I don't necessarily see an issue here - DNS solves this quite nicely, but ..... When the IP changes, you only have half the IP that needs updating, and it's a simple sed exercise to replace it in all configurations, if you have it hard coded for any reason. One of the nice things about this is that, the times I've had to re-number networks, it was a zero downtime operation with smooth transition. I'd go full DNS instead of LUA style, myself (which is what I've done), so renumbering doesn't impact me at all. Everything is updated/correct within a minute or so of address change. Sadly, as I've noted before, my last mile doesn't have IPv6 service so I've had to tunnel that, but I've re-numbered numerous times since I first started tunneling in 2009 without any operational impact. At most, dyndns provider updates for external services were the only low hanging fruit that had any impact. -----Original Message----- From: Chris Woodfield via NANOG <nanog@lists.nanog.org> Sent: Monday, December 1, 2025 5:47 PM To: North American Network Operators Group <nanog@lists.nanog.org> Cc: Bryan Fields <Bryan@bryanfields.net>; Chris Woodfield <chris@woodfield.tech> Subject: Re: IPv6 Performance (was Re: IPv4 Pricing) I’ll chime in my personal beef with IPv6, or at least, my home ISP’s implementation… Unless I want to pay $$$ for a “business-class” service for my home, my IP allocations, both IPv4 and V6, are not statically assigned. While they don’t change often, they have in the past. Now, if I want to assign static addresses for devices within my home network, I don’t have a problem with v4 - everything’s RFC1918, so if the public IP changes, NBD, and I can even do it with DHCP client IDs. However, if my IPv6 PD changes and my home devices all have GUAs assigned via SLAAC, then… guess what - every IPv6 device address in my network just changed. Oops. Practically, I’ve worked around this by manually assigning LUAs to the devices that need static v6 addresses, like my SAN and the machines that do NFS mounts from it. But 1. that’s more than annoyingly clunky - hardly the improved experience that IPv6 promised - and 2. weren’t we trying to get away from LUAs in the first place? -Chris
On Dec 1, 2025, at 13:44, Bryan Fields via NANOG <nanog@lists.nanog.org> wrote:
On 12/1/25 14:22, Jared Mauch via NANOG wrote:
I find myself having to tether off their networks when I’m on IPv4 only networks to access things like my hypervisors and other assets that are IPv6-only because they have superior networking these days.
While I'll agree v6 is easy and should be deployed I have to take issue with the current as-built being superior.
At least once or twice a month I'm downloading something and will find the IPv4 to transfer significantly faster. Case in point, I downloaded the proxmox iso yesterday to a colo server with 50g uplinks. It loafed at 2.4 mbytes/s using default wget, which of course preferred ipv6. Adding -4 to wget made that shoot up to 80 mbytes/s.
This is ipv6 behavior I've seen time and time again. I'm unsure where problems like these lie in the network, other than it's not mine or my peers. I've seen the same issues with v6 paths to the same server bounce around the west coast and back, whilst IPv4 is 6 hops and 12 ms away.
This is exactly the sort of thing that holds IPv6 back by giving it a bad name. -- Bryan Fields
727-409-1194 - Voice http://bryanfields.net _______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/AP A2YIX47NF7U65G2HIBAPHT3X6EWRIG/
_______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/AWW6EP3W...