At an IETF gathering last November, a first‑time attendee watched the debut of TCP BBR, a congestion‑control algorithm that drew enthusiastic yet wary reactions. The same meeting featured a flood of talks extolling IPv6 as the long‑awaited successor to IPv4, with many participants convinced the transition was imminent. Meanwhile, the IPv6 track showcased deployment hurdles, from NAT incompatibilities to sparse tooling.

The author traced IPv6’s tangled origins to early networking choices. Telephone circuits originally required no addresses, and when time‑division multiplexing introduced virtual circuits, IP addresses emerged solely for routing between point‑to‑point links. LANs later adopted bus topologies with their own 8‑bit layer‑2 schemes, such as ARCnet, before Ethernet’s 48‑bit MAC scheme became universal. Universities then added bridging and spanning‑tree to avoid loops.

Because Ethernet MACs are factory‑assigned and non‑hierarchical, large networks resorted to hardware bridges that learned which physical segment each address lived on. This bridging layer, while fast, produced broadcast storms and opaque routing tables that engineers could not easily debug. Thus IPv6 stalls at the edge.