What is IPv8: the protocol that solves IPv4 exhaustion without breaking what you have

IPv8 is a network protocol proposal published as an Internet-Draft at the IETF on 15 April 2026 that aims to replace IPv4 by fixing its two major problems at the root: address exhaustion and the fragmentation of network management. It does so while maintaining full backwards compatibility with IPv4 — with no need for dual-stack or forced migration — and using 64-bit addresses that embed the Autonomous System Number (ASN) as a routing prefix.

This week the draft draft-thain-ipv8-01 generated quite a bit of noise in the networking community. Some have called it “AI-generated”; others have read it more carefully and see some interesting proposals. We read it top to bottom and here’s what it proposes, where the problem comes from, and why it might — or might not — make sense.

What problem does IPv8 solve?

IPv4, the protocol that has been running since the 1980s, uses 32-bit addresses allowing just over 4 billion addresses — which seemed infinite in the 80s but were exhausted in 2011. When the IPs ran out, CGNAT (Carrier-Grade NAT) was invented: your ISP puts you and hundreds of other customers behind a single public IP. The consequences are well known: direct device-to-device communication becomes complicated (goodbye to many online games, torrents, direct VoIP…), traceability logs become a nightmare, and latency increases. It’s not an elegant solution; it’s a patch.

IPv6 was designed to fix exactly this. With 128 bits, it offers an address space so enormous it barely makes sense to write it out. The problem is that IPv6 has been standardised for over 25 years and still isn’t the dominant protocol. The reason is the transition model: IPv6 is not backwards compatible with IPv4, which forced the dual-stack model where every device, application, and network had to support both protocols simultaneously. That turned out to be commercially unacceptable for most organisations. If CGNAT was working well enough, why migrate?

How IPv8 works: address format and IPv4 compatibility

Here’s the interesting part. IPv8 proposes 64-bit addresses with the following format:

r.r.r.r.n.n.n.n

The first 32 bits (r.r.r.r) are a routing prefix based on the Autonomous System Number (ASN) of the operator or company. The other 32 bits (n.n.n.n) are the host address, with the same semantics as a regular IPv4 address. Total space: 2^64 — over 18 trillion addresses, with 4,294,967,296 hosts available for each registered ASN.

The most important design choice is backwards compatibility. An IPv4 address is represented in IPv8 as 0.0.0.0.n.n.n.n: when the routing prefix is all zeros, standard IPv4 rules apply. This means IPv4 is a proper subset of IPv8. No device, application, or network needs to be modified. No D-day, no forced migration, no dual-stack.

Beyond addresses, the proposal includes a complete ecosystem of protocols: DHCP8 to deliver all network configuration in a single response, DNS8 for name resolution, BGP8 for routing with mandatory route validation against a WHOIS8 registry, OAuth2/JWT-based authentication, and unified telemetry. All managed through a Zone Server acting as gateway, DNS server, NTP server, authentication manager, and network monitor in one.

Why IPv8 might succeed where IPv6 has failed

If anything has held IPv6 back, it’s the transition model. IPv8 addresses this far more pragmatically: instead of requiring dual-stack, it makes IPv4 coexist inside the IPv8 address space itself. The 0.0.0.0 prefix acts as “legacy mode,” and any traffic with that prefix is routed with the same old IPv4 rules. No breakage, no “either everything is migrated or nothing works.”

The other argument in its favour is unified network management. One of the big problems with current networks is that every service has its own authentication model, its own log format, its own tools. IPv8 proposes a common identity model for the entire infrastructure, which would simplify administration, security, and incident detection.

Criticisms and limitations

To be honest: this is an individual draft, not a proposal backed by any IETF working group. Anyone can publish an Internet-Draft, and the process for something like this to become a standard is long, complex, and requires a lot of consensus. The community has been seeing “IPv4++” proposals for years that never materialise.

Part of the criticism points out that the document mixes real problems with solutions that add complexity, and that the dependence on centralised services like WHOIS8 could create new single points of failure. There’s also debate about whether a new protocol is really needed, or whether it would be enough to keep pushing IPv6 with better transition mechanisms.

What is undeniable is that the problem it’s trying to solve is real. IPv4 has been held together with duct tape for decades, and IPv6 hasn’t achieved the mass displacement that was expected. If someone manages to propose a transition that doesn’t require throwing anything away and unifies network management along the way, they at least deserve to be read carefully.

The full draft is available on the IETF Datatracker. Worth a look if you work with networks, if only to see where some people are looking when they think about the future of the internet.

More info: https://ipv8.es/en/