Do You Need IPv4 Anymore?

Original Article ↗ Hacker News Discussion ↗

Do We Still Need IPv4?

  • Many argue “yes” in practice: too many services, tools, and platforms (notably several AWS services, some major sites) are still IPv4-only or incomplete on IPv6.
  • Others say end-user devices don’t strictly need IPv4 if the ISP or cloud edge provides NAT64/translation; IPv4 can be confined to the provider side.
  • Some self-hosters report that lack of IPv4 (or CGNAT) blocks P2P activities, game servers, port scanning, and generally “participating” in the internet, not just consuming it.

IPv6 Deployment Reality (ISPs, Mobile, Cloud)

  • Experiences vary widely:
    • Some ISPs are IPv4-only; others provide solid dual-stack, /56–/48 prefixes, and good DHCPv6-PD.
    • Some regions with Spectrum or Virgin report buggy/partial IPv6 (flaky connectivity, single-usable address from a larger prefix).
    • Comcast is cited as an example of strong IPv6 support; some users fail to use it due to router config.
  • Mobile: some networks are IPv6-only with translation (e.g., 464XLAT, DS-Lite), but others still hand out IPv4-only addresses (e.g., some Verizon cases).

Home Networking and Self‑Hosting

  • IPv6 can simplify self-hosting when ISPs block port forwarding on IPv4; public IPv6 on LAN plus DNS proxying works well for some.
  • Others find local IPv6 “overly complex” and prefer IPv4+NAT internally, seeing NAT as a feature for simplicity and implicit protection.

NAT, Firewalls, and Security

  • Strong disagreement:
    • One side views NAT as essential security and topology hiding, sometimes even mandated (e.g., compliance wanting hidden internal structure).
    • The other side insists security comes from stateful firewalls and default-deny rules, not address translation; IPv6 can emulate NAT-like protection without rewriting addresses.
  • Concerns about IPv6 “glass house” exposure are countered by pointing out default-deny CPE behavior, privacy/stable non-MAC-based addresses, and large address space making scanning hard.

Operations, Addressing, and Internal Networks

  • Pro‑IPv6 operators cite: easier large-scale addressing (ephemeral VMs, containers), avoiding RFC1918 collisions (e.g., mergers, VPC peering), and simple /64-per-subnet design.
  • Skeptics say most environments don’t hit IPv4 limits; renumbering or careful subnet planning is easier than a stack transition.

Future and IPv4-as-a-Service

  • IPv4 is expected to persist but become increasingly “legacy” and potentially costly; some foresee IPv4 offered as an on-demand service atop IPv6-only cores.
  • Others remain unconvinced and plan to “ignore IPv6” until lack of IPv4 or overloaded CGNAT forces a change.