Google Hits 50% IPv6

Original Article ↗ Hacker News Discussion ↗

Google’s report that over half of its user traffic now reaches it via IPv6 has reignited debate about the protocol’s real‑world adoption and benefits. Commenters highlight that much of the growth comes from mobile and newer ISPs, while many legacy providers, corporate networks, and key services (like GitHub and some AWS features) remain IPv4‑centric, forcing costly dual‑stack setups and CGNAT workarounds. Opinions split between those who see IPv6 as technically simpler and essential for scaling and decentralization, and those who view it as overengineered, unevenly implemented, and unlikely ever to fully replace IPv4.

Adoption and traffic share

  • Google reports 50% of eyeballs on IPv6; APNIC and Cloudflare see lower global capability (42–59%) depending on methodology.
  • Adoption is uneven: France and India are very high; some ISPs (e.g., in UK, NL, PT, AR) still lack or throttle IPv6 despite years of promises.
  • Mobile networks are a major driver; many are IPv6-first or IPv6-only with IPv4 via translation.
  • Some expect further growth; others think IPv6 will stall well below 100% and IPv4 will never be turned off.

Performance and reliability

  • Pro‑IPv6: often lower latency than IPv4 behind CGNAT; big content (YouTube, major social sites) commonly served over IPv6; many users unknowingly rely on it daily.
  • Anti/concerned: numerous anecdotes of IPv6 endpoints misconfigured, odd BGP paths, broken mirrors, and “random” failures in CI pipelines; some users fix issues by disabling IPv6.
  • Dual stack complexity and “happy eyeballs” behavior sometimes mask IPv6 problems or favor IPv4.

Security, NAT, and privacy

  • Ongoing debate: NAT vs stateful firewall. Several argue NAT is not real security, just incidental blocking; others note it “passes the grandma test” and hides vulnerable devices.
  • Home routers typically default‑deny inbound IPv6, recreating NAT‑like safety; some CPEs have broken or non‑configurable IPv6 firewalls.
  • Old concerns about embedding MACs in addresses are mostly mitigated by privacy extensions, now default on major OSes, though not all distros.

Incentives and deployment barriers

  • Many enterprises and SaaS providers see no customer demand; dual‑stack re‑architecture is hard and risky, so they delay.
  • Some cloud services (notably parts of AWS, GitHub, Discord) still lack full IPv6, forcing NAT64/464Xlat or tunnels, especially for IPv6‑only hosts.
  • ISPs with large IPv4 pools and CGNAT see little business upside; some even charge premiums for static IPv4.

Design debates and alternatives

  • Significant hostility toward IPv6’s perceived overengineering (128‑bit hex addresses, SLAAC, NDP, notation), with recurring “just add more bits to IPv4 / IPv5” ideas.
  • Others counter that any address expansion necessarily implies a new protocol and dual‑stack/translation; IPv6 also simplifies headers, removes router fragmentation, and supports probabilistic autoconfiguration.

P2P, gaming, and centralization

  • Supporters argue IPv6 is key to restoring end‑to‑end connectivity for P2P, gaming, and self‑hosting, and to reducing Internet centralization.
  • Skeptics note that stateful firewalls, DDoS and privacy issues still push many real‑world apps toward centralized servers or relays, even with IPv6.