The Web Still Requires The Internet

At the IETF (Internet Engineering Task Force) meetings going on this week in San Francisco, Internet architects admitted that IPv6 isn't backwards compatible with IPv4. Uh Oh.

To many in the Web 2.0/Internet 2.0 world the Internet is a black box. It just works, it has always worked, and to the best of our knowledge it will continue work. But as Nemertes noted in "Internet Interrupted: Why Architectural Limitations Will Fracture The 'Net" to those closely following the architecture of the Internet, something is very very wrong. IP addresses that identify servers and users are being depleted, with many predicting shortage in the next 2-3 years due to continued growth in users and devices. IPv6 was designed to fix this shortage, but IPv6 exacerbates another growing problem. The growing number of devices and users is straining the architecture used to track locations and route information between users. IPv6 actually makes this problem worse by creating even more addresses to track. IPv4 address shortages will further strain the system as address blocks are broken up to squeeze every last available address out of the system. The hierarchical model of routing is quickly becoming a thing of the past, leading to unsustainable growth in Internet routing tables. In response, there's growing talk of using private addressing within the Internet, breaking the very basis of the Internet design of smart end-points with a dumb network core.

What does this mean for Web 2.0 and Enterprise 2.0? Innovation and performance will suffer. Should service providers deliver private addressing with network address translation, address translation gateways may impact performance of real-time applications. Providers could also use these gateways to restrict competitive applications. Address shortages could lead to an open auction market for address space, meaning an increased cost of operating Internet-connected networks. Could we potentially see a fracturing of the Internet with a return to walled-garden services as we saw a while ago with Prodigy and AOL? Maybe.

So what's the fix? Absent a "cut over day" in which the entire Internet is forced to migrate to IPv6, no widely accepted solution exists. A number of engineers are proposing a new architecture based on separating out user locations from routes, but even this approach (known as LISP) has significant scalability concerns. Others, such as Dr. John Day in his book "Patterns in Network Architectures" argue for a completely new architecture based on services rather than locations. Within the IETF and other forums, as well as among vendors and service providers, debate continues. Those who's business model relies on the continued operation of the Internet would do well to pay attention to what is going on inside the "black box."