IPv6 deployment models diverge for access vs corporate networks

Network operators adopting IPv6 face a critical choice between two transition models—IPv6-Only+IPv4aaS and IPv6-Mostly—each designed for distinct environments. A recent RIPE Labs article by Jordi Palet Martínez, a contributor to IETF’s v6ops working group, argues that conflating these approaches risks misconfigurations, unnecessary latency, and degraded user experience. The distinction hinges on whether the network is unmanaged (e.g., mobile or residential access) or managed (e.g., corporate LANs), with each model requiring different hardware, software, and operational trade-offs.

Access networks: IPv6-Only+IPv4aaS

For mobile and residential networks, IPv6-Only+IPv4aaS—particularly the 464XLAT mechanism—dominates. This model confines IPv6 to the access network’s WAN segment while preserving dual-stack functionality on customer premises (CPE LANs). A stateless NAT46 (CLAT) in the CPE translates IPv4 traffic locally, ensuring backward compatibility without requiring user intervention. The approach is standardized in RFCs 6877, 8585, and 8683, with 464XLAT as the sole mechanism supported in mobile networks.

Key advantages include public IPv4 address conservation and seamless operation for end users, regardless of device protocol support. However, the model assumes unmanaged networks where operators cannot enforce host-level configurations. Traffic between IPv4-only devices on the same LAN remains local, avoiding cloud-based translation latency. Palet Martínez notes that while 464XLAT is widely deployed in mobile networks, residential CPEs often lack CLAT support, citing complexity despite open-source implementations and existing OEM options like OpenWrt.

Corporate networks: IPv6-Mostly

IPv6-Mostly targets managed corporate networks, where administrators can enforce host-level policies. The model, outlined in IETF draft-ietf-v6ops-6mops, allows dual-stack hosts to disable IPv4 via DHCPv4 option 108, effectively behaving as IPv6-only on a per-host basis. This reduces private IPv4 address consumption, eliminates NAT44, and surfaces IPv6 connectivity issues masked by Happy Eyeballs’ fallback to IPv4. Stateful NAT64 (PLAT) or stateless NAT64 with Explicit Address Mappings (EAM) must be deployed to maintain access to IPv4-only destinations.

Unlike 464XLAT, IPv6-Mostly requires CLAT support on hosts, which has been progressively added to desktop (macOS, Linux, Windows 11) and mobile (iOS, Android) operating systems. However, mobile networks cannot use this model due to the absence of DHCPv4 on cellular interfaces. Corporate administrators must also configure mechanisms like PLAT or SIIT-DC to enable communication between IPv6-Mostly hosts and IPv4-only devices (e.g., printers, cameras), adding operational overhead.

Risks of misalignment

Applying IPv6-Mostly to unmanaged networks introduces significant drawbacks. Residential CPEs would need to maintain dual-stack WAN connectivity or integrate CLAT, PLAT, and user-friendly management interfaces—features rarely prioritized by consumer-grade hardware vendors. Traffic between IPv6-Mostly and IPv6-only devices would require cloud-based translation, increasing latency and bandwidth consumption. Conversely, using 464XLAT in corporate networks sacrifices the granular control and IPv4 address savings that IPv6-Mostly provides.

The article emphasizes that user satisfaction depends on selecting the right model for the network’s management scope. For operators, this means demanding CPEs with CLAT support for access networks and ensuring corporate networks adopt IPv6-Mostly protocols like DHCPv4 option 108 and NAT64. Palet Martínez advocates for market pressure or regulation to drive vendor adoption, noting that cost-driven procurement often overlooks long-term operational benefits.

What to watch

IETF’s v6ops working group is finalizing draft-palet-v6ops-ipv6-only to standardize terminology and use cases. Operators should monitor progress on draft-ietf-v6ops-6mops and draft-ietf-v6ops-claton, which define IPv6-Mostly and CLAT host requirements, respectively. Meanwhile, the gap between mobile/residential IPv6 adoption (approaching 75% globally) and corporate deployment (lagging) may widen without targeted incentives for enterprise-grade hardware and software support.