In many ways, 2005 was a "so what?" year for networking. After all, anyone who expected major breakthroughs on a number of key issues ended the year sorely disappointed. There's been little progress on IPv6 adoption, the US is still claiming it owns the Internet, wireless networking made little progress on fronts political or technical, and IP telephony remains hot - though whether VoIP's future lies in hardware, software, or both is still a topic that will drive a conference panel, or three, for years.

Nevertheless, the year wasn't a complete loss. For starters, PBX vendors such as Avaya, Cisco and Siemens have added improved SIP hardware to their product lines. Although complete cross-vendor interoperability is still a pipe dream, it's now possible to deliver basic functionality without pulling out all of one's hair.

On the 10 Gigabit Ethernet front, switch vendors kept increasing their product offerings. Much of the buzz surrounding 10Gig revolves around its potential as an infinitely less-expensive substitute for SONet (Synchronous Optical Network) in metropolitan networking schemes. Analysts, however, are pointing out that ever-increasing backup requirements in the datacenter might make a more plausible business case.

But the networking technology that defined 2005 is traffic optimisation. After all, few application-layer protocols were designed to run across the WAN, especially when they were developed with a certain monopoly OS in mind. The trick in most cases is to keep redundant requests on the local network, passing on only the meaningful parts of the TCP-level conversation.

Fortunately, creating the network equivalent of carpool lanes with technologies such as QoS actually works. Prioritising traffic as a set of higher-layer protocols isn't itself a new idea; the problem has always been implementing the concept, followed by the actual management of the traffic - including the inevitable breakdowns. The initial efforts in recent years led to single-role appliances, which have met with initial warm receptions, at least in the cases where the traffic proved amenable to the individual boxes' specialties.

But the one-box solutions that companies such as Juniper, Riverbed, and Swan Labs are turning out appear to be the wave of the future. By combining caching, compression, and acceleration - at both the application and TCP levels - into a single chassis, these products appeal to the natural desire to reduce complexity, even if they might seem to violate a cardinal rule of network management: Isolate it.