ANALYSIS: Why Internode's IPv6 product makes sense

South Australian ISP Internode has started selling an iPv6 service. But as Richard Chirgwin writes in his debut as our new Networking Editor, the service makes more sense than you would imagine, given anemic IPv6 uptake.

IPv6 could be described as the world's longest yawn: a standard that was born ten years ago, that addresses a real and looming problem, and yet adoption falls far behind the need.

The story is simple enough: IPv4 doesn't have enough addresses, because the Internet's too successful and its early address assignment schemes can't cope with its subsequent growth. The response, when Internet adoption started to take off in the 1990s, was to develop a new addressing scheme. But since then, no amount of advocacy in favour of IPv6 has managed to overcome the inertia of the installed base.

That makes Internode's decision to launch a no-extra-cost IPv6 service both innovative and risky. The perceived risk is that the Adelaide-based ISP will have spent money enabling a system that its customers won't use.

However, that's not how founder and CEO Simon Hackett sees the new service. His argument is that since the move to IPv6 is inevitable anyway (the only question is when), putting a service in front the customers was going to have to happen anyway.

What it needs, however, is a spur to the flanks. There is, Hackett said, a “chicken and egg” problem: either the users move first, in which case they have an IPv6 address scheme that's not supported; or the service providers move first, in which case they're offering services that aren't yet in demand among their customers.

The lack of customer demand is, paradoxically, compounded by the Internet's characteristic innovativeness. As well as IPv6's bigger address space, he explained, “there were a lot of other enhancements put into IPv6 that weren't then available in IPv4. The challenge is that every other thing that was going to be new in IPv6 has since been backported to IPv4: things like IPSec that were built into IPv6 but were easy enough to slap onto IPv4.”

As a result, one of the big selling points of any new technology – that this technology will give you capabilities you can't get any other way, a big issue for enterprise IT managers – is lost to IPv6 advocates.

“So until the addresses actually run out, the general population won't care,” Hackett said (and if that sounds like a 'peak oil' discussion to you, you're not alone).

Moreover, network address translation (NAT, aka the bane of TCP/IP purists but used by everybody down to the home network gateway user) has masked the problem of address exhaustion for years. Still, Hackett said, the long-overdue day of reckoning is going to arrive sometime in the next five years.

Hence the decision to launch now: “When the addresses run out, people are going to suddenly care a lot ... and the idea is to have the deployment challenges largely solved by then.”

Since IPv6 is already built into modern operating systems – Mac, Windows and Linux all come with IPv6 stacks – it might seem to most people that the deployment challenges have been addressed by OS vendors. That's true, but most people don't connect the home PC directly to the Internet. There's always equipment like the ADSL modem between the user and the network.

That's an issue that Internode addresses by preferring vendors that have IPv6 support. But the other end of the deployment challenge is at the service provider end. Hackett cheerfully admits that there's plenty of cost and no revenue to be had from launching a service.

“There are significant costs for deploying IPv6 for an ISP. All the routers are dual-stack: they're running IPv4 and IPv6 – they're not doing twice as much work, but there are two protocols to manage. Today, there's no direct income related to doing that.”

Similarly for the corporate customer – or even the home customer – it's likely that moving to an IPv6 implementation on Internode today could still be “more of a challenge than a benefit ... because you're firewall has to be able to understand v6 as well as v4. But the reason you should be doing it is to get the hang of it, so that when it all happens in a great hurry over the next three to five years, you have some experience doing it.”

He counsels against a forklift upgrade, however: rather than moving the whole corporate network to IPv6, he suggests a test case as the first move.

“You might want to sign up with an Internode service and get the hang of this stuff on a machine over DSL on a desktop somewhere, just to get used to it. So you could treat it as an opportunity for a soft ramp-up.”

Given the looming – some might say gently looming – crisis, it's surprising that security researchers such as Joe Klein of Command Information have criticised the roll-out of IPv6 in desktop operating systems, saying that having the protocol enabled when users don't know it's there creates a security hole they don't know how to fix.

Hackett, however, is less concerned on that score. “One of the incidental accidental benefits ... is that most consumer ADSL routers don't support IPv6 in their production firmware yet. So you've got this air-gap in that in that since you're router doesn't support v6, nothing's getting routed to the world that way.”

And into the future, his hope is that the security settings users have put into place, whether in a home router or an enterprise firewall, will be mirrored into the IPv6 world.

In spite of rampant disinterest in IPv6, Hackett suspects commercial interests may eventually come into play in encouraging adoption of the protocol. Historically, address assignment has been on an as-needed basis, with an explicit prohibition on address trading. However, as addresses become exhausted, some kind of address trading (whether licit or otherwise) will probably come into play.

So as time goes by, enterprise customers will probably find themselves making a choice about how they spend their money: will they buy into some kind of IPv4 address trade to get the addreses they want, or will they spend the money on learning how to use IPv6, and get the addresses for free with their Internet access?

Internode's bet is that customers will opt for the new over the old.

Read more on Data centre networking