The second technique of video streaming uses a special software on a server to transmit files of highly compressed films and requires a compatible player on the client for their visualization. With these systems, films are viewed in real time and not saved as files for subsequent viewing. VDOnet and Xing Technologies are the leaders in this sector and Vosaic is a promising new arrival. These systems, as opposed to the autonomous reproducers, operate on the UDP transport band of the Internet protocol (User Datagram Protocol) and have several advantages over the TCP method for dispatching video data.
The main advantage of UDP is that it doesn't require confirmation upon arrival of each data packet to destination (as with the browsers). Packets lost during trasmission aren't sent again: the reproducer simply skips them, visualizing the new incoming data.
Another advantage of the UDP video flow is that you can look at the video in a separate window outside the browser. If your connection is fast and efficient and you have plenty of memory, it is possible to play the video in a window while simultaneously viewing other Web sites with the browser. This function might not be all that interesting for viewing cinema trailers, but in all likelihood news-bulletins will one day come in this way. The server components of these server-reproducing schemes can handle standard video formats such as AVI and MPEG, but they do not allow for saving files.
Today's programs for server streaming take into consideration that a good amount of the film will be lost during transfer. Furthermore, the servers are designed to automatically decrease the data flow to slower systems. Those using 14.4 Kbps or 28.8 Kbps modems will receive fewer video frames than those with an ISDN connection at 128 Kbps. The system allows a Web site to satisfy a wide variety of visitors with the same single video file. The only requirement is that the connection to the site is effected with the suitable replay software. Whereas the players are generally distributed as freeware to the public, the servers must be purchased.
The prospective target for such servers are "content providers" and the software for the servers is usually sold on the basis of the outflow (the number of visitors that can be simultaneously connected to the server) or the total bandwitdth that the server software is able to distribute simultaneously. Though the server-player streamers are capable of a video quality superior to that of the single player solutions, the overall video quality is still a far cry from TV transmission.
A tempest of bits
Naturally, the problem is the sheer size of the files necessary to memorize high resolution films as digital data. A 30 second film in color (24 bit) at 15 frames per second, rendered on a quarter of the screen (160x120 pixels) occupies approx. 25MB of disk space. With a 28.8 Kbps modem that has a through-put of approx. 2 Kbps, the downloading of that 30 second film would take more than three hours. The same film, full screen would take more than 12 hours. Compression is one approach to the problem.
Codecs (compression / decompression systems) are software or hardware tools based on mathematical formulas that reduce the quantity of data inside a video file, often reducing to a minimum the redundancies within or between shots. Many large portions of information in a video file are repeated innumerable times: it is sufficient to eliminate the repetitions to obtain a drastic reduction of the necessary data. A deep blue sky occupies the upper half of every scene? Codecs can practically halve the size of the file by not transmitting the whole sky for each shot. Rather than repeating all those blue pixels, the compressed file will contain a formula that tells the player where to place them. The instructions can be relayed in a few bytes instead of millions.
Furthermore, since the colors of a film undergo gradual modifications between shots, with the exception of changes of scene, video Codecs can compress the files by finding the sequential redundancies. In real practice, compression is rarely such an easy matter: even a single shot of a blue sky may present subtle variations in hue and the colors will change from shot to shot. While some images lend themselves to be notably reduced by compression, others are not so fortunate.
For example, a face against a background of intense color will probably be super-compressible, perhaps as much as 90% (from 57K to 5K in one of our tests). But the same face in a scene lensed in a crowded street might only be compressible by 10% (from 57K to 51 K). Even compressing a film by 90% will not guarantee a file small enough to be rapidly transmitted in real time through the Internet. The downloading of our 30 second video would still take 18 minutes with a 28.8 Kbps modem and 2.5 minutes with an ISDN connection.
Less is better
Video codecs regulate and optimize two variables: the speed of codification and that of reproduction. The speed of reproduction, measured in frames per second, controls the uniformity of movement. The more frames, the smoother the movement. The speed of codification determines the degree of compression used to create videos to be reproduced at certain bandwidths. For example, videos codified for a speed of 21 Kbps are replayed smoothly when downloaded on a 28.8 Kbps modem, but this degree of compression often causes grainy, unfocused and imprecisely colored images. Naturally, on an intranet it is possible to achieve high quality video in windows that occupy a quarter of the screen. < P> But the Web sites that support visitors connected via telephone line must supply images that have been coded for considerably lower bit speeds, codes that produce films of far less quality. One solution, developed by ITU (the International Telecommunication Union) is the H263 Standard for low bit through-put. The H263 Standard allows for video compression independent of the rate of data transmission. With this approach, the telephone connections are able to achieve a smoother replay, close to the original speed of the frames, but at the cost of overall image definition.
Until H263 is completely defined and released, Webmasters who wish to offer films can choose from a wide selection of solutions: the simplest is to use AVI, MOV or MPEG files that can be requested through simple connections, leaving to the visitors the trouble of obtaining the necessary reproducer. By providing links to the sites where visitors can download the streaming players examined in this issue (CineWeb, ClearFusion or InterW MPEG Player, for example), Webmasters can help their patrons appreciate their video offers without problems. VivoActive has a high-performance solution that doesn't require a content provider to install another server. To offer instant visualization with software that can adjust bandwidth to support each patron at the highest level, administrators can use the VDOnet, Vosaic or Xing systems.
In the course of our tests, we have discovered that the installation of video servers with these products isn't particularly difficult for administrators who have complete control over their systems. The servers can coexist with HTTP servers like the Microsoft Internet Information Server and O'Reilly's WebSite. Webmasters who rely instead on hosting services, must convince their service providers to add this new software or install dedicated systems with their own Internet address. The patrons must simply obtain a copy of all the reproducers available on the market: with the exception of the two AVI players and of CineWeb and ClearFusion, all the others can coexist on the same system. Furthermore all the reproducers can be downloaded cost free. Only a few require registration after a period of evaluation, and some even offer free licenses. None of the reproducers can accelerate the downloading of files, but they allow the visitor a preview of the videos as they are coming in.
Last updated 7-Jan-1997
Webmaster Elsy mail to: email@example.com