It’s now been a decade since the world officially ran out of blocks of IP addresses. In early 2011 the Internet Assigned Numbers Authority (IANA) announced that it had allocated the last block of IPv4 addresses and warned ISPs to start using the new IPv6 addresses. But here we are a decade later and not one of my clients has converted to IPv6.
Networks widely use IP addresses for devices in the network. Every cellphone, computer, network router, and modem is assigned an IP address so that ISPs can route traffic to the right device. The world adopted IPv4 in 1982. This is a 32-bit address and provided almost 4.3 billion IP addresses. That was enough addresses until 2011. IPv6 uses a 128-bit IP address. This provides for 3.4 trillion trillion IP addresses, which ought to carry mankind for centuries to come. Like most of us, I hadn’t thought about this in a long time and recently went to look to see how much of the world has converted to IPv6.
At the end of 2020, around 30% of all web traffic was being routed using IPv6. A lot of the biggest US ISPs have converted to IPv6 inside of networks. At the end of 2020, Comcast had converted 74% of its traffic to IPv6; Charter was at 54%. In the cellular world, both Verizon and AT&T are routing over 80% of traffic on IPv6 while T-Mobile is close to 100%. Around the world, some of the biggest ISPs have converted to IPv6. India leads the world with over 62% countrywide adoption at the end of 2020, with the US in fourth at over 47% adoption.
But the big caveat with the above statistics is that a lot of the big ISPs are using IPv6 inside the networks but are still communicating with the outside world using IPv4. After all of the alarms were sounded in 2011, why haven’t we made the transition?
First, carriers have gotten clever in finding ways to preserve IPv4 IP addresses. For example, small ISPs and corporations are using a single external IP address to identify the entire network. This allows for the assignment of imaginary IP addresses inside the network to reach individual customers and devices, much like CLECs have reduced the number of telephone numbers needed by switching internally with imaginary numbers.
There is an extra cost for any ISP that wants to fully convert to IPv6. IPv6 is not backward compatible with IPv4, and any company that wants to route externally with IPv6 needs to maintain what is called a dual stack, meaning that every transaction in and out of the network has to route using both protocols. This adds expense but more importantly slows down the routing.
It’s also impossible to convert a network to IPv6 until all devices using the network are IPv6 compatible. This becomes less and less of an issue every year, but every ISP network still has customers and devices on the network that are not IPv6 compatible. Those customers still using a 12-year old WiFi router would go dead with a full conversion to IPv6. This is one of the primary reasons that the big ISPs and cellular carriers aren’t at 100% IPv6. There are still a million folks using old flip phones that can’t be addressed with IPv6.
There is a definite cost for not converting to IPv6. There is a grey market for buying IPv4 IP addresses and the cost per IP address has climbed in recent years. The typical price to buy an IPv4 address ranged from $24 to $29 during 2020. With all of the grant money being handed out I expect the creation of a number of new ISPs in the next year. Many of them are going to be surprised that they need to spend that much to get IP addresses.
The main reason that the conversion hasn’t happened is that nobody is pushing it. The world keeps functioning using IPv4 and no ISP feels threatened by not considering the conversion. The first small ISPs that take the plunge to IPv6 will pay the price of being first with the technology – and nobody wants to be that guinea pig. Network purists everywhere are somewhat disgusted that their employers won’t take the big plunge – but even a decade after we ran out of IP numbers, it’s still not the right time to tackle the conversion.
I have no idea what will finally set off a rush to convert because it inevitably will happen. But until then, this will be a topic that you’ll barely hear about.
Everyone dances around the big reason IPv6 never took off – it’s dumb, non-intuitive, and nobody wants to dumb it down to the level that consumers can understand.
Consider setting up an IPv4 firewall using something like iptables. You know, for example that everything in the 192.168.0.X range (or maybe 192.168.1.X) is on your home network and should be allowed to talk to everything else in that range. If you want to connect to your home network from your office network, you can probably find out that every computer in your office will appear to be coming from the same IPv4 address when it reaches your home network, so you can put an exception in your firewall to allow incoming traffic from that single IP address. I won’t say this kind of networking is super easy, but it’s not so complex that it takes reams of paper to explain it.
But then you get into IPv6 and suddenly all bets are off. How do you define your home network? How do you allow any traffic from existing and future machines on your office network if each has its own individual IPv6 address? If someone in Nigeria is trying to hack your system, how do you block all incoming connections from the subnet of addresses assigned to his provider or to the entire country? In IPv6, either these things are not intuitive at all, OR nobody feels it’s worth the effort to write a very clear and simple guide on how to do these things. Faced with the choice of doing something hard, which is struggling with a new thing you do not understand and that no one you know can explain, or doing the thing you have always done which is relatively easy by comparison, which do you suppose most people will do?
In my opinion what should have been done is to incorporate existing IPv4 addresses into IPv6. So, for example, add another octet at the start and end of each address, so an existing address like 192.168.0.1 would become 001.192.168.0.1.001. This would have both allowed for new major networks (as when new providers need number assignments) but would have also allowed for new endpoints – but most importantly it would not have required people to make a major shift in how they currently understand networking. I have often said that computer people frequently shoot themselves in the foot by insisting on the perfect when they should be willing to accept the good enough. The “good enough” in this case would have been to build on people’s current understanding of networking, not to throw everything out and start from scratch with a system that virtually no one other than a few networking geeks really understand.
Have your already written about OTARD and what is happening for companies putting 5G antennas on rooftops? I would like to know more about how this is competing with mobile companies and if this is being sold to the rural communities to get wifi currently. THanks