Poor Man's Video on Demand may be just what we want

After a year of using a personal-video-recorder, the Tivo, I've been surprised at just how profound an effect it can have on TV viewing patterns. I expected a lot, and I'm usually pretty good at predicting these things, but got even more than expected.

One of the things you learn quickly the the Tivo and its competitors, is you almost never watch live TV any more. They advertise their ability to pause live TV, but that's just because it's easy to explain, but in fact it's one of the most boring features.

Instead, you decouple when shows are on from when you watch them, and with a few exceptions, you don't care. You don't care when shows are on, and in fact you often are glad you watch them later than everybody else.

I've come to view this as a sort of "poor man's video on demand." The early dream of Video on Demand was a big server at the cable/video company with all the movies and shows you could watch stored on hard disk. You could request any video, and watch it right away. No need to rent discs or worry about when to watch. Done well, you could pause shows in the middle, stop them in the den and finish them in the bedroom, etc. All at a low price.

With the Tivo, you browse the listings of what will be on in the next few weeks, in addition to recording your regular favourite shows. Once a week I look at the movies that will be on. I point and click and select a show or movie. And then I forget about it. The menu of available shows in a given week is actually quite voluminous if you have cable or satellite. Turns out there is more than enough to keep you entertained, in spite of the fact that most of TV can be correctly dubbed a vast wasteland.

Sometime in the future, anywhere from a day to a week, I happen to scan the menu of available shows recorded on my Tivo and the movie I wanted is there. Any time from then on, I watch it at my leisure, perhaps over several days, just like a DVD I might own, and then I delete it.

What I've learned is that I don't actually care about this gap from selection to viewing. Yes, I can't watch the show that caught my attention right there, right now. There is always something else in the Tivo that I can watch. The times I need to get a particular show now, now, now are vanishingly rare, and it's definitely not worth making the system expensive to make this happen.

I'm not the only one to discover this. All Tivo owners know it, as do all customers of DVD rental firm NetFlix. With NetFlix, you pick DVDs to watch from a vast menu. They show up in the mail a few days later. You get a week to watch them. That you wait a few days is something that, remarkably, you don't care about.

I must warn you that's a vicarious report of NetFlix's affect. Since I got the Tivo, I have not been to the video rental store in over a year. Why should I? My Tivo always had several 4-star movies sitting in it that I want to watch. Or it has the very best episodes of some TV series, leaving out the dreck. This effect may not last forever but it has been remarkable since I used to visit Blockbuster every week.

It turns out we're happy to not get instant gratification, not if something else is at hand which we will enjoy just as much. And it turns out that the cost of delivering delayed gratification -- what I might call "Video on Request" is vastly less than that of the VoD instant-play dream. I begin to wonder if the VoD dream will ever come to be. Designers aimed too much for the "demand" part when what is important is not the instant delivery, but the viewer choice and flexibility.

Cheap delayed VoD

The key to delayed video on demand is that it can rely on a PVR with hard disk at the customer location. This means you can have very low quality bandwidth from the video source to the home. Low quality as in not fast enough to deliver video, and low quality as in having too much congestion and latency to deliver live video.

Low quality bandwidth is really cheap, because it's the bandwidth that's left over after people do the necessary overprovisioning to handle the big-paying live applications. When there's a live user interacting over the net, you want reliable, low-latency connections with lots of bandwidth. So you buy plenty, and sometimes you use it all but most of the time there's bandwidth to spare. It will just go unused if you don't find an application for it.

Non-interactive applications, like E-mail and background file transfer, can use the spare bandwidth cheap. That's because they can buy it on an "I don't care when this packet gets there" basis. The other applications go to the head of the line, and pay for it. The background applications live in the spare unused spaces.

If we don't care that we can't watch our video right away, we can set up a transfer of it to our PVR. This is just a plain old file transfer. The bytes can come in any order. Perhaps the last minute of the movie arrives first because my neighbour was also getting the film but started earlier than I did. Perhaps one block is lost by unreliable transmission, and is sent again. We don't care. Perhaps my private bandwidth is only 1 megabit on average where video requires 2. Again, not a problem.

Some time later, perhaps just a half hour, perhaps a whole day, and the video is there for me to see, on my terms.

To make this cheap, it is best to mount disk farms at points nearer the users. These disk farms are also cheap, because their only purpose is to cache. That means, if somebody using the same ISP or cable company wanted a given video, a copy stays around for other people at the same location, or nearby. The most popular shows are almost always ready to be delivered from this cache, but if not, they can be fetched from more remote locations.

It's also possible to not have this disk-farm at all. If you have two-way bandwidth (which some DSLs have but not most cable ISPs) the transfers can take place -- with authorization -- from customer to customer. That means no video is every sent from the master sources as long as somebody near the target has a copy. A very small cache might hold the most popular items -- new releases, current TV shows etc.

The disk farm is cheap. If a disk fails, who cares? It just means you do an extra re-transfer. It's slow, because you don't need to access the bytes in a specific order. You grab the ones that the disk head happens to be passing over.

Indeed, with such a system, one might wonder who needs broadcast TV? Well, there are things we want to watch live -- like news, sports, and those few special "water-cooler shows" that people need to see the moment they are available. So we need a moderate amount of the bandwidth for such material, and we're doing fine with the existing broadcasting architectures in cable, satellite and over-the-air. But we only need a very small fraction in this live state. The rest is prime for personalized delivery. In fact, the popular shows, if not popular enough to be sent over broadcast bandwidth for efficiency could still be pre-delivered to PVRs in an embargoed state -- encrypted, with the final decryptor key being delivered at the official "live" time. Then everybody could watch Survivor at the same time if that's what the studio wanted.

This system could allow any viewer complete access to the complete catalog of all TV and movies every made -- with a slight delay. And it could do it with the facilities we have today -- a $300 box in the home, and the megabit download channels we can already deliver to most homes.

Will it happen? That's another story. The broadcast lobby would fight any change like this that takes them out of the picture. After all, who needs TV networks if studios can deliver shows to viewers directly and cheaply? Who needs cable companies except to be an ISP and provider of the small number of live news channels? So those companies will of course fight this.

The future will probably provide something in between. And perhaps perhaps the bandwidth to provide live HDTV personalized to the home -- so you can pick a show, watch it now and pause it -- will become cheap sooner than I imagine. Though many VoD projects have failed hoping for this to become economical.

The home PVR solution will likely be better for the consumer just as the PC was chosen over the mainframe by 1980s corporations. There are many arguments why shared resources (be it mainframe timesharing, or video distribution from a cable head-end) are much more efficient than giving each person their own equipment. But the latter course tends to foster innovation and choice. In this case, giving everybody their own PVR and using cheap bandwidth turns out to be even cheaper than the centralized approach.