Recently, I’ve had the opportunity to not only do some research around our new Internet addressing protocol but also become fully IPv6 enabled. Not me personally - I’ve yet to get that Ethernet implant sorted out - but my internet connection and the devices that hang off it are now IPv6.

Yes, it’s 2011, and I’m finally using IPv6 natively and not via a tunnel over IPv4. The new protocol comes courtesy of my ISP, Snap, alongside a rare beast: a Fritz!Box VDSL2 router that understands IPv6.

You might think that IPv6 is very slow in coming and you’d be right. Shifting to the new protocol has become an enormous task, one that grows every year as the older IPv4-based internet expands at a huge rate.

Right now, the IPv4 network is larger than the four billion or so addresses it contains. That’s because of Network Address Translation or NAT. This was a hack devised in the 1990s to deal with address exhaustion - yes, that uncomfortable problem has been known for a very long time.

NAT ”hides„ a huge number of networks behind one or more internet-routable addresses. If you have multiple gadgets running on a LAN managed by a single router in your home using DHCP, that includes your network. This saves precious IPv4 addresses, but it also breaks the end-to-end principle of the internet and leads to undesirable effects such as port exhaustion: each TCP and UDP address is limited to a maximum of 65,536 ”ports„ over which the actual bytes flow, so if you have a single, routable IP address that serves thousands of NAT’ed ones, performance on sites like Google Maps that open hundreds of connections will suck.

NAT also adds complexity, can be difficult to scale, and it has arguably slowed down the uptake of IPv6.

Due to the big address space in IPv6, there’s no need for NAT. Until all the billions and billions of devices behind a NAT gateway become IPv6-enabled - and some won’t ever, as they’re too old - that address saver isn’t going away however, and nor is IPv4.

This is one of the big challenges with rolling out IPv6: not only do you build a new internet and give existing devices a new address; you also have to make it coexist with the old IPv4 internet which is still growing, thanks to NAT.

In practice, the dual IPv4-IPv6 stack internet leads to some annoying problems. Software and operating systems are normally designed to use one address type or the other. If a connection over IPv6 is slow to establish or fails completely, falling back to IPv4 can take many seconds. There’s no indication to users what’s going on, either. So, for example, in your web browser, you may just experience blank pages and sometimes, time-outs.

A new algorithm called Happy Eyeballs is at the IETF Internet Draft stage, and it attempts to fix the dual-stack failover problem by switching over faster to, say, IPv4 as well as remembering which connections don’t work.

Another issue that I noticed is that IPv6 network paths aren’t as optimised as IPv4 ones. Google over IPv4 is served to me via a local cache, a mere 6ms away. Over IPv6, it the latency is anywhere from 60 to 330ms. Not a show-stopper and it will be sorted out eventually, but there are many similar cases featuring large content delivery networks and caches.

In simple terms, people aren’t going to want IPv6 if time-outs occur when they browse popular sites like Facebook, or if the performance is really slow due to high latency and constricted pipes.

Even if your ISP does provide IPv6 connectivity, getting it up and running isn’t easy. The vast majority of new operating systems know about native, tunnelled and 6to4 IPv6. Your smartphones and tablets most likely do too. However, the DSL modem that connects you to the internet probably doesn’t, or only badly. Finding residential and small business network gear that supports IPv6 can be difficult, unfortunately.

On the security side, it’s worth noting that dual-stacks can lead to end-to-end connectivity being established over IPv6 to vulnerable devices. IPv4 firewalls are a dime a dozen on consumer gear but IPv6 ones are an unknown quantity.

The Fritz!Box I have comes with an IPv6 firewall, but it’s difficult to configure. What’s more, while the admin GUI on the Fritz!Box provides ample info about what’s going on with the NAT’ed IPv4 network, it only shows the IPv6 address on the router’s internet-facing interface. More work needed there, I’d say.

Even with all the problems mentioned above, IPv6 is worthwhile and yes, you should nag your ISP about having it turned on. The only way we’ll get more choice of residential routers, improved network paths and performance and more is through more IPv6-enabled customers. It will be something of a slog though, so be patient.

- Juha Saarinen