Connect your Claude, Codex or OpenClaw for full transcripts on millions of podcasts. Two week free trial, cancel anytime.

Start free trial →
AJ Allan Jude Allan Jude Works on FreeBSD and ZFS storage systems and has co-written books on ZFS administration, including FreeBSD Mastery: Advanced ZFS. Hosts BSD Now and 2.5 Admins,...

“people's reaction of just disable IPv6 is exactly why IPv6 has not gotten very far. I had hoped that it was getting better when everybody's phone was kind of depending on it to avoid carrier grade NAT, but it seems like we only care about it on the phone and we've not applied any of those lessons to the rest of the internet.”

2.5 Admins · Hands-On Engineering Podcasts · October 2026

“people's reaction of just disable IPv6 is exactly why IPv6 has not gotten very far. I had hoped that it was getting better when everybody's phone was kind of depending on it to avoid carrier grade NAT, but it seems like we only care about it on the phone and we've not applied any of those lessons to the rest of the internet.” — Allan Jude, 2.5 Admins

Context

The hosts were discussing a report that a pre-installed Intel driver rewrites the IPv6 flow label to zero, breaking Microsoft 365 and Azure services behind load balancers, and that many people respond by disabling IPv6. Allan says that reflex is part of why IPv6 adoption has stalled.

Transcript

2.5 Admins Around 16:38 into the episode
Allan Jude

…wouldn't be as obviously, it's just broken as this. So yeah, this is a very terrible driver feature, which is almost certainly a bug. I don't think there's much value in ever setting that flow label to zero. It's probably trying to do more with it and screwing it up. But that's pretty terrible.

Jim Salter

…major network interface manufacturers doing something as bloody Johnson stupid as resetting flow labels to zero. Like Alan said, that's got to be a bug. There's no valid reason for it. And frankly, it strikes me as a very much romper room bug that really never should have made it out of the shop.

Allan Jude

…catch this in the lab with iperf and TCP dump. It would probably take somebody a long time to figure out this is what was happening. But once you knew what to look for, it's obvious what is happening. And it does seem that they should fix it. And also, if Microsoft has known for a while that it's the problem, why does it keep reinstalling the broken version? But to Jim's point about adoption, people's reaction of just disable IPv6 is exactly why IPv6 has not gotten very far. I had hoped that it was getting better when everybody's phone was kind of depending on it to avoid carrier grade NAT, but it seems like we only care about it on the phone and we've not applied any of those lessons to the rest of the internet. You

Jim Salter

know, Alan, you also perhaps accidentally raised another really good question, which is when you've got one tech giant like Intel that's delivering software and software updates via the Windows catalog, which Microsoft controls, at what point does Microsoft step in and take ownership and say, no…

Allan Jude

Yeah. And I also wonder how tempted Azure and Amazon and so are to change their load balancer to patch around this problem so that people can at least still use their computers that have this broken software on it, but at the same time, not have to ruin all their load balancing and run around with…

Jim Salter

I'm going to guess they weren't very tempted at all to patch around this issue because since it was only IPv6, it was probably only affecting about the third or fourth nine of their traffic.

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

More from 2.5 Admins