Just as Google is finally fessing up that video can kill the Internet, AT&T is learning a similar lesson (WSJ subs only):
AT&T’s big bet on using Internet technology to vault ahead of rival cable operators in the television-distribution business is beginning to look more like a long shot.
The telecom giant says it has rolled out its so-called U-verse service in 11 cities. But that’s four fewer than promised, and the technology seems to remain mostly in the trial phase. AT&T executives acknowledge they aren’t fully marketing U-verse because the service can’t yet handle a surge of customers. AT&T counted just 3,000 customers at the end of the fourth quarter, unchanged from three months earlier.
Meanwhile, AT&T executives last month admitted for the first time that there were problems with the software for U-verse provided by Microsoft, its primary vendor on the project. That’s a concern not just for AT&T, but for telecom companies world-wide that bought Microsoft technology to run TV services using Internet protocol, or IP, to transmit signals.
It isn’t clear how serious the problems are because AT&T and Microsoft executives won’t discuss them. An AT&T spokesman attempted to play down the situation, calling it “a little fine tuning.” A Microsoft spokesman said the technology was “on track.”
But the delays plaguing U-verse have fed criticism that AT&T and Microsoft overreached, trying to get more out of Internet technology than it’s capable of delivering at this time. The skeptics include vendors, former employees and competitors. Surprisingly, one of the challenges they believe has tripped up AT&T is something the earliest TV sets could do easily: switch channels instantaneously.
Partnering with MS was a big mistake, and the technical approach is a big leap. Unlike Verizon, which runs fiber to the home with most of the channels in simultaneously moving in multicast streams, the U-verse system is video-on-demand over copper wire with a handful of channels to each home (like 4). So they need massive numbers of servers and with channel-changing at the central office. I wrote a patent application once for a rapid channel-changing system on a network like this, and I can tell you it’s a hard problem (though not an insoluble one, heh.) The basic problem is that you need to buffer up some data before you start displaying in order to have jitter protection, and the time it takes to fill that buffer causes delay in channel changing. MS demonstrated a fast channel-changing system at CES a year ago, but making something like that work on a real network is a very different problem from making it work in a demo.
The Internet is great for personalized programming, and not so great for huge amounts of bulk data. AT&T better get a better partner immediately and some better wires down the road if they’re ever going to get this thing to sing.