Digital Journal

The State of Streaming Media

by: Christopher Harz

We keep hearing about the great new killer app of the Internet: streaming multimedia. Headlines and newscasters blare that soon anyone with a digital video camera and a PC workstation at home will be able to show their work by streaming it over the Internet, leading to thousands of low-cost independent films becoming available over zillions of Web sites. There is just one small problem. It can’t be done.

It is, of course, possible to stream video and audio over the Internet. Streaming is the delivery of content from a Web server to a user’s browser over an Internet Protocol (IP) network. It allows you to see a video clip before it has finished downloading. In essence, it downloads a small file and then lets the user view the video and audio "on the fly" as they come in. Older formats such as AVI or QuickTime 1 required that the whole file be downloaded before viewing could begin. Streaming works its magic by compressing the audio/video at the source and sending it to a "player" at the user end, essentially a decoder that decompresses the material for play. A player is a plug-in to a browser; common streaming players such as RealNetworks or Windows Media are already included in the newest versions of browsers such as Internet Explorer or Netscape.

The use of streaming media is catching on like wildfire. Short clips shown in windows a couple of inches wide on the user’s screen are commonly used to illustrate Web news broadcasts or databases, or for rich media ads. Audio streaming via MP3 and other formats is, in fact, very good, and the audio portion of live music webcasts can approach broadcast quality. But the promise of "TV-like video" over the regular Internet is a blatant lie. It can’t be done for two reasons: bandwidth and latency.

The bandwidth constraint is easy to demonstrate. Full-screen video at 30 fps, even with good compression, is still over a megabyte in size, which translates to a bandwidth requirement of over 8 megabits per second (transmission is always expressed in bits per second, whereas memory is described as bytes). Trying to cram this through a modem capable of only 56 or 128 thousand bits per second is like trying to push a watermelon through a garden hose. New broadband pipelines like cable modems or DSL will alleviate this problem with speeds of around 1.5 megabits per second. But there is a second problem that generally gets ignored by the wild optimists, namely the latency found on the Internet due to the high number of connections, traffic loads, and outages. Latency delays are irrelevant for text messages, which is what the Internet was originally designed for. But for video they are fatal – users want to see pictures in the proper sequence and timing, but instead they get video with varying speeds and "dropouts" lasting from seconds to minutes. The reasons for the delay lies in the high number of connections -- a signal traveling from Los Angeles to Moscow could go through dozens of connections, each of which add delay and error. A server going out anywhere along the way will mean a blank screen at the user end, as could an excess of traffic along any of these links. Because of these problems, industry insiders now admit that it is impossible to stream high-quality video over the regular Internet.

So what can be done? The answer is being provided by a number of "video enabling" companies that were on view at the recent Telecon West Show in Anaheim, California. Because the demand for high-quality streaming video is so strong, such companies are becoming the darlings of venture capitalists, who consider enabling video transmission one of the "hot" areas of Internet technology. Companies such as FVC.com, Cisco Systems, SightPath, and Sandpiper try to solve the aforementioned problems in several ways.

The first stage is at the content provider end, where the enabler may help film and format the video correctly, so it can be compressed easily. A program with quick "jump cuts" of the type seen on MTV is very hard to compress, since codecs (compression/decompression algorithms) usually leave out some video frames and then depend on interpolation between the remaining frames to recreate them at the user end; very fast lateral movements or quick edits make this hard to do.

The second and most important thing an enabler does is route the content directly to the gateway access of a "backbone," a high-speed Internet link that is like a toll road, i.e., it’s very fast, but it’s not for everybody, and it costs extra to use. Some enablers use land lines; VStream, for instance, links directly to a UUNet backbone. Others, such as Sandpiper, bypass the crowded "public" Internet roads by use of satellite links.

One of the most impressive demos at Telecon West showed a video being shot at a facility provided by Play, Inc., which provides a compact traveling studio complete with virtual sets allowing computer-created backgrounds, hooked up to a three-foot dish that transmitted to a satellite; any user with a dish and access to the satellite could get the webcast live at broadband speeds. If the receiver is a cable company, for instance, it could then retransmit the signal to its subscribers via cable modems.

A third service an enabler may offer is mirroring, transmitting the material to a number of remote servers where it is stored. When the user demands the video, he receives it from a gateway that is electronically near him, thereby reducing transmission delays; since the content is cached at multiple sites, he can fall back to a second server in case the primary one goes out.

A fourth service an enabler can offer is overall network management – watching over a live webcast to make sure it is proceeding as planned, re-routing if there are any problems, and monitoring the viewers to assure they are getting a quality product.

The bottom line, then, is that high-quality video is possible over Internet links – but not with normal home access, or over the public Internet.

There is one major exception to all these barriers to transmitting high-quality content over the Internet, and that is for animation. It is totally possible to transmit full-screen high-quality animated content over the Internet today. This is done by using special protocols that in essence break the content into two parts, the animated models (either 2-D or 3-D) and backgrounds on the one hand, and the movement data or vectors on the other hand. The models and background are first downloaded to the user in a small file that may take a minute or so to download. After that the only thing he gets over the Internet is the movement and audio data, which are small enough to be handled by 56K or 128K modems. The user’s desktop does all the work, rendering each scene with the models it has already stored, and moving each character according to the vector data coming in. If there is a glitch on the Internet and the description of how a character is moving is delayed, the software can keep the character moving by "dead-reckoning algorithms," making intelligent guesses as to where it would probably be going. When reconnection is established, any necessary corrections are made as subtly as possible. This technology, first developed by the military for large-scale war games, is now being used for online multiplayer computer game networks and other realtime animated web content. For an example of high-quality (and creative) 2D animated content using the Flash 4 format, see Spumco’s Web site. For an example of great 3-D animated content using the BDE format, see the Brilliant Digital Entertainment site.

Even though high-quality video cannot be viewed via today’s public Internet, the access to specialized, private parts of the net is increasing rapidly. As a result, many companies looking to the future are buying the rights to "short format" films (generally 30 minutes or less), such as Atom Films, SightSound, iFilm, and MediaTrip. Such companies have the resources to enter into relationships with the owners of backbones to transmit their material.

What does all this mean to the budding independent producer that wants to start showing his movies over the Net? First, he should not expect anyone to watch films that he sends over a 56K modem; he needs to get in touch with one of the enabling companies that can help him onto the Internet’s high-speed toll roads. Second, he may want to seriously consider producing his content as animation. The current advantage that animation enjoys over straight video in being easily transmittable over the public Internet will last at least 5 to 10 years. And on the Internet, that is a very, very long time.

return to the articles section...