A few weeks ago, David Siegel of Global Crossing looked at some high-profile websites and found that none of them have made the switch to IPv6
, the supposed replacement for today's 32-bit Internet addressing scheme. The IPv6 protocols have been finalized for a decade, and major operating systems have supported it for several years, so one would expect Internet-savvy companies like Google and Microsoft to have started running IPv6 versions of their sites. But it appears that so far, nothing of the sort has happened. Indeed, progress toward an IPv6-based Internet appears to be at a virtual standstill. The situation becomes less mysterious when one realizes that the primary rationale for the switch to IPv6 -- the exhaustion of the IP address space -- is basically bogus. It's true that if Internet governance authorities continue handing out IP addresses, they'll eventually run out. But the best solution to this isn't necessarily a massively disruptive transition to a totally new addressing scheme. It may very well be a lot cheaper to continue working within the constraints of the existing address space. Technologies like NAT
allow many users to share a single IP address. And Internet governance bodies can facilitate the creation of a robust market for unused IP addresses, so that those who need additional IP addresses can easily purchase them from someone who has more than they need. For example, Apple, Ford, General Electric, and several other Fortune 500 companies currently control blocks of 16 million IP addresses each. These companies should be given a straightforward way to auction off the unused portions of their blocks for the use of other Internet firms. There would be plenty of IP addresses to go around if firms had a financial incentive to give up unused addresses.
An interesting analogy here is to the continued health of the x86 architecture that now lies at the heart of virtually all desktop and notebook computers. For decades, people have been predicting that x86 was on its last legs because it is a clumsy, register-starved architectures. In the early 1990s, everyone expected RISC chips like the PowerPC and Alpha to clobber x86-based chips in performance. In the late 1990s, Intel itself bought into the hype and attempted to push the computer industry to its brand new Itanium architecture. Yet the predicted demise of x86 never happened. The x86 platform had extremely broad support in the industry, and it has turned out that the costs of limping along with a crappy architecture are smaller than the costs of switching to an entirely new one. I think something similar may be true of IPv4 addressing. As cramped as its address space may be, the costs of switching the entire Internet to a new addressing scheme will be enormous, and the benefits are far from obvious. So my prediction is that IPv6 will continue to be "just around the corner" for the foreseeable future, as the bulk of network owners find it more affordable to just make do with the addressing scheme they're already using.