Swimming against the stream

By David Walker (Google profile)

For more than a decade now, technologists and investors alike have dreamt of delivering "video on demand", of letting consumers see whatever they want at the click of a button. Today, the dream lives as "Internet streaming video", delivered to PC client software like RealPlayer and Windows Media Player. Much of the corporate world assumes that this technology will soon deliver streaming video that looks just like the Saturday night movie. Typical is this October 2000 forecast by U.S. Bancorp Piper Jaffray analyst Gene Munster:

"We believe that streaming media is the next macro-growth driver on the Internet ... We believe the Internet of tomorrow (two to five years from now) will resemble the television of today in terms of audio and video quality, while enabling users to control the media viewing experience."

Or try this January 2000 press release. "Despite having lower quality than broadcast video, streaming media threatens the status quo of the broadcasting industry, according to report by Cahners In-Stat Group. 'For at least the next five years, there will be a battle royale between existing, traditional media companies and streaming media on the Internet,' says Gerry Kaufhold, analyst with Cahners In-Stat Group."

Wrong. Web video won't happen this way, not by 2005. In image quality, reliability and economics, streaming video lags far, far behind television. For the next few years, people betting on TV-style streaming video risk great disappointment and monetary loss.

First, consider streaming video's image quality. To approximate that quality, sit your portable television a couple of metres inside your front door. Then walk outside, close the screen door and walk back about fifteen metres. Start blinking rapidly. Now, how does the picture look?

Streaming video looks this way because its images must be compressed even for "broadband" Internet connections. Today's jerky streaming video, with its credit-card-sized screen, uses around around 40 kilobytes per second. An uncompressed, digitised picture on your standard TV uses contains more than 30 megabytes of data per second - almost a thousand times more.

This disparity between TV and Internet won't soon disappear. Compression will allow streaming video to approach TV quality without using up 30 megabytes per second of bandwidth. But as far as we know today, TV-quality streaming will require substantially more bandwidth than is available to most broadband Internet customers today. The bulk of the modems out there in audience-land right now, sucking in no more than 6 kilobytes of data per second, will never permit decent streaming video. And those slow modems are here to stay. Jupiter Media Metrix estimates that by 2005, just 36 per cent of US online households will have broadband Internet. The Australian figure will be significantly lower, especially given Telstra's patchy broadband Internet service standards.

Streaming video's reliability problems can make its image-quality problems look almost trivial. We take TV's continuous picture for granted. But most video streams pause or die automatically after a few minutes. No surprises there; the Internet was designed to get the message through, not to get it through immediately, and its network of cables and routers allows capacity bottlenecks all over the place. Internet providers lack any incentive to spend the sorts of money needed to give us TV-grade video reliability. Despite technical advances in content delivery, that situation will remain for some years to come.

It's likely that some of streaming's enthusiasts simply don't appreciate how hard consumers have to wrestle with streaming video. In a January 2001 report called Quantifying the Stream, Jupiter Media Metrix spelt it out:

"Most consumers have had bad experiences with online streaming video. Figuring out just how bad their experiences are is a longstanding problem ... Streaming media technologies have evolved to surmount the inadequacies of TCP/IP-based networks in delivering time- and latency-sensitive content. Codecs such as RealVideo and Windows Media compress complex video data feeds so that limited-bandwidth networks can handle them. Reverse-proxy caching, content mirroring, and streaming delivery networks all reduce dependence on uncertain Internet backbones. While these technologies are intended to slay the foes of jitter, latency, and packet loss, it remains difficult to quantify how well they accomplish that objective."

Beyond its technical limitations, streaming must confront grim economic realities. Most Internet facilities cost much more to set up than they do to run - that is, they offer "low variable costs". Streaming video, though, sports impressively large variable costs. Delivering a megabyte of streaming video to one person will set you back perhaps four Australian cents (the actual figure could be higher or lower). Streaming industry-standard slightly-jerky broadband video at 40 kilobytes per second into a credit-card-sized window, you'll eat perhaps two-and-a-half megabytes per minute. In other words, you'll pay around a dollar for every ten minutes of small, jerky video - or around $8,000 per user per 24 hours. (Your production costs are extra, of course). Good luck finding an advertiser willing to meet the cost of delivering that data. On these numbers, Kerry Packer and the other network owners can all sleep comfortably in their beds.

Uncomfortable facts like these explain why Forrester Research principal analyst Josh Bernoff was quoted as telling an iHollywood Forum in August that "streaming media does not have a future." They also explain why Jupiter Media Metrix predicts that by 2005 only nine percent of US Web advertising will be streaming media.

So why the streaming hype? Many people expect streaming video to duplicate television for the same reason that their predecessors expected the telephone to duplicate the gramophone and send symphonies down the wire; people always expect new media to duplicate old, but they never do.

Others flock to video streaming because they expect the Internet to produce an endless flow of technical wonders, because they believe the Internet is still a baby medium. In fact, the Net medium is in early adulthood; it already is what it will be.

Still others rave about streaming video because it's in their interests to keep video streamed. Downloadable video, passed from computer to computer as a file, may make more sense. It may even generate more excitement among actual users. But large entertainment businesses love streaming because it leaves them in control. Users can't save streamed video without permission from the streamer.

In the immediate future, the real question is how different Web video will be from TV. (Unstreamed, download-and-wait Internet video looks promising, in part because hard disk storage space is expanding so fast.)

One day streaming video will look just like the Sunday night movie. Just don't bet on it happening this year, or next.