Most anyone that considers IPv6 implementation and proposes implementation solutions looks at it in terms of IPv4/IPv6 coexistence. When the IPv6-capable devices in a network are dual stacked, coexistence is relatively simple: The dual stack devices can speak to both IPv4-only and IPv6-only devices. The problem with dual stacks is that they require both an IPv4 and an IPv6 address, and that misses the fundamental point that we’re deploying IPv6 because we are quickly approaching a time when IPv4 addresses are no longer available. Dual stacking would have been the right approach to transition five years ago, but it is less and less viable as new IPv4 addresses become unavailable.
Dual stacking is still the right approach to existing devices that already have IPv4 addresses and can acquire a new IPv6 address (either with existing capabilities or via a software upgrade), but its not a good solution for new devices and networks that require new addresses.
The larger problem is how to make IPv4-only and IPv6-only devices speak to each other. That requires translation, a prickly problem in terms of complexity, scalability, and security.
Behind the coexistence discussions is a principle that IPv4 and IPv6 coexistence will be necessary for the foreseeable future. That’s the right approach to take, but I think that there is a point in the not-too-distant future when IPv4 will be killed off. My personal opinion is that this will happen somewhere between 2014 and 2020.
The reasoning behind this is that no one wants the complexity of having to run two versions of IP in their networks. Given that IPv4 addresses will be depleted around 2011, as I’ve written about repeatedly on this blog, IPv6 deployment will beyond that date begin growing exponentially. At the point where IPv6 either is approaching or has become the majority protocol, networkers will begin actively removing IPv4 in order to simplify their operations.
Over the past few years, discussions of IPv6 implementation have focused on connecting IPv6 devices or IPv6 sites across an IPv4-only “cloud”.
I’m at IETF 72 in Dubln this week, and I’m sitting in the v6ops working group meeting as I’m writing this. There’s a shift in the focus of coexistence discussions to connecting IPv4 devices or IPv4 sites across an IPv6-only cloud.
This is an early indication of a reorientation toward eventually bringing IPv4 to end of life, by pushing it to the edges of the IPv6 network. I like hearing this subtle change of perspective; it fits with the standard approach we take in any migration project, when some legacy technology is being replaced by a newer technology: You don’t add the new technology at the edge and “push” it toward the core, you add the new technology in the core and push the older technology toward the edge – and eventually out – of the network.