We've found that the gaming sites moved rapidly to v6 due to CGNAT issues, and that helped, but as they say, can't live with it, can't live without... - Tom On Thu, Jun 19, 2025 at 2:22 PM Adam Fathauer via NANOG < nanog@lists.nanog.org> wrote:

At my last $DAYJOB (large financial with several hundred thousand employees), we enabled IPv6, and I focused on our outbound web proxies. Day one, outbound traffic shifted to about 30% native v6 outbound. When I left 3 years ago, we were regularly pushing over 50%.

Many of the big sites (Google, Facebook, Microsoft, Apple, Netflix, LinkedIn) have v6 Internet facing services that just worked.

We were also able to enable v6 on a couple of our private connections (non-Internet) with business partners. Eliminating nat vastly simplified troubleshooting!

Feel free to reach out to me directly if you have further questions about our experience.

Adam

------ Original Message ------ From "Marco Moock via NANOG" <nanog@lists.nanog.org> To nanog@lists.nanog.org Cc "Marco Moock" <mm@dorfdsl.de> Date 6/19/2025 4:16:44 PM Subject Re: IPv6 native percentage (end user perspective)

...

Am 19.06.2025 um 14:12:20 Uhr schrieb Forrest Christian (List Account) via NANOG:

...

I see numerous statistics from Google and similar sources that indicate the percentage of end users who are IPv6 native. What I'm missing are statistics going the other way - what percentage of sites (or endpoints that customers regularly connect to) are IPv6-native, from a total traffic perspective?

https://www.vyncke.org/ipv6status/

This site provides some info about IPv6 enabled web services in certain countries.

Although, it doesn't give any info about the amount of traffic.

-- Gruß Marco

Send unsolicited bulk mail to 1750335140muell@cartoonies.org _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/EXGKCJRH... _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/26DJX7WH...

The risk of not having IPv6 is too high. It will take only one major carrier or service to say "IPv4 ends on April 25th 2027. IPv4 addresses are too expensive to support new businesses so we're making the change to support new businesses because we love you." and then stand their ground on discontinuation of IPv4. Only one single Carrier. Only one Government. Only one Company. Ex: Alphabet, The Walt Disney Company, Amazon, Microsoft, Netflix, Oracle, SalesForce, Iberdrola, DBS, Visa, Mastercard, American Express. You won't do it right? Because you make money from IPv4 functions... Think you're too small to change it? Think about a Mosquito in your tent. Will you sleep if there is a mosquito with malaria in your tent with you? You don't need me to tell you to be the Mosquito. When you're first to disable IPv4 and stick to it, you are a friend of Innovation. Eyeball network users will switch to whatever provider can get them IPv6 fastest. *The FCC is pushing for a transition to IPv6, particularly within the government, they don't explicitly require ISPs to provide IPv4 service.* On Fri, Jun 20, 2025 at 9:48 AM Marco Moock via NANOG <nanog@lists.nanog.org> wrote:

Am 20.06.2025 um 06:55:14 Uhr schrieb Tom Mitchell via NANOG:

...

We've found that the gaming sites moved rapidly to v6 due to CGNAT issues, and that helped, but as they say, can't live with it, can't live without...

Which ones did?

IIRC Steam is still v4 only.

-- Gruß Marco

Send unsolicited bulk mail to 1750395314muell@cartoonies.org _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/5ZY3LT77...

Just to provide some perspective from my viewpoint: I can run dual-stack. But I don't want to, at least for a specific customer. I want a particular customer to be IPv4 or IPv6, with an eventual transition to 100% IPv6. I don't want to restart the recurring argument, but I'll just put this out there: Why bother adding the cost of supporting a dual-stack network when there is precisely zero cost for me to stick with IPv4? From a cost perspective, if I have to assign everyone an IPv4 address and an IPv6 address to deploy IPv6, why would I bother assigning the IPv6 address? I have plenty of addresses to continue handing out IPv4 addresses directly to customers for at least several years, so there is no benefit to me in adding the overhead of dealing with both IPv6 and IPv4 on a per-customer basis simultaneously. However, I'm willing to migrate (over several years) to an IPv6-only network and run a CGNAT box to access IPv4, but only once the cost of running the CGNAT box becomes negligible. Once that occurs, I want to start getting ahead of the curve and set up a CGNAT box, then begin offering only IPv6 to new customers. Of course, the size and cost of the CGNAT device are directly related to the flows and/or bandwidth, which is why I was curious about the percentages. If it's 10% IPv6, then I'm not close to where I need to be. If it's 95%, then I can (and should) start moving to IPv6. Somewhere in the middle is the threshold, not quite sure where that number is. On Thu, Jun 19, 2025 at 3:24 PM Mark Andrews via NANOG < nanog@lists.nanog.org> wrote:

You are asking the wrong question.

Switching on IPv6 doesn’t require you to switch off IPv4. You can but you don’t have to. I find it sad that ISPs still think IPv4 and IPv6 are mutually exclusive. Nobody is asking for people to switch off IPv4. They are only asking that you enable IPv6 so they can reach you without having to run the traffic though a CGN 44 or 64.

For most eyeball networks the majority of your traffic will be IPv6 the moment you turn IPv6 on as most of the large content providers offer IPv6 and implementations prefer IPv6.

Mark -- Mark Andrews

...

On 20 Jun 2025, at 06:13, Forrest Christian (List Account) via NANOG < nanog@lists.nanog.org> wrote:

I see numerous statistics from Google and similar sources that indicate the percentage of end users who are IPv6 native. What I'm missing are statistics going the other way - what percentage of sites (or endpoints that customers regularly connect to) are IPv6-native, from a total traffic perspective?

That is, if I switch to IPv6 on my eyeball network, how much of my existing traffic will I have to CGNAT in some way to reach the IPv4-only network?

We have sufficient IPv4 address resources to stick with IPv4 for the foreseeable future. However, at some point, the percentage of traffic using IPv6 becomes so high that the reasons not to move become less significant. For example, the CGNAT box becomes significantly smaller, as most of the traffic should flow around it on IPv6.

-- - Forrest _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/ZWNAGD3G...

_______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/A75BIETJ...

-- - Forrest

I don't want to restart the recurring argument, but I'll just put this out there: Why bother adding the cost of supporting a dual-stack network when there is precisely zero cost for me to stick with IPv4? From a cost perspective, if I have to assign everyone an IPv4 address and an IPv6 address to deploy IPv6, why would I bother assigning the IPv6 address? I have plenty of addresses to continue handing out IPv4 addresses directly to customers for at least several years, so there is no benefit to me in adding the overhead of dealing with both IPv6 and IPv4 on a per-customer basis simultaneously.

At a minimum, because if they're already going to be on CGNAT, it will greatly improve their general internet access experience. Even if they won't be on CGNAT, it will still, in general, improve their internet experience on things like VOIP, multiplayer gaming, and other such similarly network centric applications that can take advantage of it. The removal of NAT layers at all is a huge benefit. Users would silently see faster and more reliable functionality of existing things that can take v6 advantage. I've implemented v6 tunnels for people on v4 only ISPs just as an easy way to resolve a variety of v4 headaches/issues that we used to just consider 'normal'. Running dual stack provides all the benefits for applications that can take advantage of it, without some of the downsides incurred by v6 transition technologies like 464XLAT and the like. It should be a zero-cost outside of labor effort to implement/support. Things like NAT64 and DNS64 do work for the most part but have rough edges when applications cannot handle v6 addressing (this is still a big issue with some stuff!). Overall dual stack is the smoothest way, and can reduce V4 load and complexity - thus leading to potential hardware downsizing and cost savings. -----Original Message----- From: Forrest Christian (List Account) via NANOG <nanog@lists.nanog.org> Sent: Thursday, June 19, 2025 6:05 PM To: North American Network Operators Group <nanog@lists.nanog.org> Cc: Forrest Christian (List Account) <lists@packetflux.com> Subject: Re: IPv6 native percentage (end user perspective) Just to provide some perspective from my viewpoint: I can run dual-stack. But I don't want to, at least for a specific customer. I want a particular customer to be IPv4 or IPv6, with an eventual transition to 100% IPv6. I don't want to restart the recurring argument, but I'll just put this out there: Why bother adding the cost of supporting a dual-stack network when there is precisely zero cost for me to stick with IPv4? From a cost perspective, if I have to assign everyone an IPv4 address and an IPv6 address to deploy IPv6, why would I bother assigning the IPv6 address? I have plenty of addresses to continue handing out IPv4 addresses directly to customers for at least several years, so there is no benefit to me in adding the overhead of dealing with both IPv6 and IPv4 on a per-customer basis simultaneously. However, I'm willing to migrate (over several years) to an IPv6-only network and run a CGNAT box to access IPv4, but only once the cost of running the CGNAT box becomes negligible. Once that occurs, I want to start getting ahead of the curve and set up a CGNAT box, then begin offering only IPv6 to new customers. Of course, the size and cost of the CGNAT device are directly related to the flows and/or bandwidth, which is why I was curious about the percentages. If it's 10% IPv6, then I'm not close to where I need to be. If it's 95%, then I can (and should) start moving to IPv6. Somewhere in the middle is the threshold, not quite sure where that number is. On Thu, Jun 19, 2025 at 3:24 PM Mark Andrews via NANOG < nanog@lists.nanog.org> wrote:

You are asking the wrong question.

Switching on IPv6 doesn’t require you to switch off IPv4. You can but you don’t have to. I find it sad that ISPs still think IPv4 and IPv6 are mutually exclusive. Nobody is asking for people to switch off IPv4. They are only asking that you enable IPv6 so they can reach you without having to run the traffic though a CGN 44 or 64.

For most eyeball networks the majority of your traffic will be IPv6 the moment you turn IPv6 on as most of the large content providers offer IPv6 and implementations prefer IPv6.

Mark -- Mark Andrews

...

On 20 Jun 2025, at 06:13, Forrest Christian (List Account) via NANOG < nanog@lists.nanog.org> wrote:

I see numerous statistics from Google and similar sources that indicate the percentage of end users who are IPv6 native. What I'm missing are statistics going the other way - what percentage of sites (or endpoints that customers regularly connect to) are IPv6-native, from a total traffic perspective?

That is, if I switch to IPv6 on my eyeball network, how much of my existing traffic will I have to CGNAT in some way to reach the IPv4-only network?

We have sufficient IPv4 address resources to stick with IPv4 for the foreseeable future. However, at some point, the percentage of traffic using IPv6 becomes so high that the reasons not to move become less significant. For example, the CGNAT box becomes significantly smaller, as most of the traffic should flow around it on IPv6.

-- - Forrest _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/ZW NAGD3GM6VKKNBE3QE5HHRJ26C4UXJF/

_______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/A7 5BIETJQDTWUGEZQWSGKNE2L5SQPNHZ/

-- - Forrest _______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/FKFUZUB5...

What happens when a customer tries to connect to a ipv6 only resource? I routinely hear the argument that “we have plenty of v4 space left”, but what about the folks at the other end? If you wait until then, you’ll either start loosing customers or have to scramble to accommodate v6 then. Much better to do it slowly on your terms. Adam Adam ------ Original Message ------ From "Forrest Christian (List Account) via NANOG" <nanog@lists.nanog.org> To "North American Network Operators Group" <nanog@lists.nanog.org> Cc "Forrest Christian (List Account)" <lists@packetflux.com> Date 6/19/2025 6:05:12 PM Subject Re: IPv6 native percentage (end user perspective)

Just to provide some perspective from my viewpoint:

I can run dual-stack. But I don't want to, at least for a specific customer. I want a particular customer to be IPv4 or IPv6, with an eventual transition to 100% IPv6.

I don't want to restart the recurring argument, but I'll just put this out there: Why bother adding the cost of supporting a dual-stack network when there is precisely zero cost for me to stick with IPv4? From a cost perspective, if I have to assign everyone an IPv4 address and an IPv6 address to deploy IPv6, why would I bother assigning the IPv6 address? I have plenty of addresses to continue handing out IPv4 addresses directly to customers for at least several years, so there is no benefit to me in adding the overhead of dealing with both IPv6 and IPv4 on a per-customer basis simultaneously.

However, I'm willing to migrate (over several years) to an IPv6-only network and run a CGNAT box to access IPv4, but only once the cost of running the CGNAT box becomes negligible. Once that occurs, I want to start getting ahead of the curve and set up a CGNAT box, then begin offering only IPv6 to new customers.

Of course, the size and cost of the CGNAT device are directly related to the flows and/or bandwidth, which is why I was curious about the percentages. If it's 10% IPv6, then I'm not close to where I need to be. If it's 95%, then I can (and should) start moving to IPv6. Somewhere in the middle is the threshold, not quite sure where that number is.

On Thu, Jun 19, 2025 at 3:24 PM Mark Andrews via NANOG < nanog@lists.nanog.org> wrote:

...

You are asking the wrong question.

Switching on IPv6 doesn’t require you to switch off IPv4. You can but you don’t have to. I find it sad that ISPs still think IPv4 and IPv6 are mutually exclusive. Nobody is asking for people to switch off IPv4. They are only asking that you enable IPv6 so they can reach you without having to run the traffic though a CGN 44 or 64.

For most eyeball networks the majority of your traffic will be IPv6 the moment you turn IPv6 on as most of the large content providers offer IPv6 and implementations prefer IPv6.

Mark -- Mark Andrews

...

On 20 Jun 2025, at 06:13, Forrest Christian (List Account) via NANOG < nanog@lists.nanog.org> wrote:

I see numerous statistics from Google and similar sources that indicate the percentage of end users who are IPv6 native. What I'm missing are statistics going the other way - what percentage of sites (or endpoints that customers regularly connect to) are IPv6-native, from a total traffic perspective?

That is, if I switch to IPv6 on my eyeball network, how much of my existing traffic will I have to CGNAT in some way to reach the IPv4-only network?

We have sufficient IPv4 address resources to stick with IPv4 for the foreseeable future. However, at some point, the percentage of traffic using IPv6 becomes so high that the reasons not to move become less significant. For example, the CGNAT box becomes significantly smaller, as most of the traffic should flow around it on IPv6.

-- - Forrest _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/ZWNAGD3G...

_______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/A75BIETJ...

-- - Forrest _______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/FKFUZUB5...

That’s why IPv6-only with IPv4aaS makes sense instead of dual-stack. You don’t need to run dual-stack in most parts of the network, but the customers have dual-stack in their LANs. You don’t care anymore about if customers or destinations are IPv4, IPv6 or dual-stack. All become transparent to you. Moreover, if your customers are mainly residential, you will find that typically 85% of your traffic is going to CDNs or content providers that are already IPv6-enabled, so your NAT64 traffic will be much lower than the actual CGN traffic, and will go even lower with the time. Waiting for 100% of the content providers to be IPv6 ready will not work, always someone will be still lost in the IPv4, and for sure you don’t want to get helpdesk calls complaining about that if you switch to IPv6-only (without IPv4aaS). The issue is to have the right CPEs or CPE firmware that supports IPv6-only with IPv4aaS. Some vendors still need much more customers pressure to offer that. An alternative is of course open source, which is available for both the CPE (OpenWRT) and the NAT64 (Jool). We have done that for several customers and zero issues. Regards, Jordi @jordipalet

El 20 jun 2025, a las 7:46, Forrest Christian (List Account) via NANOG <nanog@lists.nanog.org> escribió:

Which is why I asked the original question.

My intent is not to never move to IPv6, it's just that we're in no rush to do so. Note that we've had IPv6 on-network with our own address space since 2008. It just hasn't expanded much beyond our core due to various challenges and a lack of financial incentive to do so.

If nearly 100% of the content providers were reachable via IPv6 today. I'd likely have already been switched. If that number were still in the almost-zero category, I wouldn't even consider migrating to IPv6, other than in the core and niche cases that we've been running since we started experimenting with it 18 years ago.

At some point, the incremental cost of adding the necessary hardware to support NAT for IPv6-only customers to access the legacy IPv4 internet will be low enough that it will make sense for us to deploy it. At that point, I won't see any reason to continue deploying IPv4 for new customers, and it will become our legacy protocol. We're certainly getting closer to this threshold. I don't know what that threshold is, but we are getting closer and closer to it. I do know it's going to be well before the internet shuts off IPv4.

On Thu, Jun 19, 2025 at 8:45 PM Adam Fathauer <adam@arfmail.com> wrote:

...

What happens when a customer tries to connect to a ipv6 only resource?

I routinely hear the argument that “we have plenty of v4 space left”, but what about the folks at the other end?

If you wait until then, you’ll either start loosing customers or have to scramble to accommodate v6 then. Much better to do it slowly on your terms.

Adam

Adam

------ Original Message ------ From "Forrest Christian (List Account) via NANOG" <nanog@lists.nanog.org> To "North American Network Operators Group" <nanog@lists.nanog.org> Cc "Forrest Christian (List Account)" <lists@packetflux.com> Date 6/19/2025 6:05:12 PM Subject Re: IPv6 native percentage (end user perspective)

Just to provide some perspective from my viewpoint:

I can run dual-stack. But I don't want to, at least for a specific customer. I want a particular customer to be IPv4 or IPv6, with an eventual transition to 100% IPv6.

I don't want to restart the recurring argument, but I'll just put this out there: Why bother adding the cost of supporting a dual-stack network when there is precisely zero cost for me to stick with IPv4? From a cost perspective, if I have to assign everyone an IPv4 address and an IPv6 address to deploy IPv6, why would I bother assigning the IPv6 address? I have plenty of addresses to continue handing out IPv4 addresses directly to customers for at least several years, so there is no benefit to me in adding the overhead of dealing with both IPv6 and IPv4 on a per-customer basis simultaneously.

However, I'm willing to migrate (over several years) to an IPv6-only network and run a CGNAT box to access IPv4, but only once the cost of running the CGNAT box becomes negligible. Once that occurs, I want to start getting ahead of the curve and set up a CGNAT box, then begin offering only IPv6 to new customers.

Of course, the size and cost of the CGNAT device are directly related to the flows and/or bandwidth, which is why I was curious about the percentages. If it's 10% IPv6, then I'm not close to where I need to be. If it's 95%, then I can (and should) start moving to IPv6. Somewhere in the middle is the threshold, not quite sure where that number is.

On Thu, Jun 19, 2025 at 3:24 PM Mark Andrews via NANOG < nanog@lists.nanog.org> wrote:

You are asking the wrong question.

Switching on IPv6 doesn’t require you to switch off IPv4. You can but you don’t have to. I find it sad that ISPs still think IPv4 and IPv6 are mutually exclusive. Nobody is asking for people to switch off IPv4. They are only asking that you enable IPv6 so they can reach you without having to run the traffic though a CGN 44 or 64.

For most eyeball networks the majority of your traffic will be IPv6 the moment you turn IPv6 on as most of the large content providers offer IPv6 and implementations prefer IPv6.

Mark -- Mark Andrews

...

On 20 Jun 2025, at 06:13, Forrest Christian (List Account) via NANOG < nanog@lists.nanog.org> wrote:

I see numerous statistics from Google and similar sources that indicate the percentage of end users who are IPv6 native. What I'm missing are statistics going the other way - what percentage of sites (or endpoints that customers regularly connect to) are IPv6-native, from a total traffic perspective?

That is, if I switch to IPv6 on my eyeball network, how much of my existing traffic will I have to CGNAT in some way to reach the IPv4-only network?

We have sufficient IPv4 address resources to stick with IPv4 for the foreseeable future. However, at some point, the percentage of traffic using IPv6 becomes so high that the reasons not to move become less significant. For example, the CGNAT box becomes significantly smaller, as most of the traffic should flow around it on IPv6.

-- - Forrest _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/ZWNAGD3G...

_______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/A75BIETJ...

-- - Forrest _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/FKFUZUB5...

-- - Forrest _______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/NLEG5VLB...

********************************************** IPv4 is over Are you ready for the new Internet ? http://www.theipv6company.com The IPv6 Company This electronic message contains information which may be privileged or confidential. The information is intended to be for the exclusive use of the individual(s) named above and further non-explicilty authorized disclosure, copying, distribution or use of the contents of this information, even if partially, including attached files, is strictly prohibited and will be considered a criminal offense. If you are not the intended recipient be aware that any disclosure, copying, distribution or use of the contents of this information, even if partially, including attached files, is strictly prohibited, will be considered a criminal offense, so you must reply to the original sender to inform about this communication and delete it.

I switched my home internet couple weeks ago to ipv6less, because it was 5eur MRC cheaper. I don’t think customers care about ipv6 at all. ++ytti ________________________________ Lähettäjä: Michael Thomas via NANOG <nanog@lists.nanog.org> Lähetetty: Friday, June 20, 2025 4:33:11 PM Vastaanottaja: nanog@lists.nanog.org <nanog@lists.nanog.org> Kopio: Michael Thomas <mike@mtcc.com> Aihe: Re: IPv6 native percentage (end user perspective) It's sort of surprising that providers haven't used v6 as a marketing tool. There is some new stupid Xfinity ad about Wifi going kaboom or something that makes about zero sense, but honestly most eyeball networks are differentiating themselves mostly on nothing at all (cf: "kaboom"). v6 actually does give a direct path from host to server without the network complexity which is good for latency, etc. So they could actually visualize it as a network with a NAT Goridan Knot in the middle that their new-fangled upgrade -- ipv6 -- chops in half to unleash the true speed potential. Or something. The point is that bits are pretty much commodities, so prettier looking packaging is probably what a lot of this reduces to. Heck, you could even have partner programs that tout the content providers who work with your new tech (::snort:: 30 year old "new"). For the ones who've already upgraded, it's free publicity as well. Mike, forgive my lack of caffeine On 6/19/25 11:58 PM, jordi.palet--- via NANOG wrote:

That’s why IPv6-only with IPv4aaS makes sense instead of dual-stack.

You don’t need to run dual-stack in most parts of the network, but the customers have dual-stack in their LANs. You don’t care anymore about if customers or destinations are IPv4, IPv6 or dual-stack. All become transparent to you.

Moreover, if your customers are mainly residential, you will find that typically 85% of your traffic is going to CDNs or content providers that are already IPv6-enabled, so your NAT64 traffic will be much lower than the actual CGN traffic, and will go even lower with the time.

Waiting for 100% of the content providers to be IPv6 ready will not work, always someone will be still lost in the IPv4, and for sure you don’t want to get helpdesk calls complaining about that if you switch to IPv6-only (without IPv4aaS).

The issue is to have the right CPEs or CPE firmware that supports IPv6-only with IPv4aaS. Some vendors still need much more customers pressure to offer that. An alternative is of course open source, which is available for both the CPE (OpenWRT) and the NAT64 (Jool). We have done that for several customers and zero issues.

Regards, Jordi

@jordipalet

...

El 20 jun 2025, a las 7:46, Forrest Christian (List Account) via NANOG <nanog@lists.nanog.org> escribió:

Which is why I asked the original question.

My intent is not to never move to IPv6, it's just that we're in no rush to do so. Note that we've had IPv6 on-network with our own address space since 2008. It just hasn't expanded much beyond our core due to various challenges and a lack of financial incentive to do so.

If nearly 100% of the content providers were reachable via IPv6 today. I'd likely have already been switched. If that number were still in the almost-zero category, I wouldn't even consider migrating to IPv6, other than in the core and niche cases that we've been running since we started experimenting with it 18 years ago.

At some point, the incremental cost of adding the necessary hardware to support NAT for IPv6-only customers to access the legacy IPv4 internet will be low enough that it will make sense for us to deploy it. At that point, I won't see any reason to continue deploying IPv4 for new customers, and it will become our legacy protocol. We're certainly getting closer to this threshold. I don't know what that threshold is, but we are getting closer and closer to it. I do know it's going to be well before the internet shuts off IPv4.

On Thu, Jun 19, 2025 at 8:45 PM Adam Fathauer <adam@arfmail.com> wrote:

...

What happens when a customer tries to connect to a ipv6 only resource?

I routinely hear the argument that “we have plenty of v4 space left”, but what about the folks at the other end?

If you wait until then, you’ll either start loosing customers or have to scramble to accommodate v6 then. Much better to do it slowly on your terms.

Adam

Adam

------ Original Message ------ From "Forrest Christian (List Account) via NANOG" <nanog@lists.nanog.org> To "North American Network Operators Group" <nanog@lists.nanog.org> Cc "Forrest Christian (List Account)" <lists@packetflux.com> Date 6/19/2025 6:05:12 PM Subject Re: IPv6 native percentage (end user perspective)

Just to provide some perspective from my viewpoint:

I can run dual-stack. But I don't want to, at least for a specific customer. I want a particular customer to be IPv4 or IPv6, with an eventual transition to 100% IPv6.

I don't want to restart the recurring argument, but I'll just put this out there: Why bother adding the cost of supporting a dual-stack network when there is precisely zero cost for me to stick with IPv4? From a cost perspective, if I have to assign everyone an IPv4 address and an IPv6 address to deploy IPv6, why would I bother assigning the IPv6 address? I have plenty of addresses to continue handing out IPv4 addresses directly to customers for at least several years, so there is no benefit to me in adding the overhead of dealing with both IPv6 and IPv4 on a per-customer basis simultaneously.

However, I'm willing to migrate (over several years) to an IPv6-only network and run a CGNAT box to access IPv4, but only once the cost of running the CGNAT box becomes negligible. Once that occurs, I want to start getting ahead of the curve and set up a CGNAT box, then begin offering only IPv6 to new customers.

Of course, the size and cost of the CGNAT device are directly related to the flows and/or bandwidth, which is why I was curious about the percentages. If it's 10% IPv6, then I'm not close to where I need to be. If it's 95%, then I can (and should) start moving to IPv6. Somewhere in the middle is the threshold, not quite sure where that number is.

On Thu, Jun 19, 2025 at 3:24 PM Mark Andrews via NANOG < nanog@lists.nanog.org> wrote:

You are asking the wrong question.

Switching on IPv6 doesn’t require you to switch off IPv4. You can but you don’t have to. I find it sad that ISPs still think IPv4 and IPv6 are mutually exclusive. Nobody is asking for people to switch off IPv4. They are only asking that you enable IPv6 so they can reach you without having to run the traffic though a CGN 44 or 64.

For most eyeball networks the majority of your traffic will be IPv6 the moment you turn IPv6 on as most of the large content providers offer IPv6 and implementations prefer IPv6.

Mark -- Mark Andrews

...

On 20 Jun 2025, at 06:13, Forrest Christian (List Account) via NANOG < nanog@lists.nanog.org> wrote: I see numerous statistics from Google and similar sources that indicate the percentage of end users who are IPv6 native. What I'm missing are statistics going the other way - what percentage of sites (or endpoints that customers regularly connect to) are IPv6-native, from a total traffic perspective?

That is, if I switch to IPv6 on my eyeball network, how much of my existing traffic will I have to CGNAT in some way to reach the IPv4-only network?

We have sufficient IPv4 address resources to stick with IPv4 for the foreseeable future. However, at some point, the percentage of traffic using IPv6 becomes so high that the reasons not to move become less significant. For example, the CGNAT box becomes significantly smaller, as most of the traffic should flow around it on IPv6.

-- - Forrest _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/ZWNAGD3G...

_______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/A75BIETJ...

-- - Forrest _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/FKFUZUB5...

-- - Forrest _______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/NLEG5VLB...

********************************************** IPv4 is over Are you ready for the new Internet ? http://www.theipv6company.com The IPv6 Company

This electronic message contains information which may be privileged or confidential. The information is intended to be for the exclusive use of the individual(s) named above and further non-explicilty authorized disclosure, copying, distribution or use of the contents of this information, even if partially, including attached files, is strictly prohibited and will be considered a criminal offense. If you are not the intended recipient be aware that any disclosure, copying, distribution or use of the contents of this information, even if partially, including attached files, is strictly prohibited, will be considered a criminal offense, so you must reply to the original sender to inform about this communication and delete it.

_______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/ZVYKHGCU...

NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/HFBDSWGN...

Just a bit of prodding (wink, wink). Do you know that Astound Broadband (aka RCN) does NOT support IPv6 at the customer level and has not announced when they will do so. Are they kind of light on technical talent over there? Just asking. Unfortunately, some people in condo buildings don't have a choice as the condo associations make corrupt deals with providers to shut out competitors. On Friday, June 20th, 2025 at 8:58 AM, Saku Ytti via NANOG <nanog@lists.nanog.org> wrote:

I switched my home internet couple weeks ago to ipv6less, because it was 5eur MRC cheaper. I don’t think customers care about ipv6 at all.

++ytti ________________________________ Lähettäjä: Michael Thomas via NANOG nanog@lists.nanog.org

Lähetetty: Friday, June 20, 2025 4:33:11 PM Vastaanottaja: nanog@lists.nanog.org nanog@lists.nanog.org

Kopio: Michael Thomas mike@mtcc.com

Aihe: Re: IPv6 native percentage (end user perspective)

It's sort of surprising that providers haven't used v6 as a marketing tool. There is some new stupid Xfinity ad about Wifi going kaboom or something that makes about zero sense, but honestly most eyeball networks are differentiating themselves mostly on nothing at all (cf: "kaboom"). v6 actually does give a direct path from host to server without the network complexity which is good for latency, etc. So they could actually visualize it as a network with a NAT Goridan Knot in the middle that their new-fangled upgrade -- ipv6 -- chops in half to unleash the true speed potential. Or something.

The point is that bits are pretty much commodities, so prettier looking packaging is probably what a lot of this reduces to. Heck, you could even have partner programs that tout the content providers who work with your new tech (::snort:: 30 year old "new"). For the ones who've already upgraded, it's free publicity as well.

Mike, forgive my lack of caffeine

On 6/19/25 11:58 PM, jordi.palet--- via NANOG wrote:

...

That’s why IPv6-only with IPv4aaS makes sense instead of dual-stack.

You don’t need to run dual-stack in most parts of the network, but the customers have dual-stack in their LANs. You don’t care anymore about if customers or destinations are IPv4, IPv6 or dual-stack. All become transparent to you.

Moreover, if your customers are mainly residential, you will find that typically 85% of your traffic is going to CDNs or content providers that are already IPv6-enabled, so your NAT64 traffic will be much lower than the actual CGN traffic, and will go even lower with the time.

Waiting for 100% of the content providers to be IPv6 ready will not work, always someone will be still lost in the IPv4, and for sure you don’t want to get helpdesk calls complaining about that if you switch to IPv6-only (without IPv4aaS).

The issue is to have the right CPEs or CPE firmware that supports IPv6-only with IPv4aaS. Some vendors still need much more customers pressure to offer that. An alternative is of course open source, which is available for both the CPE (OpenWRT) and the NAT64 (Jool). We have done that for several customers and zero issues.

Regards, Jordi

@jordipalet

...

El 20 jun 2025, a las 7:46, Forrest Christian (List Account) via NANOG nanog@lists.nanog.org escribió:

Which is why I asked the original question.

My intent is not to never move to IPv6, it's just that we're in no rush to do so. Note that we've had IPv6 on-network with our own address space since 2008. It just hasn't expanded much beyond our core due to various challenges and a lack of financial incentive to do so.

If nearly 100% of the content providers were reachable via IPv6 today. I'd likely have already been switched. If that number were still in the almost-zero category, I wouldn't even consider migrating to IPv6, other than in the core and niche cases that we've been running since we started experimenting with it 18 years ago.

At some point, the incremental cost of adding the necessary hardware to support NAT for IPv6-only customers to access the legacy IPv4 internet will be low enough that it will make sense for us to deploy it. At that point, I won't see any reason to continue deploying IPv4 for new customers, and it will become our legacy protocol. We're certainly getting closer to this threshold. I don't know what that threshold is, but we are getting closer and closer to it. I do know it's going to be well before the internet shuts off IPv4.

On Thu, Jun 19, 2025 at 8:45 PM Adam Fathauer adam@arfmail.com wrote:

...

What happens when a customer tries to connect to a ipv6 only resource?

I routinely hear the argument that “we have plenty of v4 space left”, but what about the folks at the other end?

If you wait until then, you’ll either start loosing customers or have to scramble to accommodate v6 then. Much better to do it slowly on your terms.

Adam

Adam

------ Original Message ------ From "Forrest Christian (List Account) via NANOG" nanog@lists.nanog.org To "North American Network Operators Group" nanog@lists.nanog.org Cc "Forrest Christian (List Account)" lists@packetflux.com Date 6/19/2025 6:05:12 PM Subject Re: IPv6 native percentage (end user perspective)

Just to provide some perspective from my viewpoint:

I can run dual-stack. But I don't want to, at least for a specific customer. I want a particular customer to be IPv4 or IPv6, with an eventual transition to 100% IPv6.

I don't want to restart the recurring argument, but I'll just put this out there: Why bother adding the cost of supporting a dual-stack network when there is precisely zero cost for me to stick with IPv4? From a cost perspective, if I have to assign everyone an IPv4 address and an IPv6 address to deploy IPv6, why would I bother assigning the IPv6 address? I have plenty of addresses to continue handing out IPv4 addresses directly to customers for at least several years, so there is no benefit to me in adding the overhead of dealing with both IPv6 and IPv4 on a per-customer basis simultaneously.

However, I'm willing to migrate (over several years) to an IPv6-only network and run a CGNAT box to access IPv4, but only once the cost of running the CGNAT box becomes negligible. Once that occurs, I want to start getting ahead of the curve and set up a CGNAT box, then begin offering only IPv6 to new customers.

Of course, the size and cost of the CGNAT device are directly related to the flows and/or bandwidth, which is why I was curious about the percentages. If it's 10% IPv6, then I'm not close to where I need to be. If it's 95%, then I can (and should) start moving to IPv6. Somewhere in the middle is the threshold, not quite sure where that number is.

On Thu, Jun 19, 2025 at 3:24 PM Mark Andrews via NANOG < nanog@lists.nanog.org> wrote:

You are asking the wrong question.

Switching on IPv6 doesn’t require you to switch off IPv4. You can but you don’t have to. I find it sad that ISPs still think IPv4 and IPv6 are mutually exclusive. Nobody is asking for people to switch off IPv4. They are only asking that you enable IPv6 so they can reach you without having to run the traffic though a CGN 44 or 64.

For most eyeball networks the majority of your traffic will be IPv6 the moment you turn IPv6 on as most of the large content providers offer IPv6 and implementations prefer IPv6.

Mark -- Mark Andrews

...

On 20 Jun 2025, at 06:13, Forrest Christian (List Account) via NANOG < nanog@lists.nanog.org> wrote: I see numerous statistics from Google and similar sources that indicate the percentage of end users who are IPv6 native. What I'm missing are statistics going the other way - what percentage of sites (or endpoints that customers regularly connect to) are IPv6-native, from a total traffic perspective?

That is, if I switch to IPv6 on my eyeball network, how much of my existing traffic will I have to CGNAT in some way to reach the IPv4-only network?

We have sufficient IPv4 address resources to stick with IPv4 for the foreseeable future. However, at some point, the percentage of traffic using IPv6 becomes so high that the reasons not to move become less significant. For example, the CGNAT box becomes significantly smaller, as most of the traffic should flow around it on IPv6.

-- - Forrest _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/ZWNAGD3G...

_______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/A75BIETJ...

-- - Forrest _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/FKFUZUB5...

-- - Forrest _______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/NLEG5VLB...

********************************************** IPv4 is over Are you ready for the new Internet ? http://www.theipv6company.com The IPv6 Company

This electronic message contains information which may be privileged or confidential. The information is intended to be for the exclusive use of the individual(s) named above and further non-explicilty authorized disclosure, copying, distribution or use of the contents of this information, even if partially, including attached files, is strictly prohibited and will be considered a criminal offense. If you are not the intended recipient be aware that any disclosure, copying, distribution or use of the contents of this information, even if partially, including attached files, is strictly prohibited, will be considered a criminal offense, so you must reply to the original sender to inform about this communication and delete it.

_______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/ZVYKHGCU...

_______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/HFBDSWGN... _______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/VKHPPE3J...

On 6/20/25 15:58, Saku Ytti via NANOG wrote:

I switched my home internet couple weeks ago to ipv6less, because it was 5eur MRC cheaper. I don’t think customers care about ipv6 at all.

Customers never did care. Mark.

We are getting closer to dual stacking our subs. In the meantime, adding mams interfaces from a new ms-mpc-128g card, into the existing ams0 interface, thus doubling the cgnat capacity, was as easy as adding a Ethernet link to and ae bundle interface. Aaron

On Jun 19, 2025, at 5:43 PM, Niels Bakker via NANOG <nanog@lists.nanog.org> wrote:

I'll bite:

* Forrest Christian [Fri 20 Jun 2025, 00:06 CEST]:

...

I don't want to restart the recurring argument, but I'll just put this out there: Why bother adding the cost of supporting a dual-stack network when there is precisely zero cost for me to stick with IPv4? From a cost perspective, if I have to assign everyone an IPv4 address and an IPv6 address to deploy IPv6, why would I bother assigning the IPv6 address?

What if everybody thought that way? Would we ever get to a position where we could even consider turning off IPv4 altogether?

(We must consider eventually turning off IPv4. We've run out twice now, the first time we innovated our way out with CIDR, the second time there's no other option on the table but IPv6. The lack of IPv4 is currently a global drag on non-financial innovation.)

...

I have plenty of addresses to continue handing out IPv4 addresses directly to customers for at least several years, so there is no benefit to me in adding the overhead of dealing with both IPv6 and IPv4 on a per-customer basis simultaneously.

For now. While your competitors are gaining valuable experience with IPv6. And put some customers behind CGNAT, freeing up IPv4 addresses they can monetise in different ways, like sell or rent out as subnets.

Have fun scaling your CGNAT boxes to all the traffic from the customer base you'll eventually have to put behind them. If you had run dual stack then a lot of traffic wouldn't need to traverse them.

-- Niels. _______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/VJO36DQ4...

I think what we need is for some big tech companies to sign a contract with each other that they start dropping IPv4 at their network edge in 2035 or so. This would then signal the market that you're going to need to deploy IPv6 and that you can do IPv6 only, because IPv4 only networks will have to figure out translation in their edges. And I think they should be motivated to do this, to get rid of the requirement of purchasing IPv4 addresses. However they probably will always be able to sink that cost in their products, and the real companies suffering from access to IPv4 spaces are competitors who never start. So it might be a good anti-competitive strategy to keep the IPv4 dream alive.

I'm reasonably sure that those big tech companies are (closely) tracking their numbers, and know pretty accurately how much traffic (direct correlation: revenue) they would lose if they switched to v6 only. Ideally with projections on how much that loss may be in 2030, 2035, ... (with lots of assumptions, sure). What would possibly make them decide to drop 5% or 1% or 0.5% or even 0.1% of their potential customers (and revenue) at that time "for the good of the internet"? Robert

A reminder that all USA civilian federal agencies are supposed to have 80% of their endpoints on IPv6-only networks (i.e. no native PV4) by the end of this fiscal year, the end of September. And every product the federal government has bought for the past ten years or so is supposed to have been USGv6 compliant. Anyone who wants to keep, or hopes to someday get, some of that federal $$$, you best be IPv6-ready. On Thu, Jun 19, 2025 at 3:06 PM Forrest Christian (List Account) via NANOG < nanog@lists.nanog.org> wrote:

Just to provide some perspective from my viewpoint:

I can run dual-stack. But I don't want to, at least for a specific customer. I want a particular customer to be IPv4 or IPv6, with an eventual transition to 100% IPv6.

I don't want to restart the recurring argument, but I'll just put this out there: Why bother adding the cost of supporting a dual-stack network when there is precisely zero cost for me to stick with IPv4? From a cost perspective, if I have to assign everyone an IPv4 address and an IPv6 address to deploy IPv6, why would I bother assigning the IPv6 address? I have plenty of addresses to continue handing out IPv4 addresses directly to customers for at least several years, so there is no benefit to me in adding the overhead of dealing with both IPv6 and IPv4 on a per-customer basis simultaneously.

However, I'm willing to migrate (over several years) to an IPv6-only network and run a CGNAT box to access IPv4, but only once the cost of running the CGNAT box becomes negligible. Once that occurs, I want to start getting ahead of the curve and set up a CGNAT box, then begin offering only IPv6 to new customers.

Of course, the size and cost of the CGNAT device are directly related to the flows and/or bandwidth, which is why I was curious about the percentages. If it's 10% IPv6, then I'm not close to where I need to be. If it's 95%, then I can (and should) start moving to IPv6. Somewhere in the middle is the threshold, not quite sure where that number is.

On Thu, Jun 19, 2025 at 3:24 PM Mark Andrews via NANOG < nanog@lists.nanog.org> wrote:

...

You are asking the wrong question.

Switching on IPv6 doesn’t require you to switch off IPv4. You can but you don’t have to. I find it sad that ISPs still think IPv4 and IPv6 are mutually exclusive. Nobody is asking for people to switch off IPv4. They are only asking that you enable IPv6 so they can reach you without having to run the traffic though a CGN 44 or 64.

For most eyeball networks the majority of your traffic will be IPv6 the moment you turn IPv6 on as most of the large content providers offer IPv6 and implementations prefer IPv6.

Mark -- Mark Andrews

...

On 20 Jun 2025, at 06:13, Forrest Christian (List Account) via NANOG < nanog@lists.nanog.org> wrote:

I see numerous statistics from Google and similar sources that indicate the percentage of end users who are IPv6 native. What I'm missing are statistics going the other way - what percentage of sites (or endpoints that customers regularly connect to) are IPv6-native, from a total traffic perspective?

That is, if I switch to IPv6 on my eyeball network, how much of my existing traffic will I have to CGNAT in some way to reach the IPv4-only network?

We have sufficient IPv4 address resources to stick with IPv4 for the foreseeable future. However, at some point, the percentage of traffic using IPv6 becomes so high that the reasons not to move become less significant. For example, the CGNAT box becomes significantly smaller, as most of the traffic should flow around it on IPv6.

-- - Forrest _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/ZWNAGD3G...

...

_______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/A75BIETJ...

-- - Forrest _______________________________________________ NANOG mailing list

https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/FKFUZUB5...

Essentially labor (support) costs: Training support staff on IPv6, including helping customers provision/use IPv6. Adding IPv6 to internally developed databases and provisioning tools. Adding support to IPAM and our address-allocation and tracking tools to support IPv6 (we are largely statically addressed to the customer for historical reasons). Configuring IPv6 across the network. Troubleshooting/supporting two protocols that can break instead of one. Many of these have become less expensive than before. For example, 10 years ago, I would have added "upgrade all the gear that doesn't support IPv6," but I think we're mostly beyond that. On Fri, Jun 20, 2025 at 8:04 AM Tom Mitchell <tmitchell@netelastic.com> wrote:

"Why bother adding the cost of supporting a dual-stack network when there is precisely zero cost for me to stick with IPv4? "

What is the incremental cost?

- Tom

-- - Forrest