One great thing about the A-IMS architecture that Verizon Wireless recently unveiled is that it will improve CDMA cellular carriers’ ability run end-to-end VoIP over the fancy new IP networks they’re building. An even better thing may be that it will allow them to avoid running VoIP over those same IP networks. After all, no operator wants to let a little thing like network buildout force it to offer a service before it’s ready to.

Verizon says A-IMS, which stands for Advance to IMS, will fill a lot of gaps that IMS (IP Multimedia Subsystem) technology currently leaves open. It developed the design in collaboration with Cisco, Lucent, Motorola, Nortel and Qualcomm, its main equipment suppliers. A particular goal of the effort, it says, was to ease the transition from current networks to the all-IP networks of the future. IMS will be a key ingredient in ensuring that fixed and mobile services of all kinds can work together on such networks.

The new architecture will offer several useful technical advances, according to Verizon. It will take a comprehensive, network-wide approach to security. It will treat SIP and non-SIP-based applications equally. And it will provide end-to-end management for VoIP applications running over IP mobile networks, to ensure quality of service.

Verizon sounds like it has good reasons for wanting these new capabilities. Take support for non-SIP applications. "In the real world we have a pretty broad set of products," says Ed Salas, VP for network planning and strategy. "A lot of them are…non-SIP, and I would at a minimum expect that as we move forward and evolve the core of our network, we’re going to want to be able to interoperate and support all of our legacy applications, as well as any new ones that may be SIP-based. It’s also my sense that development communities are going to continue to develop in some of these non-SIP areas. So we’re trying to create a framework that enables a sensible coexistence of both types of applications and the need for them to perhaps interplay."

Its arguments for end-to-end VoIP support sound equally compelling. "VoIP is probably among the most challenging applications a wireless operator can support, if you’re going to do it well," says Salas. "We needed to have an architecture that accommodated the end-to-end quality management of a given session, not just managing quality in segments." The segments requiring coordinated management included the air interface or bearer, the backhaul and transport network, the application and the "distant end," he explains. 

Looking at the proposed advances side-by-side, however, highlights a fascinating fact: that making it possible to run non-SIP applications over a network as it evolves to IMS by definition also makes it possible to continue running non-VoIP mobile phone services almost indefinitely. And cellular carriers might have good reasons to shun VoIP as they build out their shiny new multimedia-capable IMS infrastructures.

One obvious reason is that they’re doing fine selling voice and high-speed data services separately. Currently, notes In-Stat principal analyst Keith Nissen, they may charge $60 to $80 dollars or more per month for voice, and another $60 for high-speed data — but only to those same voice customers. In short, wireless broadband data serves as a lucrative add-on to an already lucrative voice service.

But once they start offering VoIP over broadband cellular technologies like cdma2000 EV-DO Rev A, the high-speed data comes first, and the voice rides on top of it. When that happens, their entire business model threatens to collapse.

For one thing, it becomes a lot harder to argue that customers should pay for voice calls on an expensive per-minute basis, when those calls are traveling over the kind of IP network users typically pay a flat monthly fee to access. It also becomes harder to argue that they should pay a premium for using the broadband network for things like wireless Web surfing, since they’re already paying for it to get their voice service.  

Yet cellular carriers will have to successfully make — or perhaps finesse — such arguments. If they don’t, they could find themselves reduced to the dreaded fate of serving as mere fat IP pipes, able only to charge a flat rate for unlimited VoIP calls over their networks. And that would be a fate almost worse than bankruptcy. As In-Stat’s Nissen puts it, "If you’re raking in money forcing people to buy your phones and to pay 80 bucks a month for 800 minutes of time, then why would you want to go to IP where everything is flat rate?" Sticking as long as possible with TDM cellular service, which requires no new infrastructure investment, would make more sense.

It all means that moving their voice services to end-to-end IP, i.e. to VoIP, while keeping their traditional minutes-based business model intact, may be the most important — and trickiest — challenge cellular operators will face over the next decade. Doing it successfully take all the skill, persuasiveness, discretion, luck and timing they can muster.

That’s why cellular carriers may find it wiser to keep a low profile about the entire subject of using VoIP in their networks and services. It’s also why they would be better off making the various arguments, and the transition, on their own terms and at their own pace, rather than having it dictated by their IMS buildout. The A-IMS architecture would allow them to do just that.