JB: Welcome to the CREN Virtual Seminar Expert series for Spring of 1998 on Campus Communications Strategies. Whether you are joining us by phone or on the Internet, you are here because it's time. This is Judith Boettcher of CREN, one of your hosts for today's session. And today's co-host is Greg Marks from MERIT. Good afternoon, Greg. As our topic today is The Future of the Campus Network, do you have any particular challenges with your network access from home or on the road?

GM: Yeah, I usually can't get anything like the bandwidth that I would like to get, and the vendors don't seem to be moving very rapidly to provide that. It's very frustrating.

JB: That's right. Our choices, in fact, are not expanding as quickly as they might, are they?

GM: No, there's a lot of publicity but not that much action.

JB:. Well, our guest expert today is David Wasley from the Office of the President at the University of California system. He is, and I must say I really do like this title, he is the Infrastructure Planner, and he focuses on strategic planning for information technology, and particularly focusing on communications technology. David has been active nationally in developing technology strategy for the Internet2 Initiative, and within California, he's project manager for CALREN 2, their California statewide network that will become part of Internet2. Welcome, David, and thanks for being here today.

DW: Thanks, Judith. It's my pleasure to join you again for this seminar. I think it's excellent that we are making use of the technology that we would like to see more broadly used for distance learning and teaching and as part of our daily lives.

JB: David, I know we're going to get into more detail about Internet2 later in our session, but maybe you'd like to just say a few words about how the work on CALREN is progressing. Do you think you'll be ready to connect to Internet2 when it's ready?

DW: Oh, absolutely. In fact, CALREN 2 is very exciting and one of my major preoccupations these days. We're coming along very well. We are in the process of connecting 15 campuses with OC12 level connectivity, that's 600 megabits per second approximately. We will have that infrastructure in place by later this spring, and be connected to the VBNS around that time, certainly no later than the beginning of summer. And we hope to be able to demonstrate some of the applications over this network for the Internet2 meeting in San Francisco in September.

JB: Oh, wow.

GM: Very exciting.

JB: That is exciting.

GM: I should remind all of the folks that are participating that you can join the session in either of two ways. You can join the phone conference directly by dialing 734-647-2803, and you'll be immediately here on the air with all of us. Or you can come in via the Events link off the CREN homepage at www.cren.net. Either way, if you want to send us a question via e-mail, just either click on the Website's point for e-mail or send it to ccs@cren.net. That'll reach all of us here. Let me also point out that it will be question and answer throughout, and we really would appreciate hearing from you questions and comments at any point along the way.

JB: Very good, thanks, Greg. To all of our listeners, this is our second series of Campus Communication Strategies Tech Talk, and I'd like to remind everyone that if you do miss a session and want to pick up on some of it, that you can in fact replay the sessions with the real audio from the net at any time. They are out there and they are available. David, let's go ahead and get started on our major topic for today, and that is campus networking issues and how we get linked up to the world. One area in which the campuses generally share a concern today is that we keep looking and hearing about how we ought to be moving towards virtual universities, distance learning, distributed learning, and yet all of these programs depend on good access to online resources, no matter where our students might be. What's happening here? Are there new technologies evolving that can help meet these needs?

DW: Yes, I think that there are developments in several different arenas, and that people involved in support of campus networking and campus networking strategies need to understand how all of these interoperate. At the one level, on the campus itself, we're building digital libraries, that is, online resources that will enhance and support the distance learning and virtual university concepts. On the other hand, there is increasing bandwidth in the traditional Internet that we know and love (we hope) and that bandwidth will be used to deliver a lot of this material. And the third level--I think that this is the newest and probably the least well-understood area--is getting the high speed access into individual homes or wherever you happen to be working. And this third area is what I like to think of as an urban area network, that is the equivalent of what we have on our campuses, but distributed out through the metropolitan areas, providing relatively high speed connections into our homes. By combining these three links, that is, the metropolitan area network, the increased capacity of the traditional I nternet service providers and the higher capacity on our campuses, that will provide the communications infrastructure in order to make a reality out of the virtual universities.

GM: Where do you see dial-in modems fit into that picture? They're certainly not very high speed. Is this something other than just simply what you have today, but you're working hard to get past that?

DW: Well, I think like many technologies, for example, if you look at computers themselves. When computers were first invented, there was one kind of computer, a single service, as it were, and everyone made use of it, whether you were just adding a few numbers or trying to do a digital library. Nowadays, there is a tremendous diversity in compute powers, and you have computers in your wristwatches and your microwave ovens, and you have supercomputers with terabytes of memory for different kinds of problems. And in the same way, modems are one of the mechanisms that we use to communicate. The reason I focused on the higher bandwidth solutions in discussing the virtual university requirements is that the amount of information that we want to transmit is so great that traditional modem technology really is not going to be sufficient to support it. If you're trying to download full text with graphics and perhaps motion, you're simply not going to have the patience to sit there and wait for it to show up. So basically you'll drop out of the virtual university because it will be too frustrating. So I think modems will have a place. Certainly, modems are the technology of choice today for those that have to periodically connect to the Internet or to their campuses.

JB: Let's stay with the dial-up modems for just a moment, David. What direction are those going? If I don't want to use a dial-up modem, are there any other choices or options today to solve the problem of higher bandwidth into the home?

DW: There are some options that are emerging. One option which has been available for some time is ISDN, the Integrated Services Digital Network, which provides 64 kilobit channels, approximately twice the speed of the high speed modems today. And you can combine channels, so you can get multiples of 100 kilobits or so through (inaudible). The problem with ISDN is it's not available everywhere, and it tends to be quite expensive. The traditional service bureaus that offer dial-up modem access do so at $20 to $30 a month, whereas ISDN access is typically $30 to $100 or so a month, plus you end up paying usage charges. So it's not taken off very fast. There's been a lot of market resistance to ISDN. The emerging technologies are things like cable modems and things like the DSL series of technology. We often refer to it as XDSL, where X can be replaced with A for Asynchronous or H for High Speed, etc., etc. So the XDSL technologies and the cable modem technologies are available in an increasing number of areas in the country, and these will provide up to multi megabit speeds into the home.

JB: Are you testing any of these right now out in California?

DW: We within the University of California don't yet have access to the cable modem infrastructures, although we are negotiating with a number of the providers for that access. We hope as part of CALREN 2 to interconnect with the XDSL service providers in California to be able to support the very high speed access into homes via that technology.

JB: Okay. In terms of when we're planning programs and things like that with these technologies, is it safe to say that probably for the next two or three years we need to really plan on the capacity at the level of the dial-up modems?

DW: I think it's safe to say that for a couple of years, that will be by far the most prevalent mechanism for access to resources remotely. But I think that if the XDSL technologies roll out as successfully as at least the vendors believe that they will, that very quickly the community of interest that really wants to participate in distance learning and virtual universities will quickly understand that this is quite an exciting opportunity, and I think that it will change very quickly. But it depends a great deal upon the market and how it's positioned.

GM: There's some real important issues, though, of the economic model underlying this rollout and whether the university itself is expecting it will have to pay some cost of the rollout into the community, or whether the university will run the service itself but charge end users a cost-recovery kind of rate, or whether the university will simply encourage private vendors to go after this, and it has important implications in terms of who has control as well as is this an add-on cost for the university or is this an add-on cost for the individual students and faculty members?

DW: Absolutely, Greg. You raised a very central question about this whole issue of communications infrastructure and how it's supported. And I think it's more than just a technical question. It's very largely an administrative and a philosophical question, and I think that all of us in our universities have to struggle with this question. Let me try to couch it or at least offer some components of this problem to think about. First of all, I like to be guided by where we want to be in, say, five years, and if we looked at the evolution of the telephone system, it started out as very much a private enterprise system. If you wanted a telephone system in 1890, you built it yourself, and you were lucky if you could call somebody down the block because they might not have one. And it evolved over time to the point where it's a utility, like electricity, and so there is a whole market economy around supporting this thing and delivering the service to your doorstep. So I think that's where we want to go ultimately with data communication services, very high speed data communication services. So I don't think the university wants to get involved in building the urban area network. I think we very much want to look to vendors and partners to do that, but we might want to spur them on with some incentive, for example, a critical mass to start a market. The University of California at Berkeley, for example, has 40,000 people associated with it. If we could get 10% of those to be interested in the service, that would be far more than is needed to define a critical mass for that market. So we can prod the vendors in a number of ways. But a more fundamental question is, how are the costs of these things allocated? And I think that here we have a real uphill struggle, based on the history of data communications. Data communications started on the campuses, at least, as a free service, very much like the library. In fact, we sort of looked at the library model as a model for funding this kind of service. The problem is that you cannot keep up with demand with a library or patronage funding model. As demand increases, you have no ability to increase the source of your funding other than diverting funds from other activities. And so I think the campuses really are faced with a very fundamental question, which is how should the cost of this kind of service be allocated? In many cases, campuses are now charging for dial-up modem services that they operate. Berkeley, for example, has a $10 a month service for dial-up modems. Many other campuses are outsourcing the dial-up modem services, getting rid of their modem pools and simply telling their community that they should find one of the dozens of Internet service providers out there that offer dial-up services. I use an Internet service provider because I travel a lot, so wherever I am in the country, I can simply dial a local number and gain access to my Internet services. So the cost of dial-up access to the campus is more and more being shifted to the user, and I think that this is a perfectly appropriate thing to do. It's very much like having cable television in your house. There's public broadcasting service, educational programs on PBS. We don't expect PBS to pay for a cable service into our homes, nor the university to pay for a cable service into our homes. We provide for ourselves this communication service because we know that it's an important part of what we want to do in life. And I think Internet access is very similar. So I think the concept, then, of looking at cost allocation models that involve the end user to some extent--maybe subsidized, but at least to some extent--is where we have to go.

JB: It sounds, David, as if the universities really got into this and providing it free set up some expectations that we now have to struggle with. So we move to an economic model where the user, in fact, starts paying for what they get.

DW:. Absolutely. And I think that if you look at resources that are constrained, that maybe an economic model that is largely subsidized is okay, but a resource that needs to expand really needs to have an economy model that closes the loop back to the end user. In other words, if the end user demands more service, the end user has some responsibility for finding the funding to support that service.

JB: Isn't this linked with the fact that I know that a lot of folks, myself included, figured out, well, why shouldn't the university pay for this, because it really doesn't cost that much. Is that in fact changing, or what's happening that's causing the cost equation to be changing now?

DW: There are two things that are happening, and we've been talking about one aspect of it. I think maybe in a few minutes, we can talk about the other aspect of it. The two aspects I'm thinking of are the cost of the access to the campus network from off campus and the cost of access from the campus network to the broader Internet in general. There's of course a third component, which is the cost of the campus net work per se, and I don't want to get involved in that too much, because that's a different kind of model. But the cost of access from the outside world to the campus is actually not cheap. It's far more expensive per computer than the cost of access on the campus. An Ethernet connection on a campus may cost a couple hundred dollars to install, and then the ongoing costs are fairly minimal. The cost of a dial-up modem is on the order to $60 or $70 a month. Not a month, I'm sorry, a year, and so -- I'm sorry. These numbers are dropping out of my head. It is, in fact, about $60 or $70 a month, not a year. It sounds like a lot of money, but that is in fact the total cost of equipment, amortizing the equipment, the telephone lines and the staff support to keep these things working.

JB: And the $60 or $70 a month, David, is that on campus, then?

DW: Yes, that's if you have a modem pool on campus (inaudible) size, 400 or 500 modems, and you're operating that. Now, I can say that one campus that I'm familiar with has had about 600 modems for a population of 30,000 students. That is far too few. It really should be more like 3,000 modems for 30,000 students, but they can't afford to put in 3,000 modems because this cost of $50, $60, $70 a month is just unbearable, unless there's some way to recover that cost from the end user.

JB: So in other words, as people are using the network more, then that's another factor that's causing the general costs of the network to increase.

DW: Yes, and as people use the network more and more, then the cost of access from the campus network to the broader Internet also needs to increase. The cost of that access is becoming very significant for campuses as well.

JB: What about the stuff that is actually going over the network. We haven't talked about Internet2 yet, but we're talking about certainly some of the applications that are starting to evolve on the net and some of the demands of those particular applications, then, over the network. Right now, as a user, I don't have to pay any more if I'm just doing e-mail versus if I'm doing a lot of graphic and video work over the net.

DW: That's right. In fact, if you're like most people on most campuses, you don't have to pay anything at all. You might pay for the network connection in some sort of flat rate way, but I don'' know of any campus that is charging differentially by what people are actually using the network for. But I think with Internet2, that we are going to see a need for a more granular economic model. That is, the impact on the network of sending electronic mail is far different than the impact on the network of sending real time, full-motion video. The costs involved in supporting that infrastructure are far, far higher for the latter case than for the former case. And so it depends on how the campus wants to allocate costs, but it seems fairly straightforward that the campus would like to say to the user who is doing real time video, "You're going to have to share more of the cost than this person over here on the other side of the room that's just sending e-mail." And that's simply because the demand that that one person's placing on the network is so much higher than the other. To keep up with the demand requires that that cost be borne somehow. So with Internet2, we're really looking at a differentiation of Internet services into a wide variety of things, so you can request fourth class delivery of e-mail if you want and pay very little, and you can request first class delivery of video streams and you will pay more.

JB: That's a really interesting comparison, in terms of levels of service in postal service versus the levels of service on the network. That's probably a pretty good analogy, will help people understand some of the cost implications that we're talking about.

DW: Yeah, I think it's very similar.

GM: You raised Internet2, and I'd like to talk about another aspect of this, which is if we're delivering these services to this urban area network, the residential user and so on, and we know that in many instances we're going to be going through various commercial vendors, there certainly is a huge time delay between the point at which some of the learning materials and so on will be delivered on campuses and when we've worked sufficiently with various vendors for whether it's an ADSL based or cable modem based or whatever based access, so that they understand quality of service in a way that is going to work smoothly with what we're doing on campus. Isn't that going to slow down our effort a great deal? Isn't a lot of the virtual university still going to end up on campus for a long time?

DW: I think that you've raised quite a number of interesting points, and in order to try to sort that out, let me try to separate that issue of the virtual university per se with the issue of getting the communication services sufficiently broadly distributed. I think one model for the virtual university is that existing campuses would work together to offer courses via the existing or Internet2 networks in their areas of specialty. For example, if the campus has a particularly strong physics department, then maybe they would offer virtual university courses in physics which students at other campuses could take via the Internet2 network. So the virtual university concept per se is not inhibited by lack of access to homes. It would be broadened by access to homes because then the students wouldn't necessarily have to be at any particular campus, which definitely raises other issues, of course. But I think that the access to homes certainly is something that we have to work very closely with vendors, the communication service providers to make sure that they do understand the requirements. I know from personal experience that trying to work with cable television service providers to (inaudible) data communications infrastructure is like talking Greek. They don't have any notion of what that kind of service that entails, the level of reliability, the level of quality, the response to problems. They're used to somebody having a grainy picture on the television or calling once in a while. The telephone companies, by the other token, who will support the XDSL services, are used to telephone calls, and if you can get dial tone and you can dial, then your phone's all right, as opposed to is there an error on the line that's causing packets to be lost. Well, the line's up and packets are getting through. It's just that not all of them are getting through. So I think that what we might see is the emergence of special purpose communication service providers, or more specifically, data communications providers such as the at-home cable modem service which specializes in providing data communication services over cable television infrastructure. Or things like special phone companies making use of the XDSL technology, either a subsidiary of the existing phone companies or separate service providers, that will take responsibility for that end to end service. So in any case, I think that we in the university community have to be very proactive in working with communication providers in our areas to insure that the quality of the service is there when we need it.

JB: You have certainly raised an awful lot of questions, David. Those are interesting concepts, that universities in fact do really have to communicate with the providers to insure that we communicate clearly the kinds of requirements that we need for the virtual university environments that we want to have.

DW: Right.

GM: You touched upon not only the cost of access to the campus from off campus, but also the cost of campus access to the Internet. You want to come back to that? Because certainly high speed access across the world, not just the United States, that's bad enough. But lots of exciting possibilities, lots of exciting dollar figures.

DW: Yes, indeed. Well, this, of course, fits into a larger concern that many of us have, which is the economic model of the Internet in general. We are used to the economic model of the telephone system, where we know if we make a local phone call, it's kind of a flat rate, costs a dime, or whatever. And if we make a long-distance call to Japan or Australia, it's going to cost us a lot more and we better be careful how long we stay on the line. With the Internet, there is the same problem, but we don't think of it in that way. We think of our Internet connection as being a flat rate, and we don't care where the bits go. And I think that we have to start being concerned about that, and I think we have to work with the Internet service provider industry in trying to come to a consensus on economic models that don't destroy the value that we've all found in the Internet, but at the same time, provide economic incentive for the service providers to improve the quality of their service and make it affordable at the same time to us. The cost of Internet access from a campus to the broader Internet has gone up manyfold since the Internet began in earnest in the early 80's. Up until quite late in the 80's, many campuses were paying well under $10,000 a year for their Internet access. And students loved it, faculty loved it. It was a very useful service. By now, in the mid to late 90's, the costs of Internet service for most campuses is well up in several hundred thousand dollars a year and rapidly approaching half a million dollars a year. And with Internet2 and differentiated services, it's going to go even higher. So it becomes a very significant issue for campuses about how to allocate that service, how to insure that the quality of that service is worth what they're paying for it, and ultimately, how to justify the expenditure that the campus is faced with.

GM: Do you have a model for what you're going to do within CALREN?

DW: We have some ideas, but obviously, these need to evolve, partly because there's no really good way of collecting billing information on the Internet. You certainly don't want to charge per packet. I think packet tax, as we call it, is a pretty silly way to do it. But you do want to find a way with some sort of algorithm of identifying the impact that a certain stream of traffic has on the costs of the network, the costs of providing an adequate service infrastructure. And so I look at things like the amount of traffic that you feed into the network at a time when the network is more than 80% at capacity. So you're then participating with everyone else in essentially overloading the network, and it seems to me the cost of using the network at that point should be higher than the cost of using the network in the middle of the night when the traffic is pretty low. Another aspect of it is, how far do the bits go? It's relatively inexpensive for Berkeley to send bits to Stanford, which is 50 miles down the road. It's actually quite expensive to send them to Sweden, because now they have to go not only across country but across the Atlantic Ocean and those links are very, very expensive. So if I help to congest a very expensive pipe running across the Atlantic so the service provider is forced to increase the capacity of that link, that impacts cost considerably. And we have no recognition of that now. So I think the two aspects of Internet traffic flow have to be looked at. One is, in fact, what I call horizons. Over what horizon does this thing go? Or beyond what horizons? And the other is, what is the actual impact on the infrastructure? These are hard problems. They're not easy to solve.

JB: In terms of the planning for all of this and the sharing of the expenditure of all this, I know that one of the working groups for Internet2, David, is the working group on the quality of service. And in terms of certainly the questions about when I want to use the network and how far they have to go really directly impact program planning for a virtual university class, for example. If in fact we do want to have things that are activities that are synchronous, certainly the time becomes very important. But if in fact, the time doesn't have to be synchronous and we can do things at lower demand times, then the cost in fact might be less. At the same time, if I really need to have a lot of packets going over at one particular time, I'm going to have to be able to have some confidence that that will in fact happen. What is Internet2 doing to help address that confidence and that ability to really be able to predict with confidence that I'm going to be able to use the network and get the packets across when I need them?

DW: That, of course, is directly in the area of quality of service, as you suggested. And the whole issue of how to deliver quality of service and insure or guarantee at some level that that quality of service will be available when you need it is one of the problems that we hope to solve in the Internet2 projects. It's not a solved problem now. Let me offer you some suggestions about what that might look like. First of all, I think it is very difficult to guarantee in an absolute way that you will get a certain data rate with a certain error rate between any two points in a statistically multiplexed infrastructure. It is inherently chaotic. Chaotic in the mathematical sense. You cannot predict what traffic you're going to see. It depends entirely on what 10 or 20 million people do around the world. And so all you can do is describe this thing in a statistical way. You can say that 98% of the time, by our measurements in the last three months, we get a certainly quality of service with a certain loss rate. So I think that the notion of quality will have to be defined in some statistical way, so we might, for example, talk about a fourth class service for e-mail which would have no particular guarantee of bandwidth, but maybe a guarantee of packet loss that's no more than 1%. One percent is actually very high, but no more than 1% packet loss for e-mail, because you know you can retransmit it and it will get there. For video, you might want to have a loss rate of no more than .1% or .05%, because you can't afford to lose packets with video. And you might want to have a guaranteed bandwidth of some sort, so you can say something like 99% of the time, your packets will have a bandwidth available to them of no less than 10 megabits and your packet loss rate will be no more than .1% 99.5% of the time. Some statistical measure like that makes a whole lot more sense to me than to try to define an absolute guarantee. Now, having said that, you have to translate that into the nature of the network infrastructure. So measurements come into play. One of the Internet working groups is measurements, where we try to characterize the actual behavior of the network and the kinds of traffic flows that are on the network so that from an engineering point of view, we can say if we are to achieve 99% capacity, etc., and low loss--if we are to achieve that, given the kind of traffic flow that we measure here, we need to improve the infrastructure in the following way. In other words, we as engineers, it will give us the parameters that we need to do the forward engineering to insure that the quality of services are met and the guarantees are actually followed through on. It's a complicated picture, but you see where I'm going. First of all, we have to determine what quality is. Second of all, we have to figure out a way of engineering the infrastructure to deliver that quality.

JB: And also then to track it so that you can then loop back the economic pieces of it, right?

DW: Absolutely, that's the third piece. How do you allocate the cost of doing this?

GM: And then get everybody that's a part of the whole communications process to each do it the same way, or in a way that interfaces at a standard level.

DW: Oh, absolutely. Doing it in the same way is critical. Even telephone services don't always charge in the same way. In some places you have flat rates for ISDN, and other places you have message unit rates for ISDN. But suppose we had one infrastructure provider that charged for the delivery of packets to you, and another one somewhere else in the world that charged for transmission of packets from you? Now, at one end of that equation, somebody pays nothing, and at the other end, they pay twice.

JB: I was thinking maybe we ought to do the airline model.

DW: Which is?

JB: Which is it just kind of depends upon when you go in. It can be anything at all.

DW: You never know what your tickets will cost!

JB: David, there's another area that often comes up when we're talking about distributed learning and just learning in general, and that is access--not just access to the network, but access to certain and particular kinds of resources. I know that in some cases, for example, there are examples where students who are on campus in fact do have access to certain library kinds of resources, but that students who are off campus in fact, even though they have good technological access to the campus, can't access those resources. And in fact, in somewhat of an ironic situation, they have to travel to campus to get them.

DW: Yes, and that is a very interesting aspect of the evolution of our communication services in that we would like to make the communication service itself transparent to the individual, so it doesn't matter where the individual is. We've talked a lot about high bandwidth into the homes and what you do when you're traveling, etc., and all of this is sort of with the implication that the communication service is not an impediment. On the other hand, it is an impediment if the access to resources is controlled by where you are connected, so we have to move beyond that to a point where individuals making use of the network can identify themselves appropriately in some electronic way, roughly equivalent to a drivers' license or a credit card or some other identity that we're used to in the analog world. So I think that electronic or digital authentication systems are becoming critical to the kinds of things that we want to do on our campuses and with virtual universities and access to resources. You mentioned the digital library resources. It's not only on campus that these resources are becoming available. Publishers themselves are putting resources online that are protected by contractual arrangements, licensing if you will, and so only the community that is licensed to access that information can do so. But that community is certainly not in the publishers' warehouse. That community could be anywhere in the world, so it's the same problem. It's not just campuses, it's really the general electronic commerce problem of identity and authentication in the network world. So one of the things that we're trying to do at the University of California is to develop standards for a universal authentication system such that everyone associated with the university can have an identity, an appropriately defined identity, that will allow them to make use of resources wherever they happen to be, whether it's at home or on the campus or traveling. And I think that that's a very critical aspect to making these networks useful because without that, as you point out, it's sort of moot having high bandwidth access. So then let me follow on from that to say that, beyond authentication, i.e., having an identity per person, you also have to understand the characteristics of that person. And we don't think about this so much consciously in the real world. Somebody shows you a student body card, they're really showing you two things. One is, they're showing you their name--John Doe, or whatever--and the other thing is they're combining an attribute with that name, which is that of "student." Now, you may not know whether they're a graduate student or an undergraduate unless that's also on the card. If it is, then it's a third attribute, etc. So we not only have to do authentication, which is identification of a particular individual, we also all have to have ways of associating attributes with that individual. So we're also developing, then, attribute databases which can be used for authorization when gaining access to resources.

JB: We are getting into a very complex and interesting area here, David. I just want to take just a moment right now and remind the folks that are out there that now is also a very good time for them to direct any questions that they might have specifically to you by sending us an e-mail. I didn't want to interrupt your train of thought, which was so good, but I do want to invite our folks to come in. Having interrupted you on the authentication and also the building attribute databases and all that, does this all link back also to Internet2?

DW: Absolutely. As we talked about earlier, the economic models that will become necessary for the viable, long-term support of Internet2 require that we allocate costs on some level, whether it's real or artificial money, back to the individual users who are making an impact on the network. The only way we will know who those users are is through some authentication system that is part of the digital environment, so I think that the authentication and ultimately authorization issues are very much a part of the Internet2 project. Perhaps not on day one, but definitely in the long term because it's the only way it's going to be viable. For example, if quality of service is actually technically possible--and I think it will be within this year, probably--then, how do you allocate it? If you don't allocate it, if you allow anyone to request any quality of service with no feedback, then there's absolutely no reason why anyone would request less than the very highest quality of service. Why should I suffer? Doesn't cost me anything to request the highest quality of service, so I'll just request the highest quality. And if everybody does this, then there is no quality of service because it goes back to you're overloading your network. So in order for differentiated services to actually be effective, you'd have to have some way of making choices and decisions about when it's appropriate to use a higher quality of service and when it's not. And that gets to authentication and authorization.

JB: Sounds like we all have an awful lot to learn here on this yet. In terms of the quality of service and in terms of allocation, some of the things that you're saying also remind me about the situation we have on campuses with classrooms, where faculty in fact kind of queue up to get better classrooms vs. noisy classrooms or less technologically advanced classrooms.

GM: Right, and anybody that's been in charge, whether it's classrooms or offices, space management at a university or a business or anyplace knows that that's probably as big a war as any that you'd want to face.

DW: We have a phrase that we use --"Space -- the final frontier!"

GM: And what you said also about the authorization and authentication, it seems to me that that's an area that, as universities find more of these resources such as the digital library resources and they want to know, "Was such-and-such a student taking this test over the Internet?" or a whole set of things as to who's out there, and then you extend that back to your urban area network--the person isn't on campus, so you've got to have an interface with those people that supports all the same thing. This authorization and authentication is really very important.

DW: Yes, I think it's very critical, and I would add that some of the points that you mentioned also raise concerns about privacy. And I think we always have to be sensitive to that set of concerns. Whereas in some sense we need to know who's out there, particularly if somebody's taking a test, obviously we have to associate that with some individual so we can update their records appropriately. But in another sense, there are certain cases we don't want to know who the individual is, or don't need to. For example, making use of online digital library resources. It may be that the only authorization that's required is that this person is a member of the university community or perhaps a faculty member of the university community. In that case, we want the authentication step to lead to an anonymizing of the authorization so that the service only sees the authorization and not the authentication of the individual, you see, and it's a subtle point, but I think it's very important in the university environment.,

GM: Very important point. Anybody that's wrestled with issues over the years with someone coming in wanting to know who read what books in a library instantly understands that one.

DW: Absolutely. You don't even what to have records of such sort. I think that applies to a lot of different areas.

GM: I think we're running close to closing out.

JB: I think we are. I can't believe our time has gone already today, David.

GM: We've had too much fun today.

JB: Really interesting distinctions and questions that we have on the table for us. Let me just thank all the seminar members for being here and to be sure to use the e-mail of ccs@cren.net for any follow-up questions. Also want to be sure to invite you to join the third Expert Event in this series on Wednesday, February 25 at 4 PM Eastern Standard Time when our guest expert will be Ken Klingenstein from the University of Colorado, and we'll be talking with Ken about the challenges of managing these networks, and also techniques for making it easier. Be sure to view the Virtual Seminar sections on CD on network management, if you haven't seen them, before next time. Also watch for your message if you are signed up for a special announcement. If not, just go to the www.cren.net web page and the entire schedule is right there. Thanks to everyone who helped make this possible today: the board of CREN, the Corporation for Research and Educational Networking; our guest expert, David Wasley from the University of California; Greg Marks and his team at MERIT; in Michigan, Jason Luke the audio encoder from UM Online; and all of you who joined in. You were here because it's time.