IPv8: The Bold New Protocol That Wants to Replace the Internet’s Plumbing

Every few years, someone proposes a new version of the Internet Protocol and the networking community erupts in spirited debate. In April 2026, that someone is J. Thain of One Limited, whose IETF Internet-Draft — draft-thain-ipv8-01 — proposes nothing less than a ground-up reimagining of how packets move across the internet, under the banner of Internet Protocol Version 8, or IPv8.

The proposal has drawn thousands of eyeballs on Hacker News, sharp criticism from veteran network engineers, and a fair amount of confusion about a basic question: whatever happened to IPv5, IPv6, and IPv7? To understand where IPv8 fits — or doesn’t — it helps to first trace the strange numbering history of the protocol that underpins the modern internet.

A brief and confusing history of IP version numbers

If you have only ever heard of IPv4 and IPv6, you are not alone. Most everyday internet users have no reason to encounter the others. But the version number field in every IP packet is just four bits wide, allowing values from 0 to 15. Several of those slots have already been claimed — and then quietly abandoned.

IPv4, codified in RFC 791 in 1981, is the protocol that still carries the vast majority of internet traffic today. IPv5 was never a public internet protocol at all; it was reserved for the experimental Internet Stream Protocol, a multimedia streaming experiment that was eventually discontinued. IPv6, finalised in 1998 and slowly deployed ever since, is the official successor to IPv4, expanding addresses from 32 bits (about 4.3 billion addresses) to 128 bits — enough for every atom on Earth to have a unique identifier, with room to spare.

Where did IPv7 go?

 In the early 1990s, as address exhaustion loomed, the IETF convened its “IPng Area” to evaluate competing successor proposals. One of those was TP/IX — informally called IPv7 — which proposed expanding addresses from 32 to 64 bits while keeping the rest of IPv4’s architecture largely intact.
A parallel proposal, Pip, was assigned version number 8. Neither won. The IETF chose a more ambitious design — Simple Internet Protocol Plus, or SIPP — which became IPv6. IPv7 was formally obsoleted. IPv8 (the Pip protocol) quietly followed.
Version number 9 briefly appeared in an April Fools’ RFC in 1994. Today, the 2026 draft reclaims the “IPv8” label for an entirely new proposal, with no connection to the original Pip protocol.

What IPv8 actually proposes

The 2026 draft describes IPv8 not merely as a new addressing scheme but as a “managed network protocol suite” — a comprehensive rethinking of how networks authenticate devices, deliver configuration, resolve names, and route traffic. Its ambitions go well beyond swapping out address formats.

At its core, the proposal introduces a 64-bit address space. Each Autonomous System Number (ASN) holder — the organisations that run pieces of the internet — would receive 2³² host addresses, or about 4.3 billion, which the draft argues is sufficient for any organisation at any scale without the need for address exhaustion workarounds like Carrier-Grade NAT.

Beyond addressing, IPv8 proposes to unify several services that today operate independently. A new DHCP8 mechanism would deliver everything a device needs — its address, DNS servers, time synchronisation, authentication tokens — in a single lease response. A new DNS8 system would validate every outgoing packet against a registered route in WHOIS8, a proposed critical infrastructure directory. Network telemetry, access control, and name resolution would all flow through a single “Zone Server” platform running in each network segment.

Security is a first-class concern. East-west traffic — communication between devices inside a network — would be isolated by ACL8 zone controls, with devices permitted to speak only to designated service gateways. North-south traffic, heading out to the internet, would be validated at egress. The proposal also introduces BGP8, a version of the internet’s routing protocol in which the global routing table is bounded by the number of ASNs rather than the number of prefixes — a structural change intended to arrest the explosive growth of routing tables that today strain internet infrastructure.

IPv6 vs. IPv8: a direct comparison

IPv6 and the 2026 IPv8 proposal share a common origin story — both were conceived as responses to the limitations of IPv4 — but they represent very different philosophies about what a next-generation protocol should do.

IPv6’s address space is dramatically larger — 128 bits versus IPv8’s 64 — but the draft argues that raw address size is no longer the primary problem. With 2³² addresses per ASN, IPv8 contends that no organisation will ever face exhaustion, and that the massive IPv6 space is largely unnecessary overhead. Critics disagree, pointing out that IoT growth, dynamic allocation, and privacy-preserving address rotation all consume addresses faster than expected.

Where IPv6 is essentially IPv4 with a larger address field and some refinements, IPv8 is a far more interventionist proposal. It mandates a Zone Server infrastructure, OAuth2 authentication at the network layer, and certified NIC firmware — changes that would touch every device, router, switch, and software stack in existence.

Why “IPv8” and not “IPv7”?

The numbering choice is one of the most common questions the draft provokes. If IPv6 is the current standard, why not propose IPv7 as its successor?

The answer lies in that 1990s standards competition. Version number 7 was assigned to the TP/IX proposal during the IPng evaluation process and, even though TP/IX lost and was subsequently obsoleted, the version number itself remains historically claimed. The IANA (Internet Assigned Numbers Authority) tracks all version number assignments, and version 7 is recorded as used — even if the protocol it was used for is long dead.

Version 8 was similarly assigned to the Pip protocol in the 1990s, which raises an awkward question: by the same logic, why is version 8 available? The 2026 draft author appears to treat the version number as reclaimable, given that both the Pip protocol and its successor proposals were abandoned and never standardised. Whether IANA would formally reassign the number is a separate matter — and one that would only arise if the draft were ever on a serious path to standardisation, which, given current reception, seems unlikely in the near term.

The reception: enthusiasm meets scepticism

The networking community’s reaction has been mixed, to put it charitably. Engineers who have spent careers on IPv6 deployment have raised pointed objections. The backward-compatibility claim, they argue, is more aspirational than real: any existing router, switch ASIC, or host stack that encounters a packet with version number 8 in the header will simply drop it, as the header format is entirely different from IPv4.

The proposal also demands a sweeping new stack — new socket APIs, new DNS record types, a new ARP mechanism, revised OSPF and BGP, mandatory OAuth2 on every switch port, and persistent TCP connections from every end device to a Zone Server. Far from requiring no modification, critics say, it requires modifying everything simultaneously.

Supporters of the general direction find merit in the ambition. The internet’s management plane — how networks are configured, monitored, and secured — genuinely has not kept pace with the scale and complexity of modern infrastructure. The proposal’s push toward a unified identity model and coherent telemetry addresses real operational pain. The debate, then, is not really about whether the problems are real, but whether a single sweeping protocol overhaul is the right medicine.

What comes next

The draft is currently set to expire in October 2026 unless updated. It carries no IETF endorsement and has no formal standing in the standards process. For IPv8 to proceed, it would need to attract a working group, survive multiple rounds of expert review, and ultimately demonstrate interoperability in real-world deployments — a path that has taken IPv6 over two decades and is still not complete.

In the meantime, IPv6 deployment continues its slow, steady march. The more likely trajectory for internet protocol evolution is not a dramatic version-number leap but the kind of incremental extension and overlay that has always characterised the internet’s growth: more encryption here, better routing security there, and ever-better tooling to manage the complexity underneath.

IPv8 is, for now, a thought experiment with an IETF draft number. Whether it becomes anything more will depend not on the elegance of its design, but on whether anyone is willing to do the unglamorous work of making it run on real hardware in a real network.