IPv6 is Over
I’ve flipped or flopped a couple of times on IPv6. I’m doing it once again…IPv6 is over, done, not going to happen in any significant way.
I’m not saying that it won’t be developed. There have been pretty decent stacks out there for 10 years. I’m not saying that it won’t be deployed. It’s been deployed in overlay and native networks for quite a while. Despite a major push from the US government for a mandated overlay network, IPv6 is not going to make a noticeable dent in the commercial IT space.
I was at a DARPA meeting in 1993 where the two candidate protocols for IPv6 were debated. By 1997 the IPv4 address shortage was on everyone’s minds, and I took part in an effort to mitigate it without a forklift upgrade to v6. From 2000 to 2007 I taught a course that included a two-hour lecture on v6. I always stated that IPv6, “was definitely going to happen.”
Today, I’m not so sure.
There is just one motivating factor for v6: a large address space. If you surf around the web you might find arguments that v6 also has better autoconfiguration, security, and mobility support. However, v4 does fine with DHCP, IPsec and Mobile IPv4. The v4 address space crunch is largely alleviated by NATs, which are not the disaster they they were made out to be. I can’t remember the last time I couldn’t get an application to work because of a NAT. Sure, NATs make application development and deployment a bit more complicated, but will native IPv6 deployment be simple?
Despite the fact that IPv6 is commercially available in routers and end-host operating systems, it does not have the mileage that v4 does. To my knowledge, it has never been tested under extreme real-world scenarios, in large networks. The cost of deploying IPv6 includes hardware and software upgrades, retraining IT staff, downtime of mission-critical systems, and debugging all of the strange new issues and problems that inevitably result from a large-magnitude change. Compare this to the cost of staying with v4, and few organizations outside of governments and universities can afford to switch to v6.
Sure, v6 will likely continue to be deployed in small islands. We might even find that a cellular operator or two deploys v6 phones. If they can make the v6 upgrade transparent to the user, more power to them. But any business is cost-conscious and profit oriented, and will likely spend the extra money needed for v6 on a profit-generating value-added feature that is not transparent to the user.
So I’m out on a limb here and only time will tell if I’m right, but I’m taking the position that in 20 years IPv6 will not have a significantly larger market share than it does today.
January 5th, 2008 at 9:51 pm
Mike -
Interesting position to take… While it is true that NAT eliminates much of the readily apparent pressure to move to IPv6, it doesn’t address the underlying requirement that ISP’s have to growth hundreds of new customers while only adding 1 route to the global default-free routing table. Once the RIR’s are no longer able to provide IPv4 blocks from the free pool, ISP’s can certainly continue to connect customers via smaller and smaller IPv4 address assignments, but that will cause a dramatic increase in routing once those IPv4 address blocks are either customer-supplied or obtained from non-RIR sources.
The ability to assign new address blocks to ISPs gives us both address space for new customers, *and* ability to add customers with very little additional routes in the worldwide routing table. NAT only addresses the former problem.
/John
January 7th, 2008 at 8:57 pm
There’s still little or no incentive for the ISP customers to move toward IPv6. It might make sense for ISPs to adopt v4 over v6 tunnels as long as it is transparent to their customers. But that still doesn’t solve the v4 route problems.
January 26th, 2008 at 7:15 am
[...] on IPv6: A few weeks ago, David Siegel of Global Crossing looked at some high-profile websites and found [...]