IPv8

by doener • 2 months ago

15 13 2 months ago

doener• 2 months ago

"Lucien,

I see no need for this IPv8. IPv6 was carefully engineered over many years and while not perfect, works and is deployed. What problem are you trying to solve? I seem to have missed that."

https://seclists.org/nanog/2026/May/9

mmooss2• 2 months ago

That link is confusing. That message is from Gary Sparkes. Sparkes quotes a message from Jamie Thain [0], the author of the OP [1]. Thain quotes Lucien Hoydic who wrote that paragraph many messages prior [2] but for some reason Thain doesn't indicate it's a quote (e.g., Thain doesn't prefix the quote with ">") so it looks like Thain said it.

The best starting point is April 29th in the April archive, where Thain begins the discussion:

https://seclists.org/nanog/2026/Apr/date.html

The whole thread is people in the establishment mocking the idea, which is common to innovations good and bad, adopted and not.

[0] https://seclists.org/nanog/2026/May/6

[1] Per the FAQ: "Who is behind it?" "J. Thain (One Limited), author of draft-thain-ipv8-02 and the companion specifications covering routing protocols, RINE, ZoneServer, WHOIS8, NetLog8, and Update8."

[2] https://seclists.org/nanog/2026/Apr/130

lschueller• 2 months ago

Sorry, but how can an ip layer be an >integrated network management< ?

The github repo is completely empty. No draft / documentation? Looking forward to see a bit more of the technicalities

Bender2• 2 months ago

I keep seeing IPv8 (or maybe IPv4 V2) being suggested on HN. I do not see this happening in my lifetime. IPv6 while not perfect has taken 14+ years to reach almost 50% adoption most of which is from wireless and VPS providers and that has started to plateau. IPv8 if approved would have to work it's way through the daunting myriad of hardware obstacles, middle boxes, routers, operating systems and so much more that IPv6 had to complete. There are still a massive number of ISP's that need to implement IPv6 just to get people out of the CG-NAT mess. That and / or reclaim all the squatted, wasted and unused IPv4 space which is another topic in and of itself. I could clone the entire existing internet with the wasted space in IPv4 not even counting multicast. Just as an example with a couple orders from the head of the US and UK military several /8's could be freed up in a year. After that make each non ISP company that has anything more than a /20 justify each and every IP via on-site audits, quarterly. Part out ever "Future use" /8 and make non ISP companies justify their bids.

Given that IPv6 exists I can not imagine the people that integrated it now suddenly adding yet another routing solution. I would expect a majority of them to say something to the effect of, "Finish deploying IPv6 first as it already exists then we can talk about what gaps remain."

kstrauser2• 2 months ago

several /8's could be freed up in a year

But remember that each /8 freed with great time and expense (because organizations with a /8 can also afford a fleet of lawyers the size of Rhode Island) adds approximately 0.4% to the IPv4 pool. There aren't that many of those ripe for the plucking, either.

Bender2• 2 months ago

    grep -c Future iana_ipv4.txt [1]
    16

There's also about a dozen /8's that the US and UK defense industries do not use. Most of the upper portion of the multicast /8's could be repurposed and that operating systems are not aware of. There about a dozen /8's that would indeed require lawsuits or they could be annexed by governments. I could probably fit most of the existing internet in the unused and squatted space. There was a time companies had to justify their allocations. We could make IANA / ARIN / RIPE / AFRINIC and others great again. We could also take back 38/8 again as that has always been mismanaged ever since the PSINet incident.

Either way, IPv6 is already a thing and people can use that as well.

This list [1] is a bit ambiguous. Much of what looks used is not really used.

[1] https://www.iana.org/assignments/ipv4-address-space/ipv4-add...

kstrauser2• 2 months ago

This is idiotic. I'm sorry/not-sorry for the bluntness, but it just is. It's "100% backward compatible", but one of the first projects is:

Kernel build of the IPv8 stack
- Linux kernel branch implementing the IPv8 packet header (RFC §6)
- AF_INET8 / sockaddr_in8 socket API surface
- ARP8 dual-probe neighbour capability discovery
- 8to4 tunnelling for IPv8-over-IPv4 transit

So aside from having to reinvent everything, including kernel-level packet header processing, it's A-OK! It also adds:

Security by protocol
- East–west isolation via ACL8. North–south egress validated against DNS8 + WHOIS8.

...for those who were tired of being able to ping other hosts on your LAN without, let's see, checking with a layer 7 protocol first.

In an interview, the creator said[0]:

Another IPv8 feature is what Thain calls a “Zone server” that his draft explains “runs every service a network segment requires: address assignment (DHCP8), name resolution (DNS8), time synchronisation (NTP8), telemetry collection (NetLog8), authentication caching (OAuth8), route validation (WHOIS8 resolver), access control enforcement (ACL8), and IPv4/IPv8 translation (XLATE8).”

Boy, you thought people had opinions on systemd? Wait until they get a peek at this all-your-eggs-in-one-basket critical infrastructure.

I mean this seriously: there's no part of this that I want anything to do with. Zero. Nothing. I'm only being so harsh here because I don't want to leave any wiggle room of doubt that perhaps I think any of it is redeemable, because it ain't.

[0]https://www.theregister.com/networks/2026/05/12/veteran-netw...

jaysonvantuyl• 2 months ago

I feel like this completely misses the entire point (or at least half the functionality) of ASNs.

ASNs give you an identity at the peering level that can be used to dynamically provision routes; so it is used to evolve (over time) where a given network-level address lives. Making it part of the address now makes that mapping vastly less flexible.

The list of things it breaks is breathtaking:

- Transferring a subnet between providers can't be done (the ASN part changes).

- Providing the same services on multiple providers doesn't work (different ASNs can't advertise the same address).

- Various types of multicast domain management break (global networks that use IBGP for internet multicast routing need to figure out their prefixes).

Then there's the bootstrapping problem with the mix of layer 3 / layer 7. And JWT used everywhere? And no integrated solutions for things like DNSSEC (warts and all)? No apparent attention to distributed implementations of (now consolidated) core infrastructure?

Say what you want about systemd, at least there is arguably someone applying a different (but consistent) design to a real problem. This "proposal" feels like someone with limited understanding of the design principles behind the Internet opened up ChatGPT and asked it to fix "problems" that they barely understand.

I try not to be mean to the well-meaning, but... well... some people aspire to operate at a level for which they are entirely unqualified.

omoikane• 2 months ago

See also:

https://news.ycombinator.com/item?id=47788857 - Internet Protocol Version 8 (IPv8) (2026-04-16, 111 comments)

By the way:

After 25 years of dual-stack effort, IPv6 still carries a minority of global traffic.

It reached 50% last month, I wouldn't call that a minority.

https://news.ycombinator.com/item?id=47777894 - IPv6 traffic crosses the 50% mark (2026-04-15, 621 comments)

mmooss2• 2 months ago

Some interesting information from the author, Jamie Thain, where they introduce the idea on nanog:

https://seclists.org/nanog/2026/Apr/96

I joined this list because, as part of IPv8, I am creating a BGPv8. Inside BGPv8, two new protocols CF (Cost Factor), weigh cost factors along the routes to produce a better metric. It's a hybrid of EIRGP mixed with BGP to create better engineering results.

I also as part of CF created Sun Tzu which is the protocol that watches CF and gives you a CF score of reliability. Do I trust my partnership with you?

Now, beyond an on-slaught of IPv8 is stupid, IPv6 solves every problem, etc, etc. That's not my discussion point. My point isn't "should I even propose IPv8" my point is what would be the best result for operators?

So the things to know, IPV8 is NOT a 64 bit addressing system.

It is a 32 bit routing system with a 32 bit addressing system.

A Routing Number = ASNs plus others.

8.8.8.8 would become 15169.8.8.8.8

https://www.ietf.org/archive/id/draft-thain-ipv8-02.html

queenkjuul• 2 months ago

LLMs must really love convincing people that they should replace IPv6

viraptor• 2 months ago

No, let's not do this. This is a weird perversion of how open internet should work. Too much control and management layers on the way. This honestly feels like some corp's fake-grassroot campaign. (Inviting donations and looking for interested hackers... yet something feels off...) It continues to pop up in various places without a serious rationale that can't be immediately dismissed as at least stretching the truth. I've not seen any known, serious people discuss it and be excited about it anywhere. This really smells BAD to me.

hackeraccount• 2 months ago

Let's give up a lot of things for a protocol that seems more backwards compatible (but realistically isn't).

IPv8

Related Stories