Skip to content

RFC 015 – Mandatory IPv6 Adoption#15

Open
hitchhooker wants to merge 1 commit intomainfrom
rfc15
Open

RFC 015 – Mandatory IPv6 Adoption#15
hitchhooker wants to merge 1 commit intomainfrom
rfc15

Conversation

@hitchhooker
Copy link
Contributor

Summary

All IBP members must provide IPv6 addresses for their endpoints by January 31,
2026.

Motivation

JAM is an IPv6-only network planned for launch Q3 2026. IBP proxies must support
IPv6 by Q1 2026 to be JAM-ready.

Why All Members

AAAA records override A records in DNS resolution. If only partial members
provide IPv6 addresses, GeoDNS will route to the reduced set of IPv6-capable
endpoints. This degrades latency and coverage for all users as fewer geographic
endpoints are available.

100% member adoption is required to maintain current service quality when IPv6
is enabled.

Proposal

Requirement Deadline
All members submit IPv6 addresses to config January 31, 2026
IPv6 endpoints operational and reachable January 31, 2026

Members not compliant by deadline become ineligible for bandwidth reimbursement
at RFC-014 rates until compliant.

Rationale

JAM's IPv6-only architecture requires IBP infrastructure to be fully
IPv6-capable before launch. Q1 2026 compliance gives 6 months buffer before Q3
2026 JAM launch for testing and integration.


Vote

On the GitHub RFC PR:

  • 👍 Thumbs up = Aye
  • 👎 Thumbs down = Nay

No reaction = abstain.


@dcolley
Copy link

dcolley commented Nov 26, 2025

If "JAM is an IPv6-only network", do we need ipv6 internally?

@hitchhooker
Copy link
Contributor Author

In IPv6 you don't really have internal and external networking with network address translations so all nodes with IPv6 are also external and truly peer-to-peer without holepunching hacks. I'm not sure what kind of network setup you have but you probably need BGP peering session with your provider to get v6 routing table to both of your edge routers.

Most transit providers will give you a /48 or /56 for free with existing contract — just ask. Once you have the prefix:

  1. Add address-families=ipv6 to your existing BGP peer config (or better probably to create separate v6 session)
  2. Announce your prefix from both routers
  3. Run VRRPv3 (v3 handles both v4 and v6) for your internal gateway failover(or run anycast(birdc) on all servers announcing same ip for haproxy redundancy)
  4. Make sure ICMPv6 passes your firewall — NDP (neighbor discovery) breaks without it, this is the most common mistake

If you're not running your own ASN and just have provider-assigned addresses, it's even simpler — they handle the routing, you just need to enable v6 on interfaces and set the gateway.

Your VMs become globally routable. No more NAT state tables, no more hole punching for p2p connections. Validators can peer directly. So hardening with iptables/nftables on all VM kinda must have.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants