Now we know: The secret is to get John Donovan off the teleprompter.

The AT&T senior executive vice president of technology and network operations was delivering a stiff keynote at the Open Networking Summit Tuesday, giving a PR-sanitized description of the carrier's new Domain 2.0 policy. (People said it's the exact speech he delivered last week at Mobile World Congress.)

But when he got off the teleprompter to start the Q&A session, Donovan came alive. Suddenly free to use terms like "southbound" and "APIs," Donovan gave generous and spirited responses to a host of audience questions — and in the process, gave us new insight into just how strongly Domain 2.0 is intended to change AT&T.

Donovan is a believer. He wants to change not only AT&T's network but also its internal policies and philosophies. He embraces SDN "because there's no army that can hold back an economic principle whose time has come," as he told the crowd. "What I'm talking about is radically reshaping the entire wide-area network."

Here's some of what he shared.

Domain 2.0 doesn't ignore the installed base.

"What we're looking for in 2014 are beachhead projects that can move us from an old Domain 1 architecture to a Domain 2 architecture. We're calling it D1.5," Donovan said. The idea is to fit controllers onto AT&T's current platforms, extending the lives of those platforms.

The beachhead projects are meant to happen quickly. AT&T has already identified "a half dozen" of them, Donovan said. Moving into 2015, AT&T hopes to start introducing applications that were born in the cloud, a buildup that should lead to a "crescendo effect," Donovan said.

AT&T really is serious about working with startups.

AT&T indicated as much by selecting Tail-f and Affirmed Networks as Domain 2.0 vendors, but Donovan was adamant that these weren't token appointments. Through facilities such as the AT&T Foundry sites, the carrier has met with 1,200 carriers in the last 36 months or so, Donovan said.

And AT&T has already engaged some of them.
"Intucell last year was bought by Cisco. We met them when they were 4 people, and we took a chance with them," he said. "They took out 20 basis points of dropped calls on a network that was carrying 15 billion calls a day."

The wireless network will be virtualized from the core outward.

The wireless access network could benefit from SDN-like architectures (and in fact uses some similar principles already), but converting it means too much work for too little immediate gain, Donovan said.

So, the first parts of the wireless network to be converted will be those that involve "fewer data centers and are latency-tolerant," he said. "Things like policy and authentication and all those core-related activities."

The access portion of the network would come later. "In the beginning those will look more like RNC pooling," which Donovan said he could count as an example of AT&T already virtualizing its network, but it really doesn't represent a new way of doing things. Substantial virtualization in this part of the network won't start for a couple of years, he said.

The old OSS regime must die.

AT&T plans to retire more than 1,000 of its current OSS applications. In their place, the carrier hopes to handle operations such as billing and provisioning through parallel processes "as opposed to the inhibiting process that defines everything that we do," Donovan said.

Security vendors need to break out of the appliance mindset.

They're addicted to the business boost of releasing new boxes — but the SDN/NFV model is meant to do away with that cycle.

"Despite a lot of the security firms being smallish and very nimble, we were, I think, the first ones to really push them into a model that gets them out of the appliances. Because if there was ever an industry that had more of an appliance model than telecommunications, it might have been security," Donovan said. The alternative AT&T wants to see: Deploy software (which allows for always having the latest feature) and get these vendors' wares to talk to one another to improve the overall intelligence of the security network.

Orchestration of many distributed controllers is the hardest part of carrier SDN.

And it's far from solved. Donovan got asked whether any vendors have any orchestration offerings that suit AT&T's goals. "We've had a lot of very interesting dialogue, we have nothing ready to deploy," he said.

AT&T is on the lookout for some answers, though. "That is a massive software project," Donovan said. "It will be a highly complex, multi-participant approach. We're going to have to even write some of it ourselves."

To that last point, he added: "Our backbone today, we manage over 4 million routes, and that's larger than you can get from any control plane, so we wrote our own. So, we're not brand new to this game."

Donovan's commitment is palpable, but he still has to get these changes accomplished while pushing against a large company's decades-long habits. "Part of what he has to do is dismantle the Bell system a little bit," said Francois Locoh-Donou, senior vice president of Ciena's global products group (and someone who's had a lot of experience dealing with carriers). Donovan is a former championship-level boxer, so he might be equipped to take on that challenge.