Services called IP Centrex, for want of a better name, are still sorting themselves out, and the name itself is no small part of the problem. Even if providers of hosted VoIP services scrupulously avoid using the C word, the mind-set is still there, dragging along the baggage of the past.

Examine hosted IP telephony solutions, and you can see two fundamental problems that stem from technological constraints and cultural issues. The first problem is the old Centrex paradigm that puts a $10 million piece of equipment in the IP cloud and shares it among customers. From the carrier's point of view, no equipment at the customer premises equals maximum efficiency. However, voice is now traveling over a best-effort network rather than dedicated circuits. Upstart carriers such as Vonage can deliver only two- or three-nines availability, which limits their utility in the enterprise market and prevents them from offering feature-rich services. Big guys such as AT&T, which control the network, might get the availability up to four-nines, but that is still far short of the five-nines capabilities of traditional phone trunks.

The centralized Centrex model also is a bad fit for distributed IP networks because it fails to exploit their fundamental resiliency. If that big central-office switch in the cloud goes down, all its customers lose their phone service. In contrast, when voice services are based on distributed IP PBX switches at the customer premises, a switch outage affects only one customer - or even just a single customer site.

The second problem with hosted VoIP is the carriers' ability to manage hosted voice services when the delivery mechanism at the customer site is the enterprise LAN. It's one thing to be an expert on the cloud side, and quite another to look into a customer's Ethernet infrastructure, see which component is causing a problem and fix it remotely.

Because there is already equipment at the customer premises that must be managed, there is a good case for an edge Centrex model that replaces the $10 million switch in the cloud with inexpensive switches at each customer site. This provides creative young carriers - the ones that are likely to drive new services - with some major advantages. They can expand incrementally as they add customers and those customers grow. They also can deliver a highly available service without requiring a high-QoS WAN link, by using the public switched telephone network for failover. And they avoid the "touch SS7 and you're regulated" problem. When hosted IP telephony is based entirely in the cloud, service providers have to file plans with the FCC and hire lobbyists, and will ultimately spend money to defeat 'Net neutrality instead of using it to develop new products and keep customers happy.