The world in which IPv6 was a good design
9 hours ago
- #Legacy Systems
- #IPv6
- #Networking
- The author attended an IETF meeting to understand IPv6's complexity, noting IPv6 was designed to simplify networking by eliminating legacy issues like bus networks, MAC addresses, ARP, and DHCP.
- Historically, networking evolved from physical circuits to LANs with ethernet and MAC addresses, leading to bridging and routing complexities, with layer 2 (bridging) and layer 3 (routing) becoming intertwined and messy.
- IPv6 envisioned a world without bus networks, broadcasts, MAC addresses, ARP, or DHCP, using point-to-point links and multicast, but legacy systems like IPv4 and ethernet prevented this clean slate, forcing continued use of layer 2 bridging.
- Mobile IP (roaming) is a key issue: current IP routing can't handle mobility without layer 2 bridging, as changing IP addresses breaks connections. Solutions like LTE use tunneling and bridging, adding latency and complexity.
- The flaw in mobile IP lies in the 4-tuple (source/destination IP and ports) for session identification. If sessions used layer 4 identifiers (e.g., large unique IDs) instead, roaming could work without bridging, as seen in protocols like QUIC or MinimaLT.
- Despite IPv6's goals, legacy systems persist, making it impossible to remove layers like ethernet addresses or DHCP. The author suggests future protocols might enable true mobility and elegance, but widespread adoption is challenging.