Many Devices, No Boundaries, One Provider

As the top techie at the top cable operator in the U.S., Comcast’s chief technology officer David Fellows needs no introduction. After all, his arrival at Comcast in 2002 came after a long career, including senior positions at Scientific-Atlanta, Siemens, AT&T Broadband and the old Continental Cablevision. In an interview, Fellows expounded on the next phase of Comcast’s—and cable’s—evolution. To sum up Dave’s World: Convergence is here (and for real this time). Telephony will actually get fun. And integration between different hardware and services will rule the day. What are the biggest technological challenges at Comcast now and in the near future? David Fellows: The thing that is on everyone’s mind here is what we call Comcast Digital Voice telephony service. We’re getting that in place, scaling that service, making a good first impression, getting at the 20% of people who are historically upset with their [telco] incumbent no matter how good a job their incumbent is doing. Getting into that phone service as the third leg of that service tripod—or maybe we’ll call it a quadrapod later. But it’s that convergence. So it’s taking some ground out of that voice space for later integration with data and video services. We have spent more time on infrastructure and AT&T Broadband integration than other MSOs. But we’re at the 15 million home spot headed for another 15 million homes this calendar year. And we’ll do a million new connects in Comcast Digital Voice [this year]. That’s not a highly technical futuristic thing, but there are a lot of people focused on that. Going out a little bit further is the innovation in the video space. Here at Comcast we have focused on what we call the RNG [next generation] of set-top boxes. These are boxes that we have defined from scratch and that use OCAP [OpenCable Application Protocol] as the operating system so that features and services can be innovated on top of that open platform. These are set-top boxes that have advanced video codecs. As we look forward to hundreds of high-definition movies on demand and lots of high-definition channels, we can begin to take advantage of MPEG-4 or VC1 encoding in addition to MPEG-2, the standard today. The third thing is that we see the video space becoming more interactive and integrated with the voice and the data space. So we’re getting DSG, or DOCSIS Set-Top Gateway, signaling in place. Finally, the fourth key innovation in this space is downloadable security—or DCAS [Downloadable Conditional Access System]—so that boxes become portable. Getting a conditional access system that’s a little bit mobile is part of that innovation in the video space. There has been some consumer confusion over voice over IP, or VOIP. Is that why you’re calling it Digital Voice instead? Fellows: The Digital Voice product we’ve rolled out is based on the PacketCable specifications from CableLabs. These are a set of specifications and protocols that use Internet protocols and equipment designed for the Internet to carry voice signals. The reason we’ve not branded it VoIP is that it doesn’t touch the Internet. This is using equipment that’s really cheap because it’s sold in Internet quantities but on our private networks we’re using that to carry VoIP. Does VoIP still present technical challenges? Fellows: We need to treat the voice packets differently than run-of-the-mill Internet protocol packets. They need a higher QoS, or quality of service, so the packets go through first if there’s ever any congestion. The packets need to arrive in the right order, so a slightly different Internet protocol is used because if you lose a voice packet, there isn’t time to go back and say, "Oops. I lost that packet. Could you send it again?" TCP-IP is the standard Web protocol. That’s the one designed in case a nuclear bomb hits the network. It’s very reliable. But we want to use that inherent reliability but under UDP, so we don’t want to retransmit voice packets. The challenges have been adapting the Internet protocols to handle this carrier class of service. I’ll give you a trivial example, but it’s indicative of what we had to go through. With the Internet, you have an extremely reliable and inexpensive network. It is the standard procedure that if you load new software into a router you can hit the reset button. The router goes down and then comes up with the new code. Well, in the Internet world, when the router goes down, the packets get rerouted around that thing and if some of the packets happen to be halfway through the switch when you hit the reset button, TCP takes care of that and they get retransmitted and the Web page gets reassembled at the far end without any problem. But in the VoIP world, if you take a switch down and there are calls going through it, those calls go down. It’s not something you do. You load the new image and then in real time on the fly—without resetting—you point to the new software and boom, you’re off and running the new code. And you overwrite the backup image of the code. There are things like that we had to go through, and we’ve gone through those. The challenges in the voice world now are just scaling that—getting millions of subscribers up and running, doing a hundred, doing a thousand, doing ten thousand, doing a hundred thousand. We’ve all been through those scale milestones. We’re now heading for a million and then 10 million. The other thing is to say, "Alright, having done this with Comcast Digital Voice, where does that lead?" Because me-too telephone service is a good thing to have, and the good news is that customers recognize what it is. They pick up the phone and hear a dial tone and know what to do when they hear dial tone. We’re all trained to interact with the phone system. But now you’re riding on Internet protocol. What about video telephony? I’ve got a videophone in my office and in my house in the Boston area, and I make video phone calls back and forth. Skype and other Internet players are already offering video VoIP products. Can we assume cable will offer a higher-quality alternative? Fellows: If it stays on our private network and not on the Internet, then we can ensure what the bit rates are and that all the packets will get through. If the service went over the public Internet, then we would be in the same boat as Skype and other people. You take your chances on what’s known as a "best efforts" networks. So what other new telephony features are on deck? Fellows: At some point, we’ll touch on the wireless space and something called Fixed Mobile Convergence: The mobile telephone converging with fixed telephones when the mobile wanders inside of a house. So we’re extending calls to be video calls as well as voice calls, extending mobility to that. I can also mix up the data world and the voice world. When I get a phone call from someone, I record a voicemail. Well, how about if the voicemail box shows up as a buddy on my buddy list? When someone leaves me a voice mail and I’m logged in somewhere on the Internet, boom—a little icon buddy pops up on my messaging window, and it’s my voice-mail buddy. I can click and listen to the voice mail. And with a couple of Internet techniques called identity and presence, that voicemail can go find me on the Internet. That’s something that’s been talked about for a long time and called Unified Messaging. E-mails that get read to me by a computer through my phone. Voicemails that get stored as audio files and show up on a buddy list on a computer. And that’s where this is all headed over the next four or five years. Internet players are trying to offer those services as well, no? Fellows: That’s right. There have been e-mail services in which you can dial a phone number and enter some codes, and a computer will read that e-mail to you. But they are not really unified in any sense. The address book you’ve got in your phone is different than the one in your computer. The voicemail box you’ve got for your cell phone is different than the voicemail box that you’ve got for your landline. Those are the kinds of things that we’re working on, with the Sprint venture, to unify them. In January, Comcast joined several MSOs to announce an effort to build OCAP-enabled set-tops. How will that lead to more converged services? Fellows: OCAP is technically called a middleware because it rides on top of an operating system. But this is a common operating system that Comcast can run its applications and services on top of. It was developed with the thought that I’ve got a television set or some piece of consumer-electronics gear. And when it comes out of the store and gets plugged into a Comcast system, the Comcast service has to come up—our guide, our access to video on demand and any other applications we have on there. The only way that I can write a set to come up on [any box] is to know what that computer language is. And we’ve defined that to be OCAP. The other thing that we need in that retail world is security. As that box wanders about the countryside, the ability to decrypt the system needs to be portable. In the first place, we solve that by putting it on a card and the consumer electronics or TV set has got a slot in it. When you come to Comcast land, I plug in a Comcast card known as a cable card. And you go to Time Warner land and they plug in their card. You go to Cablevision Systems, and they plug in their card. This is what downloadable security supersedes or augments. You can continue to use cable cards, but instead of going to Comcast and getting a physical card and plugging it in and entering a bunch of numbers, you can just plug it into a Comcast system. Comcast recognizes you and downloads what’s appropriate for Comcast. If you unplug that box and wander to Time Warner land, the same thing happens. Their service gets downloaded. So is OCAP more about portability and retail availability than adding lots of new features? Fellows: Well, it starts off being the former, but a side benefit is that if I have some innovation, I have a common platform to write against. In the old, old days, you had Pioneer set-tops and Scientific-Atlanta set-tops and General Instrument set-tops and Magnavox set-tops, and a whole bunch of things. If you wrote some cool something, you’d have to write it six or seven or eight different [times]. Now that OCAP is a standard, you can write your innovation to OCAP, and as a vendor, you’ve got a market. You can sell it to everybody in North America. And because we chose OCAP to be very similar to a standard in Europe called MHP, you actually can innovate and have a worldwide market for your Texas Hold ‘Em poker game that you can play with your TV set and remote control, or whatever. So it does have this feature and competitive innovation. If I’m going to do a videophone that comes out of a set-top—our set-tops have USB interfaces on them—let’s plug in a little USB camera-microphone, and let me use my TV set as a videophone. In the proprietary world, I have to do that for almost each individual box that’s out there. But at least when we get this platform in place, you’ve got a platform for innovation. Could it get to the point where subscribers could even pick and choose features on their set-top box? Fellows: One of the ground rules still in place in the set-top world is that the cable operator still has control over the suite of applications that can run on that box—mainly because they are leased boxes and in the technical sense still the MSO’s box—but also because of the difficulty of managing that and guaranteeing some level of decent service. Yet someone may have a video-phone application, and their neighbor may not have a video-phone application. So amongst the set of services the customer will have some say on what they have. The other plus is that we have a box called the OCUR, which stands for the OpenCable Unidirectional Receiver. That cable card slides in so the box can decrypt video programs but then re-encrypt them using a digital rights management program—and then it plugs into a computer. So we’re finding ways of bridging our cable video world with the personal computer world. How important is that portability? Some of the DBS players already offer portable gadgets. Fellows: Occasionally I’m put on the defensive by some satellite vendors that have portable solutions for satellite boxes. My response to that is that the most successful portable devices out there are the family of iPods, and most of them are actually getting loaded using cable modems. So we are actually the infrastructure for loading up your iPod or video iPod. We’re part of that ecosystem, and we don’t particularly have to do something. It’s just a service that we offer. Are content owners getting to a point where they’re more comfortable with portability, on-demand services, etc.? Historically, they have worried about protecting content as it gets saved to hard drives or flows from device to device. Fellows: We use something called conditional access. In the world of computers and software-based devices, they use something called digital rights management. One of the main differences is that within conditional access, we have hooks into hardware. There are secret keys buried into chips buried into the devices so that a cable operator can always authenticate that device. It’s a little more secure than DRM, and Hollywood has always appreciated that difference. We use the OCUR device to connect the cable system to a computer. If we connected to a portable device, we would use the DRM that the device supported. Typically, that’s Windows Media DRM. So with the OCUR, we translate from conditional access to Windows DRM or other accepted DRMs. I know that content providers trust conditional access, and when they do trust their content to a DRM, we translate between the two. But are there any areas of security that could be improved to make the content owners more comfortable? Fellows: One thing I’d like to do is to extend conditional access into portable media devices. We could extend these hardware hooks so that even the best content could go portable. But other than that, it’s just getting everyone—including us—comfortable with all the DRM systems out there. And then in a secure fashion, we can translate from one world into another. What about network architectures? With IPTV systems rolling out, how will cable compete on features? Fellows: One of the problems is that IPTV means a lot of different things to a lot of different people. So first you have to ask, what do you want your IPTV system to do. If it’s to be interactive and get access to lots of different channels, then the IP network that we have in place today that delivers video is sufficient to see us past those competitive demands. Right now, when we receive signaling off of a satellite or fiber network, we use Internet protocols to ship the content around town. We use Internet protocols to drop it off at the right hub, the right head-end, the right neighborhood. And we use the packet protocol—not IP—to translate video into our set-top boxes. In terms of switched broadcast, Time Warner Cable and Cox have deployed that in the field. We are a little bit behind them, but we are attempting to do this in conjunction with our on-demand systems. Here in Philadelphia, we have 6,000 channels on IP and available to the customer; 5,800 of them are accessed through a video-on-demand menu and 200 of them are accessed through a grid guide. So you could think of Philadelphia as having 200 channels, but at any given time there are 6,000 things to choose to watch. So we try to separate the actual IP transport, which we use today, from the guide. But eventually the guide and the search functionality will all merge into a common interface. When will you finally go completely digital? Will it happen in the next five years? Fellows: I’m going to give you three answers to that because I’m famous for saying that all-digital is not no analog. That means that with digital simulcast, every channel in our system is available in a digital format. It’s just that about 80 of them are also available in analog formats. So already, all of this IP innovation is possible in the systems where we are simulcast. The fact that we’ve got analog there means we clog up or waste some bandwidth, but we do that to be friendly to analog TV sets. The second answer for all-digital is that we’ll probably take off everything about the same time the must-carrys go off the air—so in the 2009-ish time frame. But even there, we will leave a set of analog channels on—I’m just guessing for another five years or so. That’s so basic-only subscribers or third or fourth TVs in the home can tune a set of 20 or 30 channels. I think that’s in place for another decade. And I think for another five years, 60 or so [analog] channels are in place. What’s the benefit of eventually going all-digital? Will it enable more services or just free up bandwidth on the system? Fellows: Well, the answer is yes to both of those. Every time I take back one analog channel, I can either put three high-definition channels or 12 standard channels. With switched broadcast, I can at least double the number of channels I could put into those slots. Or it’s another 100 or so high-speed data subscribers or another 200 telephone calls. So if I take back 10 channels, I could launch 60 high-definition channels using switched high-definition broadcast. Or I could go from an 8 megabit per second bit rate to a 16 or 20 megabit per second high-speed data bit rate. So the answer is a little bit of all of those. Our high-speed data will continue to get faster—especially if high-definition channels continue to multiply. And also the number of things that are available truly on demand—roughly 5,000 hours of 6,000 things we have out there today—will double up into the 10,000-hour range. The possibilities seem endless. But does Comcast have a sense of which services consumers really want and how those tastes will evolve in the next five years? Fellows: The evolution of customers’ tastes and desires…I think we’ve got a pretty good handle on. We just issued a press release on a study we did on children and video-on-demand programming. In a sense, it’s things that we didn’t think about—like peer-to-peer protocols—that have the most profound change on our systems. But even there, we know about peer-to-peer and how that will evolve from the music world into the video world. Our voice people have done studies on portals. And our GuideWorks division does focus groups all the time. In fact, they’ve got a built-in focus room—with the one-way mirrors and cameras and everything—built into their facility out in Radner, Pa. And we run people through there every day studying how they interact with our cable system. Evolutionary—I think we have a very good handle on. Occasionally, something comes out of left field that no one thought about, [like] when they unleashed this IP technology, and that’s when my job gets interesting, and you’ve got to react and study this new thing. Do next-generation architectures help you innovate more quickly? Fellows: That’s the promise of an all-IP system. When you’re using all open protocols, you sort of harness an entire world’s worth of innovators because they speak your IP language. When they innovate, the innovations can be used on your system. For example, the new Google Earth map on your computer. That’s authored in a new language that’s a combination of Java and XML—and if your set-top boxes speak that language, then the application automatically runs. And again, this OCAP middleware that we’re trying to get in our set-top boxes happens to speak Java as a programming language. So the reason to go IP is to harness this innovation and then, with that, comes the speed to market. You aren’t trying to say, "Well, that’s a cool idea. Let me author that in Scientific-Atlanta language or Motorola language or yet another language." Everyone is speaking the same language. With such ambitious plans, will Comcast need to spend more or less proportionately to meet these challenges? Fellows: The spending levels will be about the same whether we’re spending it on boxes in the home or software in the network or getting bigger pipes. Those go through phases. But with any one thing that you look at, the spending is big at the start and goes down over time. Now Comcast got in trouble, frankly, by saying, "For the current services that we launch, the budget’s going to go from 4 and a half billion to 3 and a half billion to 2 and half billion to 1 and a half billion because the upgrades have been done and the fiber has been laid." But then we said, "Oh, by the way. We’re now going to launch telephony. Here’s some new spending, so the budget’s going to remain flat." Wall Street said, "Wait a minute. You said the budget’s going down." And we said, "No. For stuff that we’re already doing, the budget is going down. We’re doing some new things." And Wall Street said, "Ah, you lied to us." And we said, "OK, our budget’s going to stay flat." And they said, "Well how can you introduce new things if you’re budget stays flat?" Well, that’s because the old spending’s going down and that leaves room for new spending. So I think that our pipes are laid. Your front lawn has already been dug up, and our fiber is in the ground. As we go to all-digital—as we take back those analog channels—it’s a virtual tripling of the capacity. Two-thirds of our bandwidth is now tied up in analog. Over time, we will take that capacity back and launch new high-definition, launch new switched services, launch new on-demand services, launch higher speed or video telephony on top of telephony. But again we’ve got money and bandwidth in the bank, and I feel pretty good about our ability to compete and to innovate and to react to customer demands.