Steve Deering of Xerox PARC (Palo Alto Research Center) might be considered the father of the Mbone, the high speed multicast portion of the internet. As a doctoral student at stanford University, he began working on multicasting for his thesis, and carried that work to PARC in 1990, where he has continued to be the leading expert on multicasting. He has also spoken at DeskCon on the topic of desktop Video conferencing (desktop VC).

View: Tell us about video on the internet today-whats feasible and what's not?

Deering: Video on the Internet has been happening on a small scale for a couple of years. What's feasible today in videoconferencing over corporate intranets,where there is usually an abundance of bandwidth,and on the public Internet where the paths between participants are uncongested and have available bandwidth.Where there is competition for the bandwidth is scarse,it doesn't work very well at all.

View: Lets get a scense of feasible scale .In a large company ,with 100 concurrent intranet conversations using VC,will transmission bog down?

Deering: That depends on the topology in use ,the capacity of the links,and the patterns of usage.100 video calls is alot to ask for a single ethernet.

View: What has held back wide spread adoption of desktop VC on the internet,to this point.

Deering: Internet telephony is already widespread.but video still requires a camera and a video capture board.Only recently have commodity workstations come equipped with the horse power,microphones,and speakers to enable the video conferencing.Previosly you had to buy alot of extra hardware.Nowthat it's bundled in ,and is fairly inexpensive on most of e platforms,we are seeing an explosion of Internet conferencing activity.

View: 12 months hence,do you expect the commodity PC will have a camera built into the monitor ,and video capture on the mother board?

Deering: Right.That is enabling technology .Wide-scale Internet video conferencing wouldn't happen if people had to go out and buy codecs and video gear.Equiptment has to be accessible.The other part is networking piece,where the main issue is capacity for video,which requires more bandwidth than you get on a 14.4kbps modem connection to the internet.We can now do marginal audio/video teleconferencing with a28.8 kbps modem. The people working on compression continually find clever ways to get more bandwidth.If your talking about getting more capacity for residential use,it's through ISDN or cable modems.For business ,It's ISDN or T-1.Other Technology enablers provide multicasting ability over the internet.

View: Steve,can you explain multicasting?

Deering: Multicasting is the technology that allows large video teleconferences.If you want a teleconference with a 100 people,and you don't have multicast capability, then you need a 100 times of the capacity,at least first-hop network.When video is sent to 100 receivers ,if there is not support for multicasting,then a separate copy of the video must be sent to each person.100 copies of the video go out over the first hop,and subsequent hops.So large-scale video teleconferencing isn't very feasible until multicast is more widely available.

View: Why use the internet for video conferencing over telephone lines?

Deering: First,Because the Internet id there,and so people want to try it .And of course,because if the different pricing models.Internet access is usually offered on a flat monthly basis,regardless of how much you use it.This creates a perception that there is no additional costs for long distance audio/videoteleconferences.You do pay for internet usage-surprise ! but just on a different basis than with telephony.There is a great debate about the economic model of the internet versus traditional telephony.Whether one prevail ,or we'll end up a hybrid of the two,is unclear.We live in exiting times,relative to the economics of the internet.

View: What is the importance of the Mbone,and why is Mick Jagger worried about it melting down?

Deering: The Mbone is a part of the internet that supports multicasting.Mick Jagger was transmitting a rock concert, and you can imagine the potential audience size would be in the millions.You cannot send that much video without multicast capability.The Mbone is just our name for that of the internet which has been upgraded to support multicast.Over time,we expect that name to go away,as the entire internet support multicast.Then there will not be a separate thing called Mbone.

View: What is needed on the internet to support multicast on a large scale.

Deering: First,support multicast routing.Second,more sophisticated traffic/congestion management and control than what is currently on the internet.Most of the routers on the internet service packets on a first come first serve basis.Congestion occurs when the number of incoming packets exceed the capacity of the next hop they want to go to. When a packet arrives,it is put on a queue for an outgoing link.If that queue is full.that packet is thrown away.most internet applications use TCP,which responds to packet loss by slowing down.That has worked to the benefit of videoconferencing over the internet,because these apps do not use TCP,they are using UDP,and are totally insensitive to congestion.They transmit at a rate based on the amount of video activity at the source,or sampling rate on the microphone. When the audio/video causes congestion at a router ,the traffic using TCP will actually back off,making room for the audio/video.IT favors us ,but it displeases the people who try to do every thing else like transfer files. Right now it is possible for a video source to monopolize a link,because there is this dumb queuing at the routers.If the routers extant had someway to prevent any single source from monopolizing an output link,then there would be more fair usage of the capacity.This is a matter of presenting audio and video from driving out all the other traffic. Because of that,we've had to build rate limiters into our applications,to run them more slowly than they would otherwise could, to make sure we always have capacity for other applications.That's a poor solution,because it means when there is no competiting traffic,cannot take advantage of the unused capacity.

View: How many companies actually have Mbone applications?

Deering: Most Mbone applications are from universities and research labs.One company I know of ,Precept,has commercial products for the Mbone .There may be others I don't know of.Most companies that provide commercial teleconferencing for the internet don't use multicasting,because multicast isn't available every where yet.So they did not want to constrain their product.The CU-See me program and others like it were designed to do their own packet replication,to enable multipaticipant conferencing. There is a chicken and the egg problem here.Getting the multicast capability deployed,so that the commercial organizations will modify their applications and use it.They would all like to be able to tale advantage ,but first the ISP's have to be persuaded to complete deployment of the facility.

View: Is it a question of ISP's being paid to provide multicast capability.

Deering: Not really.It isn't their highest priority.Plus,it is new technology and they are not so familiar with it.

View: How will companies make money from VC on the internet.

Deering: Mostly it will be a software business; that is, shrinked-wrapped conference software.Thats the basis of companies like Precept.It's possible that there will be a market for compression hardware.Much of the success of conferencing over the internet has occurred because it doesn't require special hardware,but there could emerge a market for compression boards,cameras,add on hardware.Other than that ,its consulting,and helping people figure out how to play with this stuff.

View: How rapidly will desktop VC on the internet be odopted?

Deering: We've tried teleconferences on the internet,where only some of the participants have video,and that works.If you are meeting with people you already know,it's far more important that you have good audio,and a shared white board or other collaborative tool.Video is a nice complement,but if the people already know eachother,there's less need for video.I expect many people to conference over the internet without bothering with the video part.

View: What is RSVP?

Deering: This is a new protocol to extend the internet to support resource reservations.Let's say I would like to reserve 500 kilobits for 1 hour.With RSVP,the network would come back and say yes or no,sorry we cannot guarantee you can have that capacity to Yourself.RSVP is the reservation protocol for the internet,so computers have a way to request the capacity they need.RSVP was also developed at Xerox PARC.

View: How important is RSVP?

Deering: It's unclear.Some people think it is essential happen to think the key factor is having enough capacity to support the demand.RSVP helps where there isn't enough capacity.It lets you give guarantees to some fraction of the requesters and deny bandwidth to everyone else In the long term,that is not an adequate solution.

View: What will decide whether RSVP is widely adopted?

Deering: In the public internet ,it requires the invention of a billing scheme.In a private net,the criteria for deciding who gets capacity can be set by a companies internal policies.But on the public heterogeneous net ,if I want to conference with people in Australia and Japan,that could mean going through a dozen of different Internet service providers.Why should,and how will they all agree to provide capacity to me?I cannot imagine,unless there is a billing system that provides for inter provider settlements.I think that is a long way off.