> It failed to solve the problem of impending IP address depletion
I wouldn't say so. Some mobile carriers and big data centers have used IPv6 to pretty much completely solve the problem of being able to assign a unique address to devices.
For mobile devices, moving 50% of traffic over to IPv6 means buying half as many CGNAT/v6-to-v4 boxes (of various kinds).
And on the v6-inside, unique address can be assigned. Legal requirement and court orders suck when you get "who had A.A.A.A:32800 at time T?" if you have to go through three levels of NAT to decode that. So even if a customer only accesses IPv4, having their actual handset only be assigned IPv6 makes things easier and cheaper. Even if they share an outside address, there's only one translation so the inside is unique.
For big data companies, it means not needing to solve the problem of running out of 10/8 (yes I'm aware of the other private addresses), and having an address plan problem any time they make an acquisition.
And I've seen large providers who build their whole actual network with IPv6, and only tunnel IPv4 on top of it. Huge savings in complexity and cost of IPv4 addresses.
So what I'm saying is that I've seen first hand in multiple large providers of different kinds how IPv6 is delivering incremental payoff for incremental adoption.
It doesn't have to be 100% before we get ROI.
> it is not a success.
About half of even public traffic on the most complex and distributed system ever built is IPv6.
It's going slower than I'd like, but it's definitely paying off.
There are still ATM and X.25 networks out there, so is IPv4 a failure? (admittedly, a bit hyperbolic)
I'm working on a problem right now at a large company to move a thing from IPv4 to IPv6 because the existing IPv4 solution is running out of addresses, and it's impossible (for multiple reasons) to "just add more IPv4". Can't go into details, sorry.
I should've qualified that as address exhaustion on the Internet, the side adventure of private networking has no bearing on the goal that IPng had set out to do, which was to address the impending address exhaustion. You say you wouldn't say so, but here we are, IPv4 exhausted, and IPv4 remains the incumbent. If IPv6 had succeeded, we would probably be having this very discussion on an IPv6 enabled site, the cost difference between a v4 address and a v6 address would be negligible, that is to say v6 would not be a second class citizen or an optional bolt-on to the Internet. I mean that's all that needs to be said about whether it has succeeded in what it needed to do.
> I should've qualified that as address exhaustion on the Internet
Well I addressed that too, so…
> private networking
To some extent this is a distinction without a difference. Again, as I said…
> we would probably be having this very discussion on an IPv6 enabled site
$ host news.ycombinator.com
news.ycombinator.com has address 209.216.230.207
news.ycombinator.com has IPv6 address 2606:7100:1:67::26
When IPv4 is disrupted for me, I only notice because github.com goes away.
> v6 [is] a second class citizen
It is. Except for endpoints (again) as I mentioned…
> the cost difference between a v4 address
The alternative to buying v4 is not just private addresses, as (again, as I was very specific about) private v4 addresses also have a cost.
v4 is priced according to the demand. Without IPv6 demand would be much higher, as the alternative (with CGNAT and intra org problems) would drive up the demand for more public addresses.
To say that "the cost should be equal" for IPv6 to not be a partial/in progress success misses the entire economics of addresses.
The biggest most complex system in the world shuffles half its traffic on IPv6, and rising, with million of devices without any form of IPv4 address.
This is mainly due to mobile devices only being issued ipv6 addresses by the telco 4g networks. They are the only ones using ipv6 on the millions of clients scale.
Everything supports both. We are talking about being issued only IPv6 addresses where you actually use it to connect to stuff.
Most mobile devices are only issued an IPv6 address and therefore when the masses do google searches it uses IPv6 and makes it look like there is huge adoption.
> We are talking about being issued only IPv6 addresses where you actually use it to connect to stuff.
You seem to be asserting that dual-stack machines use IPv4 by default, but that's not really true. If your machine has both IPv4 and IPv6 connectivity, browsers will in fact use IPv6 to connect to sites that support it, like Google. They prefer IPv6 by default and fall back to IPv4 if IPv6 is slower (Happy Eyeballs algorithm).
Of course, random software can mostly use whichever it wants, so I'm not claiming every process on such a machine will use IPv6, but most common stuff does.
Make sure they actually have GUA addresses, not just link-locals.
If you're on a Linux machine, check `ip -c addr show` for an address that's "scope global" and doesn't start with "f". Those are the ones you need. If you have one of those, check `getent ahosts google.com` to see if v6 is being sorted before v4 in DNS lookups, and then `wget google.com` to see if wget prints any errors connecting to the v6 address.
If you have GUA addresses and nothing is outright broken, devices and software that support v6 will prefer it over v6.
Well, I am. MacBook on a home internet connection in Arizona. Using IPv6 by default without me having ever had to do anything special to configure it.
You are simply misinformed. Either your setup doesn’t actually support IPv6 (or it’s much slower than IPv4 due to something being misconfigured), or you turned it off at some point, or you’re making a mistake in how you measure it. Because IPv6 is used by default on systems that support it. You don’t have to take my word for this, you can google it or ask someone else to try it.
Unsurprisingly Google actually does also have IPv4 addresses. What they're measuring isn't "How did you reach our servers?" but instead "Could you have reached our IPv6 servers?"
My understanding (for which I can't give you a citation) is that a tiny fraction of Google visitors are randomly chosen to try to reach IPv6 servers and measure what happens.
Because of Happy Eyeballs if you measure whether your users did use IPv6 you don't find out whether they could have done so, and so your results will be thrown off by happenstance.
APNIC's stats check for that. For the US, it makes the difference between 58.74% capable and 57.85% preferring, so it doesn't produce a huge discrepancy.
I believe your understanding here is incorrect. It doesn’t make sense that Google would claim to measure usage while actually measuring access. I can’t find anything that supports your assertion.
"When large masses of devices that use IPv6 connect to IPv6 servers it makes it look like there is huge IPv6 adoption"
I don't understand your logic. How does a large amount of devices using IPv6 to connect to IPv6 servers only "make it look" like there is IPv6 adoption but somehow it shouldn't count?
Half. The. Internet.
What a failure. /s