Hi David, IPv6-Mostly has been designed considering enterprise networks, not residencial ones. If you deploy IPv6-Mostly in residential networks, you will need the access link to keep dual-stack (so the ISP needs to keep CGN or keep getting sufficient IPv4 public addresses), because you don’t know if the residential can still have IPv4-only devices (older printers, smartTVs, IP phones, cameras, home automation, ….) that need to keep running. One additional issue is that those “local” devices, which only speak IPv4, now can’t communicate with the IPv6-Mostly enabled Windows (and other OSs) because they are configured (if the CPE supports also IPv6-Mostly), to run IPv6-only. So you can’t anymore print from your home to your own printer. Yes, it can be resolved in several ways, such as having a PLAT at the CPE, or having rules that allow the local traffic going to the printer to travel to the PLAT in the ISP, then travel back to the printer, etc. Not good solutions. The cost of implementing anything like that in CPEs is probably much higher than implementing CLAT, and with CLAT you have the advantage that you can get rid-off the CGN in the ISP, and instead use a less overloaded PLAT, which also requires much less public IPv4 addresses. And as more and more services in Internet enable dual-stack, less and less NAT64 traffic you will have (and you can even reduce the IPv4 public pools there). In enterprise networks (“managed” networks) the IT team can make sure that those old devices use the local PLAT, or have those devices in special network segments which have a front end with stateless NAT64, etc, even just port-proxy in windows/Linux can do that. There are several ways to fix that. The code for CLAT and PLAT is available in open source, multiple implementations. Is not a very complex code (at lest not the CLAT one). There are some vendors that have several CPEs with CLAT. The issue is that there hasn’t been sufficient market pressure to ask all the vendors to provide for existing devices a CLAT-enabled firmware. In my experience most of the time vendors do that if you pay for it (I’ve participated in a number of deployments doing that), but of course, they will not share that firmware in public, they just want money, which mainly will require you buy newer CPEs from them with the CLAT. T-Mobile for example, got CPEs with CLAT that provide broadband via 3G/4G/5G, but in general I don’t see other big players pushing the vendors for that. We really must change that if we want to move on. Why we want competition if we don’t ask them to support the standards? or we just believe the CPE is a disposable cost and because that, we spend more CapEx and OpEx in other parts of the network because the lack of vision when asking for CPEs? Regards, Jordi @jordipalet
El 16 jun 2026, a las 21:21, David Prall via NANOG <nanog@lists.nanog.org> escribió:
Soon enough all the client computers will have CLAT, so the CPE won't require it leaving the CPE IPv6-Only. IPv6-Mostly is being widely evangelized, with DHCP Option 108. The only client without CLAT widely available today is Windows, and it is out in Insider Preview with Fingers Crossed it comes to GA shortly.
David
On 6/16/2026 2:05 PM, William Herrin via NANOG wrote:
I don't see any future where organically IPv4 dies I do. First, service providers get fully on board with native Ipv6. Once a customer deploys IPv6 at all, they want it from all their providers so
On Mon, Jun 15, 2026 at 11:33 PM Saku Ytti via NANOG <nanog@lists.nanog.org> wrote: that their users' software isn't using second-class connectivity. However slow the IPv6 customer deployments, it eventually passes a threshold where the disincentive to stay IPv4-only as a service provider can't be ignored. Meanwhile, smaller and newer ISPs have an IPv4 acquisition problem. They can solve that problem with CGNAT, but if they've deployed IPv6 they can also solve it with PLATs running 464XLAT. The latter means they don't have to dual-stack their network, so it's cheaper and less error-prone. As long as they can get a CPE router which can do CLAT. Which is not much of a thing. Yet. The customers of the 464XLAT folks will occasionally want a public IPv4 address, so they'll solve that by tunnelling a static address for an extra charge. And like has happened with AWS, that'll become the standard: IPv6 and RFC1918 included, public IPv4 for an additional fee. As this happens, native IPv4 peering will become more challenging. With folks winnowing IPv4 from their network cores in favor of 464xlat there will be fewer places to trade IPv4 traffic. In addition, CGNAT and 464XLAT PLATs are challenging to manage. If your traffic is 90% IPv6, maybe the PLAT is a good candidate for outsourcing. So you buy PLAT service from an IPv4 exchange (IXPs: new product opportunity alert). And so on and so forth. 464xlat's CLATs never leave the CPE the same way modern POTS terminals still support rotary dialing but the IPv4 network collapses into a small core like Usenet has, serviced by long IPv6 tunnels. When I peer into my crystal ball in 2026, that's what I see. Regards, Bill Herrin
_______________________________________________ NANOG mailing list https://lists.nanog.org/archives/list/nanog@lists.nanog.org/message/NPKI5WUD...
********************************************** 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.