A Month Without IPV4 Is Like A Month Without…

Recently, there was a Mastodon post from [nixCraft] challenging people to drop their NAT routers for the month of November and use only IPv6. What would it be like to experience “No NAT November?” [Alex Haydock] decided to find out.

What did he learn? You’d imagine he’d either wholeheartedly embrace IPv6 or stagger back in and warn everyone not to mess with their configuration. Instead, he recommends you go IPv6 mostly. He notes he is only talking about a home network, not necessarily networks for a big company or an Internet carrier. That’s a different topic.

IPv6 has been around since 1998, but it has been slow to catch on. However, OS support seems universal at this point. [Alex] was able to easily switch on IPv6 only using Windows, macOS, and several Linux flavors. He didn’t use any Android devices, but they should be OK. His iOS phones were fine.

Where he did have problems was with embedded devices like the Nintendo Switch and a Steam Deck — surprising, since the Steam Deck uses Linux. Actually, the Steam device does support IPV6, it just thinks that if it doesn’t have an IPv4 network, the network must be down.

Some home networking gear also required IPv4 addresses to use their management interfaces. That’s especially funny since the devices clearly know about IPv6. They just don’t serve web pages over their IPv6 address.

Unfortunately, there are many websites that do not have IPV6 servers. That’s not as rare as you might think and [Alex] points out offenders like GitHub, Reddit, Discord, and Steam. No IPv4, no access to those and many other sites.

So despite being No NAT November, it was necessary to set up a NAT64 gateway to read IPv4-only websites. However, unlike normal IPv4 NAT (NAT44), you can use a NAT64 gateway anywhere on the network. [Alex’s] ISP hosts a NAT64 and DNS64 instance and that solved his problem.

The post goes on about other specific cases — if you’ve ever even thought about IPv6, it is worth a read. Switching over? Probably not yet, but as [Alex] points out, with a little work and perseverance, it is possible.

In addition to our earlier coverage of why IPv6 isn’t more popular, we’ve also made the arguments about why it should be.

89 thoughts on “A Month Without IPV4 Is Like A Month Without…

  1. “IPv6 has been around since 1998, but it has been slow to catch on. However, OS support seems universal at this point.”

    DOS, including FreeDOS is still on TCP/IP v4.
    Links, Lynx/DOSLynx, Dillo, Arachne, MicroWeb v2 – they rely on TCP/IP v4.

        1. Perhaps they intended to reply to the “no mouse March” thread, elsewhere in the comments here. I can imagine making that mistake; there are a lot of reply links, but the “nesting” lines are fairly low contrast. I don’t think they are a bot.

    1. This article is so nonsensical. Yes, let me just throw everything into Chaos at home and then reverse engineer everything and add 2 more appliances to support the half of the devices that don’t support IPV6.

      Not to mention, how is this “No NAT” even possible? Are home Internet providers now passing out blocks of IPv6 address to residences? If I connect hosts through a switch to the cable modem, will it pass out multiple IP6 addresses? Can I call up my provider and get a block of public IP6 addresses to distribute from my own DHCP server?? Yes let’s just throw out my firewall as well and plug my shady IoT devices raw dog right into the cable modem. I like to live dangerously.

      Oh my gosh this sounds like so much fun! I also like to ram my pinky toe into the chair legs over and over for fun.

      1. Yes ISPs do give out blocks of IPv6. Enough for multiple subnets usually. And yes you can still use a firewall within your router. And yes you can use DHCPv6, or each machine can pick a random IP within the subnet advertised by the router.

  2. Some of these stories are so dumb.
    Lets have a no flash drive Feb, or lets have a no mouse March..
    Network engineer here, and even I think this is ‘tarded beyond belief.
    Tell me you’re bored without telling me you’re bored.

        1. You could do that with IPv4 just as easily. The whole idea of running out of address space is based on the idea that somehow every single internet connected device needs a unique IP… It doesn’t… Your fridge and your toaster will share the same WAN IP

        2. Correct. IPV6 was invented to eliminate network administrators. All your devices are pawned by someone else. It’s the equivalent to giving all the prisoners a full set of keys to the prison because, using logic, the locks are all super secure so everyone is guaranteed safety. You don’t need firewalls in the corporate world because everyone has a secure key to the unfettered internet. Nirvana.

          1. Wow, this Is definitely the worst I have read today. Thank you for making me laugh, but you should not be spreading misinformation. In no way IPv6 compromises the use of firewalls. NAT Is not a firewall, although many beginners are unable to tell the difference since it “can work like a firewall” in some circumstances.
            If anything, the corporate world will no longer need to pay huge sums of money on IPv4 allocation, and deal with all the problems introduced by Nat, that break the abstraction of a uniquely addressable Internet.

    1. “No mouse March” is a fantastic idea.

      Flush out the apps that can’t be fully used without a mouse. Pretty much anything other than CAD or image manipulation should have enough hotkeys and tabstops to access every feature.

      At least once a week I accidentally create a bookmark on my work-issued Mac (because that’s the hotkey that puts the cursor in the address bar on my personal machines) and then Chrome won’t let me dismiss the confirmation dialog without taking a hand off the keyboard. The really annoying thing is that until a few months ago it was possible to dismiss that dialog with a tab or two and a spacebar.

      1. And No Touchscreen Tuesday. Find devices that can only be controlled with a touchscreen that need physical controls.

        Recently got a new car and made sure I can control everything without the touchscreen. Even the infotainment system and the “must-be-in-park” configuration setup stuff has buttons.

        Its like TOS, buttons, knobs, switches.

        1. Out of interest, what new car did you find that does all that with buttons and switches?
          I can do some stuff with buttons and switches in my car but annoyingly the AC can only be controlled by touchscreen.

    2. Consider that one of the largest problems affecting ipv6 rollout is likely complacency – namely, people are used to working around ipv4’s deficiencies, and that they’re not likely going to change how they do things, given that what they currently do is working, despite it being hacks layered on top of hacks…

      This gets those people to step outside their comfort zone to try something new.

      That doesn’t sound ‘tarded to me.

      1. Yeah, I quite like the idea. A dry run when you’re not under pressure and you can research and solve issues properly. It’s like companies that have “disaster recovery plans” they’ve never tested and don’t work when faced with the real world.

    3. what? i actually thought this was the most genuinely useful article i’d seen in ages.

      i’ve finally got the option of ipv6 and i want to know what i would experience with varying levels of commitment to it. this article reports someone else dipping their toes in the water, and their result tells me not to bother yet.

  3. I don’t think I, or more other people, need a static IP. So one advantage of IPv6 as part of a global Internet-of-Things isn’t really so relevant. Having my own non-routable subnet on IPv4 is exactly what I need for a LAN, no more, no less. Philosophically, viewing the Internet as something you only access through a gateway and don’t let into your home is a Good Thing(tm)! The separation, even if only conceptual, is part of what’s necessary to make it harder for the bad guys to get into your home.

    uIP wouldn’t be hard to port to DOS. But it would probably need to be linked like WatTCP is, and then you’d have a completely different API and you’d have to mess with patching all your favorite DOS apps to get them to work. It hasn’t happened because nobody running FreeDOS wanted IPv6 that badly to go to the trouble.

    1. I think you’ve confused routing with firewalling (and this confusion is very common).

      There’s a big difference between “I can connect these things but I choose not to” vs “I can’t connect these things even when I want to”.

        1. That is what they meant, but this is independent of if one has a static IP.

          I can put my NAT behind a static or public IP, makes no difference to the devices behind it.

          But if I want to connect to my network from the outside, it’s a big hassle with a dynamic IP. It’s doable, but it requires a 3rd party service.

      1. Why is it difficult to understand that the humidity sensor in my bathroom, my neighbor’s coffee machine, or the TV in my bedroom don’t need to be directly addressable from the outside Internet.

        Only one of those devices should ever even try to talk to the outside network. That connection should always be self initiated and as anonymous as possible.

        IPv6 evangelists are correct that there are some technical reasons for it to exist.
        They are also blind to the numerous external problems that it causes.

        Firewalls as a solution?
        Sure… Except that just shifts the burden to network admins, and now requires them in every single household. It isn’t less work. It is simply different work, and the penalty for screwing it up is a seemingly working device that is a security/privacy problem, instead of an obviously non-working device.

        I don’t disable IPv6 because I’m too lazy to implement it.
        I disable it because I know devices are going to TRY to use it, and exactly one of those devices should be communicating outside my network.

    2. It’s not about static IPs, it is about public IPs. NAT creates all kind of issues, especially if you start staging multiple of them. Normally you still get a public IP at least for land line connections like DSL or fibre, so there is no NAT router (at home) behind a NAT router (at the ISP) yet. I think the first countries have started to deviate from that, because their ISPs did not get enough IP addresses any more.

      This is not how the internet was designed. NAT should be an exception. It is not fully transparent. Also it totally prevents you from running anything at home which can be connected from the outside, like your own “cloud”.

      1. All new ISPs in the UK have been forced to use CGNAT for a few years now: the RIPE pool of IPv4 addresses is all but exhausted. The only way for me to get a public IPv4 address on my home router is to pay £8/mo for a static IP. All of the new FTTP providers are this way.

        Many other ISPs in Europe use CGNAT, even established ones. At least with some, you can get a public (but dynamic) IP address for free by turning on the DDNS service they provide.

        Otherwise, it’s 100.64.0.0/10 for you.

        1. Same here. I’m in the UK. 1gig up/down fiber, CGNAT by default or additional cost for an IP, and they aren’t even static it’s just one that that’s public facing.
          The worst thing is, they don’t offer IPv6 at all.

      2. NAT is already in common use. Every cellular provider uses CGNAT. Tmobile home internet is IPv6 only.

        Its interesting though, aside from using torrents, I had next to zero issues without IPv4.

        Bu then I also don’t know enough to know how they get around those limitations.

      3. “Also it totally prevents you from running anything at home which can be connected from the outside, like your own “cloud”.”

        What? You can totally do it, everybody does it, NAT is not only for outgoing traffic it can be for incoming traffic also combined with port forwarding…

        1. IPv6 uses link-local addresses (denoted by FE80::/10) to determine if an interface is meant for local network communication only, otherwise for the wider internet it uses global unicast addresses, which are routable across different networks.
          It’s basically got the same protection and more than NAT has, baked in. If you want your devices on the LAN to be only accessed from LAN, don’t give them a public internet IPv6 address.

          1. But what if those devices need to communicate out?

            From what I understand about the stateless addressing system, devices will generate their own unique IP within the network. That seems like it would still leak numbers of devices, tie them to traffic, and, depending on what is allowed for generating addresses, could leak some device info if manufacturers decide to make them default to a setup like MAC addresses which can link to specific brands or device models.

        2. You’re conflating NAT and firewalls. Yes, NAT basically acts as a firewall (by necessity of what it is doing – mapping ports, with unused tuples closed), but the right place to filter traffic both inbound and outbound is the firewall and those work just fine on IPv6.

          1. I don’t believe I am. Firewalls manage what traffic is allowed in and out of the network which is not what I am talking about.

            What I am saying is that a NAT prevents uniquely tying a device to its traffic external to the network. This hides info about the internal structure of the network.

        3. There are no security benefits from NAT, that’s a widespread myth. The few IPs reserved for private network are easily scannable (contrary to IPv6 nets). An NAT-Traversal is a real challenge for secure Peer-to-Peer protocols, so it makes your network less secure. What you need is proper firewalls on all of your devices.

          1. You can’t scan for anything behind a NAT unless you manage to compromise a device on the LAN first and perform your scan from there.

            A remote attacker can’t really do anything on your LAN unless they first compromise a device within it. No, that’s not a guarantee. You should still keep your services limited to the ones you need and those secured well. But that is certainly a big hoop the attacker has to get through. It’s not “no security benefit” Unless of course the attack is coming from inside the house…

            Many years ago I worked for a major ISP in tech support. We took so many angry, nasty calls from people who thought we were down but it was really just their shitty firewall software acting up. I really hate per-device firewalls after that.

            Maybe if I was connecting to something large like a campus-wide LAN at a large university or something. 10,000 CompSci students on the LAN might give me pause. Then I might agree with you. For most home users it’s just a way to ensure their stuff breaks more often and is harder to repair.

          2. You can’t just scan for private devices in a LAN unless you are already within the LAN.

            NAT does obscure what is in your internal network. And while “there is no security through obscurity” is mostly true, in real life there is some security gained from obscurity.

      4. I still get a public IPV4 address from my ISP. (Last I checked they didn’t even support IPV6 yet).

        All my devices on my LAN behind the NAT that I control is pretty ideal for me. I can forward the ports I want, let the rest drop. IPV6 looks like a lot of work, which isn’t helped by the fact that every guide I have read seems to be written by someone who’s target audience seems to be people that don’t need to read their guide in the first place because they already know it all.

    3. Of course you don’t have a need for a static address, everything for the last 20 years has been designed with no static address in mind. Having one opens new possibilities.

      Just look at all the “solutions” we have for peer-to-peer file transfer between people (dropbox, wetransfer etc.). These are only there because in general you can’t have two “clients” connect to each other. Or everything that NAT hole punching is used for, and the difficulty in finding out what the problem is when it doesn’t work.

      I don’t know what every device having a globally routable address would make possible, and neither do you. But “Everything I use now works without it, so we don’t need it” is a bit of a circular argument.

      1. It’s weird, but I don’t get to choose what other people do.

        I live in a world where many American companies choose to do geolocation for licensing content, content that I pay for.

        Personally, I actually enjoy watching streaming show, I actually don’t pirate as I make enough money to choose to contribute back to the industry such that, with a bit of hope and a prayer keep making content I enjoy.

        But geolocation on IPv6 tends to not work right now, and it tends to not work in ways that I cannot do anything about.

        I can’t fix the rest of the industry, I can’t fix everyone else usage of IPv6. I can setup my own network and do what I want with it, but if I have issues with services like this, there’s large swaths of users that have zero understanding of problems like this.

        Maybe it would work if I were American, it I’m not; nor am I trying to bypass geofencing, I’m perfectly fine with the content for my region.

        But it Does Not Work Reliably on IPv6.

  4. First day at a conference and the IPv4 WiFi was overloaded (probably ran out of IP addresses). Used my phone as a dongle over USB, but could only get an IPv6 address (phones get a real Internet-routable IPv6 address while IPv4 works via NAT). Figured out the IPv6 address for Google DNS and configured a default route and DNS server to get on the IPv6 Internet. Unfortunately the conference website wasn’t on the IPv6 Internet, but Raspberry Pi website was, so I was able to find equivalent (probably better) instructions on how to set up MicroPython on start hacking on the badge. IPv6 saved the day, but more websites need to get an IPv6 address…

  5. Actually, the article doesn’t say “many websites don’t support IPv6”, it says almost the exact opposite. GitHub, Reddit, Discord and Steam is basically the list in it’s entirety – at least for the tools that the author uses.

    The broad use of CDNs across services brings automatic support for IPv6 to vast swathes of the internet, so it’s only the small sites that are self-hosted without CDN (which have tiny audiences and so affect few people), or the huge sites that insist on doing their own CDN and not transitioning (which affect lots of people but they’re few in number and will inevitably cross over at some point).

    1. Three of these four are backed by huge non-self hosted CDNs are are global infrastructures. After all, steam with it’s half of billion of users, 140 million active simultaneously, can’t considered “small”

    2. those giant websites are hardly obscure!

      i would hate to chop off the long tail like that but the other thing is, my personal website is on ipv6 even though i haven’t done the work to accept ipv6 generally into my life (i’m too lazy to reconfigure my firewall and vpn). so i think it is something more than the ‘size’ of the institution in play.

  6. My ISP only provides dualstack lite (ig that’s what’s meant by nat64) access for me. And there have been occasions when I thought oh damn an ipv4 address would have been nice to have.

    But in the end you only really need that if you need a public ipv4 address, other than that it’s completely transparent. Even for hosting from home it’s fine most of the time because everything i’d access it from would be capable of ipv6 anyways.

    It’s a shame that renowned sites don’t even bother implementing ipv6. But then again, what does more layer of abstraction in the cloud jungle matter?

    1. That’s where I’m at too. For thing like self-hosted services it’d be incredibly useful and convenient.

      However, when I think about how even with a firewall it effectively uniquely characterizes what devices are in the internal network it gets bothersome.

  7. Once upon a time I was a CCNP, I think they tried to get too clever with IPv6, and should just have made IPv4 bigger, with legacy addresses having leading zeros maybe.

    Normals will already try to stab you in the eye with a fork before you get half way through explaining a subnet mask, I think if exposed to the start of an ecplination of IPv6 their fight or flight reaction will kick in almost immediately.

    1. Exactly what they did. Turns out that if you have one of these new addresses that eclipse the old address space you can’t communicate with the old gear.

      There is no solution. IPv4 has a full mapping into the IPv6 address space. It used to be exactly just the leading digits all zero, although that got deprecated.

  8. IPv6 is fine as it is. Usually it’s not the normal people that have been giving me problems, but rather fellow admins (including ones who claim to be network admins but don’t understand routing past default gateways and WAN/LAN Networks).

    There are tons of reasons why, especially for network admins like me, IPv6 is a million times better than IPv4. VPNs and SIP are so much better with IPv6, it’s not even funny. Setting up multiple routers is massively easier (no DHCP) as well. Troubleshooting? If a user gives me an 192.168.33.4, that tells me nothing. With IPv6 I actually know where he’s connected.

    Devices don’t have problems anymore. It’s mostly old fashioned admins and programmers who are either too stubborn or too stupid to move to IPv6.

    Subnets are also easier to explain in IPv6. The first half is for “where” and the second is for device itself. Always a /64 subnet (no need to explain that they can be bigger or smaller to 99% of people).

    That certain (big) sites are IPv4 only must be either ideology, laziness, or incompetence (or all three).

    NAT64/DNS64 is a great transition mechanism and I’ve also seen it in the wild (especially in Japan).

    1. “That certain (big) sites are IPv4 only must be either ideology, laziness, or incompetence (or all three).”

      You forgot the big one: money. I’m a network engineer (not at any of the listed sites, but a pretty big presence nonetheless, and currently IPv4-only). I’ve been working for the past year on turning up dual stack, and just got hit with, “we’re pausing the IPv6 project.” Why? We have a limited number of network engineers on our team, and a lot of things that need to get done, including a project to remove about 250 routers that are going end-of-life (guess what my new project will be?). Given the choice of turning up new features that solve problems for our users, replacing obsolete hardware with new equipment that still has vendor support and updates or rolling out another way of addressing devices on our network so our users can access the services they already can access, guess which project doesn’t get funded this quarter?

  9. Why do I have to scroll past everyone else’s comment to leave my own? 🤨
    IPv6 is kind of rubbish. Trying to manage a network with that memorizing those long tedious strings would be really annoying. It would be one thing if it was necessary, which is an idea that people tried to push, but not every connected device needs a unique IP address. Every network needs its own unique IP address, and then every device within that network has a unique IP only within that network.

    1. Why would you setup a network where you had to memorise the IP addresses? That’s what DNS is for. Your basic assertion “Every network needs its own unique IP address” already requires IPv6, it’s not possible on IPv4.

    2. “Trying to manage a network with that memorizing those long tedious strings would be really annoying.”

      It’s really not as bad as you would think. Humans are inherently lazy, and that includes network engineers who assign IPv6 addresses, so we tend to use strings that are easily memorable, as much as possible. You’ll see a lot of embedded words (“coff:ee00,” “dead:beef:cafe,” etc.) and long strings of zeroes in a lot of addresses. You’ll also see VLAN IDs or other information encoded in specific portions of the address. All these things make those long, cumbersome strings a lot more manageable.

  10. Disabling ipv4 completely actually not supported by current web infrastructure in majority of Internet coverage.. DNS, NTP, PnP… Yes, your software and hardware supports ip6, you application can create ip6 tcp connection, but ip4 is a standard fallback, multiple protocols primarily are served over ipv4, converted to ip6 where it can be done

  11. And Bill Gates thought a10mb hard drive was all you’d ever need. More unavoidable obsolescence junctures caused by dumb conservative thinking…16 more bits, no no this will do just fine mind ye young feller.

  12. About 10 years ago, I bought a new house and got a brand new DOCSIS modem from the ISP. I had, IIRC, a 500/50mbit connection. Transferring files between computers in my house was so slow. Turns out, the DOCSIS modem from the provider (all settings locked, only thing you could do was change the wifi name and password) automatically gave out IPv6 addresses and rerouted all traffic, including traffic between computers on the local network, over the internet. I’m glad I was able to buy a router, have the ISP put it in bridge mode, and have that function as my firewall and DHCP server. Haven’t been a fan of it since.

  13. Okay, many of the comments here mention the downfalls of IPv6.
    I was unaware of those. So, I need more education about those.
    Does anybody have a recommended link for me to check out in that regard?
    Yes, I can DDG, but if you have a good one in your bookmarks, would you kindly post it for me?

  14. One of the things that will keep it relegated to specialty use cases is that IPV4 has been around and workarounds to eliminate the “running out of addresses” concern are well known and work remarkably well. I began taking a Cisco CCNA course. They touched on the IPV6 and in later classes got into details. Yes, the benefits are there. But like everything else, why re-invent the wheel?

    If I have to dive into IPV6 I end up checking my work because I don’t use it very often. IPV4 I can do almost everything I need to do with little need to check reference materials to fix issues. I’m in the majority of web jockeys in that case.

      1. Externally for any company access to the world web community IPV6 makes plenty of sense.

        But when you’re up against vendors who won’t set up proper IPV6 protocols in their devices and you end up having to support an IPV4 presence in multiple nodes within a company, keeping everything at the IPV4 keeps things easy.

        For systems that could benefit from the extra layer of security IPV6 provides (finance, HR, etc.) they should be isolated and kept in an IPV6 island with translation when needed for external access.

        I am not sure if the public IP my company uses is on CGNAT or not (not my department to know). Not sure if our providers would divulge that information or expect the customer to ask to not be on a CGNAT setup and then they’d be able to charge a premium. Where IPV6 would clamp down on that practice pretty quick.

        In any case as long as the reasons to switch to IPV6 are mitigated by IPV4 the migration to it will take decades if at all.

      2. Externally for any company access to the world web community IPV6 makes plenty of sense.

        But when you’re up against vendors who won’t set up proper IPV6 protocols in their devices and you end up having to support an IPV4 presence in multiple nodes within a company, keeping everything at the IPV4 keeps things easy.

        For systems that could benefit from the extra layer of security IPV6 provides (finance, HR, etc.) they should be isolated and kept in an IPV6 island with translation when needed for external access.

        I am not sure if the public IP my company uses is on CGNAT or not (not my department to know). Not sure if our providers would divulge that information or expect the customer to ask to not be on a CGNAT setup and then they’d be able to charge a premium. Where IPV6 would clamp down on that practice pretty quick.

        In any case as long as the reasons to switch to IPV6 are mitigated by IPV4 the migration to it will take decades if at all.

    1. My cheap Netgear router runs OpenWRT, which obviously has an IPv6 firewall that blocks all incoming by default. That’s just normal behavior.

      I only get a dynamic IPv6 prefix, which meant I had to open a port to my server without specifying the prefix, and then use DDNS. But it works fine after all that, so no complaints.

  15. One of the things that will keep it relegated to specialty use cases is that IPV4 has been around and workarounds to eliminate the “running out of addresses” concern are well known and work remarkably well. I began taking a Cisco CCNA course. They touched on the IPV6 and in later classes got into details. Yes, the benefits are there. But like everything else, why re-invent the wheel?

    If I have to dive into IPV6 I end up checking my work because I don’t use it very often. IPV4 I can do almost everything I need to do with little need to check reference materials to fix issues. I’m in the majority of web jockeys in that case.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.