly to the classful architecture.