we ran OUT of IP Addresses!!

The world’s IPv4 address space didn’t run out because the math was wrong—it ran out because early IP allocation rules were built for a much smaller internet and then handed out too aggressively. IPv4 offers about 4.3 billion possible addresses (2^32 = 4,294,967,296), but the system’s original “classful” design wasted large blocks and reserved entire ranges, leaving far fewer usable addresses than the headline number suggests.

In the classful scheme, IPv4 addresses were grouped into classes A, B, C, D, and E. Classes A–C came with default subnet masks that determined how many hosts could fit in each network and how many separate networks could exist. Class A locked in the first octet as the network portion (default mask 255.0.0.0), producing extremely host-heavy networks—about 16,777,214 usable addresses per network—while leaving only about 126 possible Class A networks. Class B used a default mask of 255.255.0.0, yielding roughly 65,534 hosts per network and about 16,382 Class B networks. Class C used 255.255.255.0, which created far smaller networks—254 usable hosts each—but enabled about 2,097,150 Class C networks.

The mismatch between expected and actual demand drove the damage. Large organizations were assigned huge address blocks “just in case,” because the internet’s explosive growth and the number of devices needing addresses weren’t anticipated. The result: many networks were allocated far more addresses than they could ever use. Even when the system allowed splitting large allocations into smaller subnets, the initial “give away too many” approach left the overall pool constrained.

Two additional problems reduced usable space further. First, Classes D and E weren’t meant for ordinary unicast addressing: Class D is reserved for multicast, and Class E is experimental/reserved. Second, a chunk of what looked like normal Class A space was missing from the usable range—127.0.0.0/8—because it’s reserved for loopback. That’s why 127.0.0.1 always points back to the same device, a built-in mechanism for testing network connectivity. The transcript emphasizes how pinging different 127.x.y.z addresses still “works” because they all route to the local machine, making the large reserved block feel like wasted capacity.

By the end, the core takeaway is that IPv4’s apparent 4.3 billion addresses were undermined by class-based allocation, reserved ranges, and early mismanagement. The promised next step is a “bandaid” fix—an approach that eventually became the practical path forward when the address pool tightened.