Network Computingwww.networkcomputing.com

With the growing availability of LANs, WANs and the Internet, a great bypass temptation has arisen. View these networks as "free transport," spend $200 for codec software and a camera, and as fast as you can say, "Hey! I've got a cheap videoconferencing syst em," you will. But the experience will be about as satisfying as the failed bypass experiments with the Internet telephone.

The problem lies in the inherent nature of IP networks. The steady digital bitstream presented by the videoconferencing system's hardware or software codec must be chopped up into packets. Most networks guarantee delivery and recombination of these packets in correct order, but not at a steady rate. LAN internetworks may be subject to high latencies. Audio quality is susceptible to slight interruptions or rate variations in the data stream. Video presentations will freeze if interframe information is held up more than 200 to 300 milliseconds (ms). Developers are working hard to improve codec software and invent techniques for overcoming network limitations.

Cornell University's CU-SeeMe has been a highly influential product. Codec software to support videoconferencing over a network was first developed as a research project. Cornell's public domain version still is available free as a download (ftp://gated.cornell.edu/pub/cu-seeme).

In May 1995, White Pine Software licensed the Cornell system and turned it into a well-supported consumer software product, Enhanced CU-SeeMe. You can download a 30-day evaluation version (www.cu-seeme.com) and then buy it from White Pine for $69.

I obtained Enhanced CU-SeeMe Windows 2.1 and was particularly curious about its ability to use the Intel ProShare 200 proprietary camera/microphone/video board installed on my i486/66 PC. To make effective use of the BRI ISDN line installed at home for videoconferencing, I also have a U.S. Robotics Sportster 128-ISDN modem that uses Trumpet WinSock to connect to my Internet service p rovider (ISP).

To test Enhanced CU-SeeMe as a home visit tool, I arranged a session with a colleague who has a Macintosh connected to his ISP using a 33.8-Kbps POTS modem over a second phone line. This left his primary phone line available for voice contact. We found that when either party is sending audio through a microphone using Enhanced CU-SeeMe, his video nearly freezes. White Pine mitigates this by keeping the audio channels off and providing a "push-to-talk" button. Audio transmissions have a two- to three-second delay. Quality is acceptable, but a CU-SeeMe session is reminiscent of the old ham radio days--every sentence ends with, "over!" We got around this by placing a direct phone call. Without the audio load, grayscale pictures were being transmitted at 3 fps to 8 fps.

We also tested the WhitePineBoard collaboration feature. It worked as well as similar whiteboard software supported on other systems. The most impressive result was that we were conferencing and collaborating between two di ssimilar platforms: a Windows PC and a Mac. This shows the seminal roles that Cornell and White Pine have played in the evolution of DVC. Using a heterogeneous mix of nonstandard but existing devices, protocols and transmission media, many of the predicted functions of H.323 systems are being demonstrated in Enhanced CU-SeeMe today.

Settling Standards Just as maturation and adoption of the H.320 standards suite settled the group system field, so will the rollout this year of new standards--H.323 Internet videoconferencing and H.324 POTS--begin to weed out the DVC market.

H.320 is a group of related standards that relies on clean, guaranteed bandwidth--primarily ISDN--to support acceptable videoconferencing performance and appearance. Under H.320, H.261 defines how the video portion is coded and decoded. The video codec is a sophisticated image processor that determines what is changing in the picture and what is not. A full information frame needs to be sent only two or three times each seco nd; intermediate frames describe only those pixels that have changed. The decoder section regenerates the video display frames using an updating process. A brilliant design concept in H.320, carried over into the H.323 and H.324 suites, is that decoder operations are relatively simple and rigidly defined. Most of the energy of audio and video analysis is devoted to encoding. An encoder must produce a bitstream that follows standard rules, but its compression, prediction strategy and processing power are left to the manufacturer to implement.