The Boundaries
of the Network

Session 3b: Architectures of Innovation, continued

Thursday, September 6, 2001

David Isenberg: It occurred to me that in the last session Tom Freeburg was saying to Larry Lessig that just because you can't find them [on-line music traders] doesn't mean that they're not out there. We have an alien in our midst -- one of those great musicians that the recording industry has been totally unable to find. That's Joe Weed. Joe is not only a musician, he's also a record company owner, a producer, an arranger, a recording engineer and a studio owner. Joe, do you want to say a few words?

Joe Weed: I don't know how relevant my experience will be, but because David's a good friend, I'd like to share a little bit. First of all I do have my own independent record company. My experience all my life has been marginalized from the centralized big four or big five.

Every musician that I've ever worked with or known who's been involved with those big five has regretted it. I've spent 30 years in the music business. I don't have a single acquaintance who has done well as a piece of fodder of the majors. So that's point number one.

Point number two, we've been talking about Napster a lot today. I write a column for a bluegrass magazine and was railing about Napster, especially last summer in July when we the Erie Canal and went into a little coffee shop one morning and read about the injunction. And wrote an article for the magazine about Napster. And then I realized that I've never even done this. So I said, before it disappears, it is time for me to learn about Napster.

So I downloaded Napster on my Mac. The first thing I noticed was that I had to sign an agreement that I would never give it away, that they owned the software, that it was proprietary. I could never share it, re-engineer it, or make copies of it. So that seemed to me to be patently illogical based what they were trying to do with my and everyone else's music.

I just produced a new album called Swannee and everyone here is welcome to have one for free. Fortunately the NY Post gave us a very nice review on the third of July, the day before Foster's birthday. Stephen Foster was born July 4, 1826, the day the U.S. turned 50, the day the founding fathers Thomas Jefferson and John Adams died. Foster's music survives to this day. Many of us think of it as folk music, but he's the guy that wrote Oh Susannah, Camptown Races, Hard Times Come Again No More and Way Down Upon the Swanee River.

Stephen Foster died with 38 cents in his pocket in New York City's Bellevue hospital on January 13, 1864. ASCAP was founded on January 13, 1914, specifically to honor Foster. They chose the 50th anniversary of his death to say never again shall someone who can create this wonderful material for our country die with 38 cents in his pocket.

Most of you have heard my music. I would assume a large portion of you listen to NPR. Morning Edition and All Things Considered like to use pieces of my first album, Waltz of the Whippoorwill for their snippets between stories. Unfortunately, BMI, which is a competing organization to ASCAP worked out a deal with NPR where NPR can use our music for free because it gives us "exposure."

My dear friend Steve Palazzo, the guitar player here, says that people have died from exposure. So I think frequently about sending in with my mortgage payment a little clip from NPR. Saying here you go, this month I'm sending this in instead of money because this is my exposure, my coupon for being in the music business.

Anyway, there is a good living to be made if you are creative and if you have an understanding and creative spouse and family. There's a good living to be made in the music business, but we need to protect intellectual property as much as all you need to protect your own property.

It's fun to sit here and listen to these big issues being talked about. Especially Napster. I know a lot of what we are talking about Napster today is not music-specific, even though music is the driving thing that made it all happen. And we are talking about intellectual property and people turning on in huge numbers to this sort of function. Anyway, that's my three minutes.

Isenberg: Okay, now I'm going to turn the floor over to David Reed, who's going to continue on the theme of the Architectures of Innovation and Bob Lucky will offer his 38 cents worth later.

David P. Reed: I have too much to say and no time to say it.

Man: So speak up.

Reed: I've become completely convinced, I think I was convinced when I was born, that architecture is everything. Architecture matters. Most human thinking seems to not recognize architecture as valuable. We have this [grand old?] architecture around us.

In particular if you look at Larry's chart, the three layers, the physical layer, which things are made of; the content layer, which is what we as consumers all care about, there's this middle space that's usually viewed as in the way. Sometimes it does get in the way. But more importantly, what is viewed as necessary but not useful, if you looked at economics.

I've spent a lot of time trying to live in the world of economists because that's where the arguments always take place even though technology isn't about economics. If you look at economics, for many years, they had this architecture-free view, that economics is about atomic goods and distribution to consumers and how you got there really didn't matter. Mostly it was a pain in the ass how you got there.

Then a variety of writers started focusing on the architecture of various things. They tended not to be working in the traditional economic big schools of thought. Bob [Meeses?] is one of those guys and Coase is probably the more modern thinker along these lines. Coase worked in a law school because he couldn't be hired by an economics department.

Coase won his Nobel Prize, not just because he said transaction costs are important, but because he said structure was important, because structure is what creates the transaction costs in the first place. He was in some sense the economist who crystallized the notion that value was created or destroyed by structure and that structure actually mattered.

So I got interested in the interaction between architecture and economic value and discovered that there are a bunch of people who are interested in that question now. They tend also not to work in economics departments. They tend to work in business schools, where it's practical and applied. Carlos Baldwin is a friend of mine, and there's a guy [???] who I've read but I have not met.

What they write about is modularity and structure, among other things, and how those things create value. The reason I bring that up is that I'm still very sensitive to the fact that people can say like [Tom] Nolle did in David's SMART Letter that the Internet was created by a bunch of hippie anarchists who didn't care enough abut economics.

In fact we all did care about economics. The number of conversations we had in naive economic terms around the design of the Internet was huge, and it wasn't about anarchy, it was about designing something that could both scale (which is what I want to talk about today) and create value and pay people for the value. For example we had lots of conversations about billing and we left out billing because we thought it perverted the structure of the network. We all knew that the telephone network was about billing and that 70% of it's architecture was designed to make it billable.

Now if 70% of its architecture and 70% of its resources are devoted to designing something to be billable, the engineer inside of me says, well, maybe there's a way to do billability cheaper. Why waste 70% of the resources on billing? So we said lets put billing off, because we'd like to figure out a way to do billing later. Not because we thought you shouldn't bill.

Maybe we thought that charging for minutes is not right, charging for sessions is not right. We made lots of intuitions that charging -- that the particular charging structure -- distorted the whole thing. What we wanted to do is create incentives for people to build networks and create content and hopefully do it in a way [that was different].

So I want to talk about wireless. I'm shifting gears. My new five year intellectual focus is the scalability of wireless technology. I got fascinated by it five years ago by talking to people who were asking, "Can't we just have ad hoc networks that grow organically that are wireless and that would change the capital structure of the industry because capital would come from the users not from some capital source that has to bet that 10 years in the future that they are going to recover it?"

What we didn't want was another Iridium. Iridium was a beautiful technologic product. You cannot fault it on technology. Talk to the guys at Motorola. Not only did they build great technology, they also built great technology that updated over time could be the best technology.

But the interesting question is why didn't it work? Well the reason it didn't work is because the only way to get value out of this was to figure out what the applications were -- that's what they thought. They predicted an application that at the time made sense, and they architected the system for that application -- voice telephony from remote locations. I think they thought maybe they could eliminate the need for cellular networks with a lower cost alternative because it would be more universal.

In any case, pricing of cellular was really high at the time and they didn't predict significantly lower pricing. But what's really interesting is that they built a very brittle architecture in the large. It's not that it didn't work, it's not that it was particularly late, but rather it was aimed at the wrong target. There really isn't a market for [the Iridium] architecture at the moment that we can find. Lots of people have tried. The military is about the best and that's not a big market.

In wireless, there are a bunch of questions that are pragmatic ones, like what's the next thing in wireless. But I think the more interesting thing long term is, is wireless part of the long term solution, or is wireless something that will bother us once in a while and be an irritating thing around the edges, or a thin strip -- the last five inches of distribution so you don't have to have a cable. Exactly where does wireless fit in in this world?

The fundamental question is, is wireless scalable. It keeps coming back to that. This is where the intellectual battle is being fought, and it's in fact where the policy battle is being fought as well.

If you go to Washington, you can have lots of conversations with people and they will tell you there is a scarce supply of spectrum, just not enough. We need to do something about that, because there are some people who want to use more of it.

The question that keeps coming to mind is, is it really scarce? On the opposite side, there are people who find a lot of spectrum lying around unused. Somebody points out to me that if you walk with a spectrum analyzer through downtown New York, which you would expect to be heavily used, you'd find a few cellular peaks, which are very small in terms of bandwidth, and no other use averaged over a very long period of time.

The reason for this is regulatory. But the reason the regulatory thing exists is because we've taken a particular path. Where did that path start? It started largely with the Titanic. I've heard this from Bob Lucky.

I just want to read a quote. There are a whole lot of interesting things about the Titanic and you can find all about on the Web. Here's one quote that I found.

So [???], one of the ships that happened to be not very far from the Titanic on that fatal night sent to the Titanic the following message: "Ice Report -- much heavy pack ice and great number of large icebergs, also field ice.

The Titanic said, "Keep out, I'm working Cape Race."

In other words, the wireless operators on the Titanic were sending passenger traffic -- it was the first liner to support full passenger radio. They were sending passenger to Cape Race to relay off to various other places, and that was considered priority. Weather reports were not high priority at the time. So they received this message and they basically ignored it, because their goal was to prioritize passengers traffic.

Two things were going on there. First, it was an example of interference. This was not a message to the Titanic, this was a message to all ships on that channel. The rule was one user had priority and the other users had to shut up. The other thing that was going on, was that no one had recognized that safety-oriented traffic ought to have priority in marine applications. So the FCC was largely created around these two things.

One was safety because it was clear that radio had more of an impact on public safety than anything.

Isenberg: The FCC was created in direct response to the Titanic disaster?

Reed: No, the first structured regulations, which ended up in the FCC, and then ultimately in the way we do this worldwide, said the government has an interest in safety. The government didn't have an interest in anything else about radio, but this has been parlayed into a much more interesting question, and the source of that path has been the notion that there is only so much radio to go around.

If you got to have safety and there is only so much to go around, we ought to keep the regulation in one place because that's the place that matters the most.

Man: At that time was it still spark-gap radios?

Tom Freeburg: No, radio was much more sophisticated than that by that time.

Reed: There were some other interesting things going on. For example, radio came with a human. If you had a radio, you had an operator. And your operator worked for Marconi or Telefunken. They didn't work for the ship owner, and because of proprietary concerns. The Marconi operators were not allowed to talk to the Telfunken operators.

A lot of things were going on in radio at that time that caused the regulators to want to get involved. They wanted to force interconnection for safety. So regulation is good, there is a lot of goodness that regulation created. I just want to put that on the table.

But it's not a very old system of regulation. And the metaphor, which led to the architecture of radio, is the metaphor of land and property rights. And that fit with a very naive notion of radio that was perfectly sensible at the time, that radio was narrowband, it was possible to confine your transmissions to a narrow bandwidth, and two people could not transmit on the same frequency at the same time and expect the receivers to distinguish the two transmissions.

The technology at the time said the exciter and the demodulator/detector worked in this way that was very frequency specific and it was cheap to do that way. That was the only way you could do radio. Now radio could have evolved another way, but somebody got the great idea that there were bands of spectrum and that you could regulate according to these bands. I think that’s the last contact that regulators had with business.

A physicist knows that radio is basically an electro-magnetic field wave. And there is no such thing as frequency for magnetic fields. A frequency is a particular decomposition that Fourier invented, and people have found useful as a way decomposing what is going on in some complicated system into a manageable coating, if you will. In particular it is done by frequency.

I’m not going to dwell on physics, but the interesting question is if you transmit supposedly in a frequency band, in fact the information even in a narrow-band signal is splattered all over the spectrum, and maybe gets splattered at lower power in other areas of the spectrum. But because the transmission varies in time, it’s got components everywhere, even if you filter it a lot.

This results in some rather strange things, where people invent modulation techniques that seem to violate various physical laws by transmitting more than they can possibly transmit, because in fact they mange to encode information way outside their band, but at very low levels, so it’s somewhat like spread sampling.

Then they come out and say, I’ve created this great new radio and then when you actually look at them, the way they actually work is that they are actually transmitting more information than will fit in a band. So bands are not fixed. In fact the way we manage the bands is that every time we divide them we do something like draw little areas between the bands. They used to be called guard bands, but there’s a lot of things going on there. When we divide space up we do the same thing, so that whenever you divide something you lose a whole lot of capacity.

So the result of this property model is that as you subdivide you tend to waste more and more. Almost every radio system works this way. Except that in modern times we started understanding a few more interesting things abut the physics and exploiting them. Actually we can build really neat radio systems now that work and don’t waste the spectrum anywhere near as much. Cellular is a great example of that because it basically geographically reuses spectrum. So cellular uses a frequency band much more efficiently than, say clear channel broadcast radio.

So one interesting question is, if it becomes really easy to build cell sites, why don’t we build wireless broadcast TV networks that use cellular architectures, which would much more efficiently use the spectrum rather than transmit at really high levels in order to bleed very low power signals out to the edges.

An even more interesting question gets raised as systems get really dense. Now we live in a world where radio is really dispersed. And there are a set of engineering solutions that were practical at [a certain] point in time, like traditional broadcast radio, which has a transmitter that covers a metropolitan area. [In this architecture,] you have the fringes that don’t work very well, and if you are too close to the antenna it’s too loud, but basically it works reasonably well. You design intelligence into the receiver.

But we are going to a world where radios are everywhere -- a world is station ubiquity, if you will, and radios are two-way everywhere. The question is, do the regulations and rules and design principles that used to work very well in the dispersed situation work here? I submit that there is a technological discontinuity somewhere here where it stops working.

There are a bunch of interesting technologies which I’ll call to your attention. One is called OFDMA. It is actually being used in 3G (or 4G) network designs. OFDMA basically says you don’t have to have this wasted space between the bands. Because if you choose the center frequencies of all these transmissions to be orthogonal, that is properly spaced, then you can simultaneously decode all transmissions at once. If you simultaneously decode them, in fact, the power can bleed out between the bands and it doesn’t matter.

So in an OFDMA network, you can get much more capacity. Capacity doesn’t go down with the number of stations or the number of bands used, which is a nice property.

Tom Freeburg: I just wanted to add one thought. You are saying in general terms exactly the stuff I’ve been preaching to my management and all my potential customers for the last five years. You've got a good large part of it. I just wanted to throw in one thing. When you talk about distributed radio becoming the only form of networking, you come to a problem of, lets call it a multi-hop problem in radio, but I’m reminded of the highway system in Italy, where up until just a few years ago, the only way to go up and down the Italian coast was to weave in and out of all the highways, and it took a painful long time to get from village A to village C. Then somebody built the Autostrada system, which is a toll road system. It’s a private corporation and you pay tolls. You know what? Anybody going any distance at all uses it. There’s no question at all as to the way to go if you have any distance to go.

So I’m only suggesting that there will definitely be a place in this future that David is describing so ably here of multi-hop, mesh networking, complete peer to peer networking, whatever you care to call it. There is going to be a place for long haul toll ways. By the way, that network is already in place, what we call the Internet today, without thinking about the last mile.

Reed: I might be talking over people. I’m a little worried. So OFDMA is this interesting technology -- it’s called orthogonal frequency division multiplex access. There is also something called OFDM, which is kind of a lower level [?] that doesn’t affect the access part.

If I were George Gilder, I’d go around saying OFDMA is unlimited bandwidth. But I’m not. OFDMA is not a magic bullet. It’s just an illustration of the general phenomena that hasn’t been practical and now is. The reason it hasn’t been practical is one of my two messages. One is computation.

Computation helps. You can trade computation off to create a much richer thing. We only had coils and capacitors in the old days. OFDMA requires that you do a Fourier transform on every symbol and an inverse Fourier transfer on the other end. You can only do that if you have cycles to throw away -- that's the reason why nobody’s ever done it before.

Now there is another thing. I want to talk about space. So we have frequency, and the old property model said that to achieve the illusion that we have property that frequency is the thing that we could trade or allocate. In the old days frequency was allocated by fiat, now the proposal is to auction it. To achieve that illusion of property, we waste resources. We waste more resources the more we chop it up.

Now the other dimension is how do we manage space? [The traditional idea was that two distant cities can use the same band, but adjacent cities must use different bands.] So what you need is a careful allocation process to make sure that space is reasonably efficiently used. But the problem is that power drops off at the square -- or other power laws -- from the point of transmission.

So you have tremendous differences in power that have to be handled by receivers. That results in a definition of interference that is extremely conservative in the design of receivers, because they have to be very sensitive to very low power; on the other hand they have to handle very high power. To make practical receivers you have to waste resources in the whole way of assigning frequencies. In fact you [might] barely detect the signal [in certain circumstances, yet] but it [can be] perceived as interfering with something somewhere in the fringes. Spatial division historically in radio has been considered bad, and has created extremely wasteful use of spectrum as well. This wastefulness accretes over time because people make the rules more and more conservative. And you start to play games with the rules, rather than the technology.

The cellular guys have some motivation to start thinking about the space problem.

Man: Isn’t it true that the experience has encouraged innovation -- not intentionally -- because you need bandwidth when you are pressed on spectrum? I’m an advisor for Centrepoint Broadband Technologies, and what they can fit on a single channel is amazing. Nobody in their right mind would design anything so fiercely complex to transmit the signal unless they were backed into a corner and said you have *this* to do everything else.

Reed: I think you are actually right. If you think of innovation in isolation, yes, lots of innovation happens because of this. And in fact, if you think about it, lots of innovation happens because of the desire to control things. We got great innovation about managing digital rights management and encryption and unbreakable stuff -- there's lots of innovation going on.

It is not really a question about whether lots of innovation is happening. It’s to what end. Is the innovation being directed toward creating value for society and consumers or is it focused on creating value to the suppliers and so forth. And sometimes the question is mixed because the same innovation can be useful to . . .

Man: If you are arguing about capitalism, it’s a long story , but . . .

Reed: I’m not going to argue about capitalism. I will argue about property laws. And the reason is that I’m a capitalist but I’m not a property capitalist, I’ve decided. Capital is the deferred reward. It has nothing to do with property. There are lots of ways to get a deferred reward, only some of which can be structured as things that behave like property. What we’ve had is a bunch of radicals to say we can think of everything as property. If that were true, that would be great. I would claim that it is very bad to think of spectrum as property because in fact it results in incredible waste because there is no way to make spectrum look like property.

Here’s the final example that I think is the most interesting. If you had a fixed geographic region and you had a fixed frequency band -- a band limited system in a defined space -- there is an interesting question you might ask. I asked this question and I found out that no one could answer it.

Here's the question: What’s the scaling law that describes the capacity of this space as the number of stations goes up?

So you have stations. I don’t care whether they are randomly distributed or whether they are arrayed in rectangular coordinates. You specify the problem to that level of detail and just tell me if you have N stations in this region, how much capacity in bits per second -- actually the best measure is how much capacity in bit meters per second because if you transmit a bit twice as far you clearly got twice as much capacity in some sense.

So, how much capacity in bit meters per second? If you can get one bit, one meter that’s a bit meter per sec. If you can get one bit two meters, that’s two bit meters per second.

Man: You are missing a whole bunch . . . a huge number of variables.

Reed: Holding all other variables constant, can you answer the question, what’s the basic scaling law of the system.

The answer is you can’t tell unless you specify the architecture of the system. So you have different architectures. You could have an architecture where every time slice sends all their messages to some central point and it then sends it out. That’s kind of an access point oriented structure, or maybe a multiple access point structure. Over all such architectures, there must be a best architecture. So what’s the scaling law of the best architecture?

Well, it turns out that no one knows the answer to that. Even though it is purely a result of physics -- this is a physical system -- no one knows the answer. If you knew the answer, you’d know what the best architecture was -- at least you’d know what you were shooting for.

There are several possible answers. And this is what I want to get at -- the structure of possible answers.

The FCC model, which is also the model that is supported by the incumbents, is that as N increases, the total capacity, which we will call T, is constant. Which means that the per-station capacity, C, declines as 1 over N. Now if that were true, that would be a classic property model. The property model says you got a fixed amount of stuff -- you have to allocate it. You can’t make much more of it. It’s very expensive to make much more of it, if you can make any at all. So you need an allocator that allocates it to the best possible use. That’s the property model. For that kind of property, property rights work real well.

So the second question is, maybe it’s unlimited. In the George Gilderish sense of the word, it’s fixed only in the sense that technology can make an infinite amount. You have a total of infinity and it doesn’t really matter how many stations there are, it’s just a question of technology, so this is the unlimited case and I would claim that basically because of physics, we know that’s not true. It can’t possibly be true. Just like DWDM isn’t really unlimited. It is just a lot less limited . . .

So if that’s not the case, is there possibly a model where the total capacity goes up with the number of stations, or worse yet is the true model something worse than the FCC model, which is that total capacity goes down with the number of stations. Fortunately, we don’t have to consider that one because there is at least one known architecture, and it’s basically the architecture that Tim Sheperd sort of worked out. He was a PHD student at MIT and got his Ph.D. for this.

It basically shows that the total [capacity of the system], which should be a curve, goes up as the square root of [the number of stations] with his architecture. Now we don’t know if that’s the best architecture or not. A number of other people have done some interesting papers. By the way, that’s not so great, because it’s sort of a funny kind of scarcity. What happens is the square root of N over N is a curve that is 1 over the square root of n individual capacity. So as you get more and more stations, everybody gets less and less. It is just that they get less and less slower and slower than the FCC model, so there is a need for allocation. But what’s interesting about this is that the only thing Shepherd needed to do to get to square root of N growth was a simple thing, which is to allow stations to repeat each others traffic.

So besides computation, you get repeating. And repeating is an interesting form of cooperative behavior that isn’t normal in a market. In order to get that, I don’t know of any architecture that gets square-root-of-N growth without cooperation. That says that individual actors acting alone can’t achieve the best possible use of diverse resources.

Man: What about the tragedy of the commons?

Reed: Well maybe there is a tragedy of commons without additional rules. You need some kind of rules.

Man: Individual actors acting along will not manage the resource to the collective benefit of the whole.

Reed: Right. Which is why you need an FCC or some kind of regulation in the first model. Or you need a market that is driven by externalities, which are the value people place on the usage. What we are saying here is yes you need a market, in this model, but you don’t need a property market because subdivision doesn’t work. Cooperation works better than subdivision. But the more interesting question about this is is this the answer for the physical world, or is this just the answer for an architecture that repeats? And this is where no one knows. So what’s the answer for the physical world?

Well, I can tell you an upper bound on the physical world, and it has to do with the work of some people at AT&T Labs and others in what’s called space time coding. There is a practical application of this in the system sort of an architecture called BLAST.

Isenberg: Bell Labs is actually Lucent now.

Reed: The key idea is that if every station is transmitting and every station is receiving in a space, then every station has a slightly different path to every other station and what’s really interesting is that this works even better if the path is a multipath environment where things are reflecting and bouncing off stuff, because then the paths are really different.

You have N-squared different paths which fill in a matrix of N-squared coefficients, although they are actually impulse responses, and if you have N equations and N unknowns you can back-solve the entire matrix because you have enough information. So, in that model, which is a little bit impractical because in order to do the computation you have to assemble all the operators in one place and obviously they are distributed. In that model you get a growth that grows linearly with N and in capacity it grows flat per station.

The question is, between these two there is a whole range of architectures that are possible that involve not just cooperation, but cooperative computation. If you can share a little bit of information with your friend, he can compute on it. And if you share a little information about the impulse response between you and various partners out there, ultimately you can at least get N. I still haven’t said whether you can get more than N, but if the physics work this way, and regulators think it works the other way there is a huge problem…and that’s where I’ll stop.

[applause]

[Gap to change tape.]

Reed: (explaining his graphic) This is a fixed region, no new spectrum allowed . . . N transmitters and I’m assuming N growth, because what’s interesting to me is how does wireless capacity grow when everybody is carrying four on their belt and they all converge in this room. And cellular systems don’t work in that case.

Man: *Today’s* cellular systems don’t work.

The X-axis is N growth and the vertical axis is a measure of capacity -- I am using bit meters per second. N is hand-helds plus cell sites because right now just like in the last talk, hand-helds are just stupid terminals on a backbone. But if you take into effect that the handhelds can compute and cooperate with each other, there are a lot more resources in the system then the backbone. Unfortunately it’s not under the control of a capital provider.

Isenberg: Can I ask you to work with Joe Sterling to create a labeled version of the chart?

Freeburg: David definitely is on the right track. I want to point out two very simple things all based on the concept that you don’t need to get nearly as complicated as David was talking about and describing accurately what’s going on in the industry today. But you don’t need to get nearly that complicated to get there. Just small nudges have already moved the technology development of radio in the direction that it takes.

For instance, six years ago I published a paper that talked about, if this is an axis of some measure of spectrum efficiency or spectrum effectiveness in the direction that David was talking about, and this other axis is robustness to interference, it turns out that as this robustness to interference improves (and you don’t have to get into regions where you can work in the presence of interference stronger than your signal), as that robustness improves spectrum efficiency, the number of bits or bit meters that you can push from 1 Hz goes to infinity. All it takes is directional antennas and a very little amount of interference robustness.

Second point is that both of these views of how you manage spectrum, one is property, one is some kind of common thing, already exist today in the US, the so called UNII bands at 5GHz, [where there are] 300 MHz that we get to use as we want.

I had a lot of input into the way those rules are written and I personally am convinced it’s enough. If you like, the camel's nose is already under the tent. There's a shared kind of thing and that’s enough. There’s enough there to do everything that we are talking about and that needs to be done for the next half decade, maybe decade.

Reed: Let me respond. I think we mostly agree. I think the problem is not with the technology, but with the ability to deploy the technology, which might be due to regulation, or dumbness on the part of operators, or lack of capital or whatever. The problem is deploying these innovations. The other thing is it’s not any particular innovation, which is why I emphasize OFDMA is not the next DWDM. Don’t go invest in OFDMA companies or UWB companies. Right now you are going to hear a drum beat for wideband radio, or impulse radio, which has some of these properties. It does not come with an architecture. Basic ultra-wideband does not come with an architecture. The game is in the architecture fitting with the technology, not the other way around. And UWB is interesting, but not necessary.

Freeburg: I have bet my career at Motorola on being able to bowl something meaningful through the particular knothole that you have just described. I've been aiming at this for several years now. It’s birthing as a commercial product as we speak, very first phase simple, but I really believe, and have for a long time that’s what we are going into production with. The third trial production runs about to begin.

Isenberg: Here it is. [Holding up plastic transciever.] One of these is providing our backhaul link across Little Harbor to the Oceanographic Institution.

Freeburg: There’s no property rights in that at link at all and its just enough sophisticated technology to meet all of the stuff that David has described.

Reed: An interesting thing happens when you start talking about all these technologies, that other kinds of property rights start showing up, like the right to put the antenna on the roof or the pseudo interference that might happen if the system is designed wrong, or whatever. And you know a company like Motorola can play that game because it’s big.

Freeburg: Actually it is easier for a smaller company to play that game.

Isenberg: 3 min coupon!

Larry Lessig: So I spend my coupon to make one small plea. You know because we’ve been doing this stuff a long time, I completely agree with you about your intuitions about where this can go. But I want to talk about strategy. Because the real enemy here is not people talking about property, the real enemy is the existing FCC bureaucracy that’s talking about licensing spectrum for particular uses.

And there’s another movement out there that I agree ultimately won’t be the best way to allocate spectrum. This is the Tom [Hazlet?] view of perfect property rights. It’s the Ronald Coase view of let’s create real property rights and spectrum, which is not you are allowed to use this spectrum at this time for this particular use. Rather it is, "You have the spectrum, sell it, do whatever you want with it. It’s your resource just like this piece of property is your resource."

That view is so radical relative the existing FCC vision that it would completely blow up spectrum management if it were adopted. And I think there is a great opportunity for us to recognize what Tom’s work is doing. Tom’s implementing what you’re talking about, I think that what you’ve done with the FCC and your work is the future, it’s brilliant and great, but we get there if we ally in some sense with people who say we need a radically new way to organize spectrum. One part of it has to be a commons and this way that Tom [Freeburg] is taking advantage of in unlicensed spectrum, it may well be that a huge chunk of it’s going to be property-like. But we have to get away from the existing method, which is that people buy the rights through political power at the FCC…That’s the evil. And it’s not ideas about property, it’s corruption of the political system.

So if you talked about licenses and said you know lets have a world where we experiment with lots of commons and lots of property, but lets get away from government, centralized, Soviet control of spectrum, we could find alliance with an extremely important right wing world out there who would say, yeah, that’s where we ought to be going.

Man: Can you give examples of what you are talking about? Are you talking about radio broadcasters using the spectrum?

Lessig: The catholic model is, you buy a chunk of spectrum just like you buy a piece of property. You sell it for whatever use you want, you can use it for cellular radio or you could use it for unlicensed…whatever…you have the right. And it’s unlike existing allocation of spectrum, because in the existing system you have this particular chunk for this particular use for this particular period of time. You are not free necessarily.

Isenberg: Bob Lucky has prepared some remarks and then we’ll continue in the all hell breaks loose mode. Unless somebody wants to use a coupon here. Quick question?

Man: At some point can somebody take what this discussion is and tell us what the underlying technology . . .

Isenberg: Yes, after lunch we have Dewayne and Tom and one or two others who are going to delve into what this wireless thing is in a more nuts and bolts sort of way.

Man: If property model doesn’t work, what is the alternative model?

Isenberg: What *are* the alternative economic *models*

Bob Lucky: I don’t have any prepared remarks, in fact, I don’t know what I’m going to say. I should say that I happen to chair the technology advisory council at the FCC and Dewayne chairs the committee on spectrum management. We do talk about these issues regularly at the FCC and there is a certain willingness to listen so we are not exactly silent there. But as Larry said, there certainly are entrenched political forces that far surpass [the power of] any advisory council that they have.

But I don’t want to talk about that. I want to talk about Dark Forces. I have been a member of the Dark Forces for my entire life. I am currently with Telcordia and I used to be with Bell Labs many years. Telcordia is independent and not associated with any particular Dark Forces, but since all of our money pours down from Dark Forces I suppose that makes me a captive of these forces.

Let’s try to look at what’s going on there for the few minutes that I have. In another month it’s the witching hour for my AT&T options to expire and they are almost worthless. The question that I think about sometimes when I wake up in the morning and that is, "If I were C. M. Armstrong, what would I do?"

What would you do if you were C. Michael Armstrong?

To me you’ve been given the short end of the stick and it’s your own fault, it’s what you took, the long distance and technology undermined that. The cost structure of the way it was built, the historic way it was priced has gone out the window. Frankly I don’t know what I would do.

David, you were there for many years. Do you have a simple answer to that question?

Isenberg: No, I have some good hindsight observations, but that ain’t fair. I think that’s one of the hardest jobs in the world.

Man: I’d sell AT&T Broadband to Comcast, I’d take my retirement package and leave it to the next guy.

Man: I'd break it along customer lines -- wholesale, retail.

Man: I'd listen to my customers, which they never do.

Matt Oristano: I would destroy the whole pricing principle and install flat rate long distance and move the whole thing to IP non-switchable.

Lucky: Now the RBOCs are in a different situation and usually when you think about Dark Forces you think abut those rather than AT&T, which is underwater, not just my options but the disability charts.

What do the RBOCs want to do? They don’t want to be caught in this business of selling commodity stuff. So they all talk about moving up the value chain. Let’s sell services. Let’s get value added services and build on that. The problem is nobody knows what these are. But somehow they are there. Now the initial grasp for these higher valued services are QOS [Quality of Service], SLA’s [Service Level Agreements]. Lets have higher quality bits. My bits are better than yours. They will draw a premium. I’m not a commodity, I have good bits as opposed to these lousy bits.

I’ll just bypass the idea for a moment of provisioning VPNs. But SLAs and QOS has been this Holy Grail now for a number of years, and in fact the technologists buy into this too. And for sort of a dirty reason -- it is one of the few good technical problems around. You know, I mean lets make these bits good. You go to conferences on telecom and all the papers are on QOS, like there aren’t any other problems here that are worth looking at. So just from the standpoint of finding interesting problem, QOS is a great problem for engineers, let’s do it. Management likes it because these are superior bits that will draw a premium. So for quite a while now this has been sort of a Holy Grail.

Now how many people here believe this will work? QOS as a marketing thing.

[Very quiet audience]

I don’t see any great show of hands. I’m very skeptical myself. too, but it is amazing how the industry has bought into this.

Peter Cochrane: Well you bolt CRM on it and you have the Holy Grail.

[Laughter]

Man: . . . and CTI

Man: . . . and BLS

Lucky: It is amazing how we engineers have bought into this and never really questioned.

Isenberg: Ours is not to question why . . .

Victor Blake: If you go to conferences, I’m a complete anti-QOS person, any conference you go to, you start to talk like that, what happens is that the Worldcoms and Sprints and Genuities and Level 3s light up and say, "Victor, you’re stupid…You don’t know anything.."

Man: They’ll all be bankrupt in the next year, so . . .

Blake: Probably true, but in the meantime they wield a great amount of influence in standards organizations and their voice dominates the intellectual effort to work on problems. I’m actually amazed that no one raised their hand here because I would have thought . . .

BL: How many people would say they are anti-QOS? The forces of industry here are aligned.

Man: Their point is hey look that’s what the carriers are buying. I can make what’s right or I can make what sells.

Isenberg: QOS is about allocating a scarce resource, and it ain’t scarce.

Cochrane: The problem is there are flavors of QOS. You say to an engineer QOS, they are going to tell you about down time and performance issues. If you go to the marketeers, God knows what it is.

Man: At some point everybody is going to have to build their networks to a certain 99.999 it’s just going to be there . . . everybody’s going to have to be there.

Lucky: Yeah but maybe that will just come naturally -- the rising tide theory.

Man: I don’t know what the anti-QOS issue is.

Gigi Wang: I’m going to take a stand. I believe in QOS. From the customer side, it depends on how you look at it. If you look at QOS from the customer side you have a choice of different qualities of service and what you want to pay for. Right now for long distance, you get to punch different codes. What’s really interesting is that SingTel has 011 as their premium quality service, and 013 is the budget stuff -- 019 is VOIP.

Man: It all goes through the same pipe.

[Laughter]

Lucky: So what motivates the RBOCs right now. One very important thing is that for the first time, they are losing lines. They are losing lines big time. BellSouth a month or two ago, lost the most lines in its history. Land lines are going down. You have the substitution of wireless going on, they are losing second lines. This is a panic situation, this has been their bread and butter.

And then they got this other situation where data traffic is growing, and now 80% of the traffic is data. And yet, voice down at 20% is 80% of the revenue. So you have a total disconnect from where the traffic is and where the revenue is.

How do they survive this cataclysmic shift from where they get money out of the data? And they haven’t figured this out yet. It’s really a desperate situation. We’ve been going for a century with a very artificial pricing structure for telecom that’s not based on any kind of cost or reality whatsoever. Technology has shattered this. It’s drawn out all the underpinnings that held up this artificial thing.

So what does motivate them and what are the hopeful things? I think there are a couple things that might be helpful. We heard recently that Larry Roberts collected all this data on network traffic. It’s on their web site, Caspian Networks. For the first time instrumented the whole core and measured all the traffic going into it and found incredibly that traffic now is growing faster now than ever.

Bradner: Internet traffic.

Lucky: This is very encouraging. And he says in 6 months or so, you will have exhaustion of the embedded capacity that they are just sitting on and using up because they are afraid to buy new stuff right now. At the current rate of traffic growth you will need to start buying stuff to light the dark fiber in 6 to 9 months at the current rate of increase.

I take that as one of the few glimmers of hope. Traffic continues to grow, and that I think is good news.

There is one caveat here. Larry Roberts, as we know, is a good guy and one of the founders of the Internet is also founder of Caspian Networks and sells gigabit routers. So it is very much to his interest that the traffic is growing. But nonetheless he’s a good guy and lets assume that he . . .

Man: It can’t be verified. The source of the data is actually much better if you believe that he actually bought the right to instrument all the Tier 1 guys, as part of buying the rights he had to make it clear that he couldn’t share it with their competitors and therefore he is only aggregating it.

Man: It is not a real scientific study because he won't expose the data.

Lucky: I recommend that you look at the data. It is terribly interesting.

BellSouth said in January that Napster was 5% of their traffic and in June was 1%.

Man: Music traffic has grown slowly despite the drop in Napster.

Lucky: How many of you used Napster in the old days? Okay. How many now use some other music sharing. About two-thirds.

Let me summarize -- even though it is hard to summarize when I didn't really start. A lot of people have the idea that if you could open up this local loop bottleneck that you would stimulate the traffic and the applications to come on. And is there someway we can undo this.

My question at the beginning of the session is what limits the rate we can roll out DSL? I mean, why can’t we all get it out there? I’ve asked a number of the people in charge of DSL deployment at different companies, what are the limits. They don’t seem to know, but a lot of the elements are financial. A lot are just logistical. And a lot of them have to do with the complexity of customer care and all that.

Let me ask question. How many here do not have DSL offered where they live? See that to me is the problem. I ask that question in a number of places where I give talks to computer industry people and typically three quarters of the people will say that I don’t get it where I live. Yet if you talk to the RBOCs they say 75% of our offices are equipped and it’s available there. My theory is that computer executives don’t live in places like that.

Isenberg: Thank you very much Bob. [applause] Now does anyone have a 3-minute coupon that’s burning a hole in their pocket?

Man: Even if you get DSL out, a lot of what they are rolling out is 256 or 512 . . .

Lucky: You only have 3 kbps upstream anyway.

Man: I mean are they going to spend to do another whole cycle again?

Lucky: On the average they all get about 3 kbps per customer upstream. The real fallacy of the argument between DSL and cable modems is that there is sharing somewhere upstream [so the hidden question is] how much have you really got?

Isenberg: We have a three-minute coupon -- Peter Cochrane

Cochrane: Let me give you a low down on ADSL. First of all, it could only come out of the diseased minds of the telco.

[Laughter, applause]

Cochrane: It’s a completely stupid idea. The world is not asymmetric, it is symmetric. Why? First of all, you have a bunch of marketeers in there who don’t understand the very simple phenomenon called cross-talk.

They also don’t understand what a complete and utter ball of wool the local loop is. So you go to the laboratory and you do ADSL and you say, hey, we can do 2 megabits over 8 kilometers, we can do 10 megabits . . . the whole deal.

Then you get out into the real world and , oooooh shit . . . Why does it become uneconomic? Because suddenly you find that the cables are not like the ones in the laboratory. They are in a hell of a mess. There is more cross talk than you anticipated and as you get more DSL out there, cross talk goes up, the speed goes down and you have to employ more and more people to go out there and do customer testing.

In the UK we had this reaction, and I bet you it was the same everywhere else. Fiber Optics comes along, and the decision is made -- not on any financial basis -- to roll fiber into long lines. Devastating success. 90% of the people employed in long lines are taken out. The people in the local loop say, "Ooops, better stop fiber or we are out of a job." The way we can do this is invent ADSL. We’ll promise 50 megabits, 12 megabits, anything you like. Hold off this fiber, guys!"

Now the reality starts to bite now, because when they roll out this technology it is absolutely killing them.

I would ask you to think on this. How come all the population can buy radio, TV, HiFi, cordless phones and satellite receivers and they can install them, but they can’t be involved in the installation of wideband access in the local loop. If we had done the very simplest of things…given away these to everybody closed down the local loop, sent 50,000 people home, saved $12 billion a year in operating costs, recovered all the lead and copper out of the local loop, which is worth a fortune and then rent off the duct space to companies who can then put fiber in, but for the moment 802.11 in the street, bingo, you would have the replication of a situation that occurred in this country in the late 1930s, early 40s where it was not economic for the TV stations to build repeater stations so communities did community antennae TV (CATV), amateurs did it. When it became complex, it became the CATV industry.

[applause]

Peter Ekelund: I hear the US perspective and I think you are really quite far behind. This is too simple a tack for what is happening now. Telecom Italia in hopeless world of copper in Italy is currently going to make a decision this autumn. It is very likely that they are not going to go for ADSL. [It is likely that they] are going deploy fiber in all the ducts as a matter of strategic decision. So there is a glimmer of hope, even among the incumbents, because they know that their copper is completely useless when things go to ADSL. It’s a very big project. The decision can be expected this year. And that will have significant repercussions in industry.

Number two, on the wireless side, as an illustration. B2, the company I was part of founding, has at the moment about 30,000 customers in Stockholm, each with 10 to 100 mbps full symmetric connectivity. They are starting to buy, just accidentally these type of devices and are putting them into their apartments and their houses. And Stokab and other operators are also putting them in the streets. There are about 50,000 of them -- the total investment is about 15 million dollars, and the consumer is paying for that. There is an 11 mbps wireless network that’s virtually being created in the city of Stockholm. This is basically happening now in the hometown of Nokia and Ericsson.

David Curry: I mentioned yesterday a couple of communities in the Pacific Northwest that are converting. The business model for these communities is to build up the networks, finance revenue bonds, and invite, in an open access model, voice, data and video providers to come in. They are going to be overbuilding existing ILEC infrastructure. Now in looking at these community strategies to bring in voice and data providers, it begs the question should we approach other ILECS. Would they be interested in bringing services into their competitors’ territory? And bring them into a community.

I had meetings in the last two weeks with Qwest and Century Telephone. Forget my Canadian metaphor. In meeting with the Qwest folks, they were discussing the thinking and philosophy in Qwest these days, the dark side. The suggestion was that Nacchio and friends the Qwest folks were going to follow this model that this fellow used to describe it, and he described why Wayne Gretzky, the hockey player, was successful. Gretzky as you know is not the biggest hockey player in the world, or the fastest skater, nor did he shoot the puck the hardest.

Gretzky says he's successful because, "I don’t worry about where the puck is. I spend all my time thinking about where it’s going and being there first."

This Qwest middle management fellow said that’s Nacchio's thinking. Now, whether they behave that way remains to be seen. I posed the question over lunch with them. I said, if a community in your serving area is going to overbuild your network and take your dial-tone customers away from you, what would you do? Without a moments hesitation, we’d bring our dial tone into that head end to keep those customers.

A week later, I met with Century telephone, which is a largely rural, 3-million access line, non-contiguous service provider, a local exchange provider, all around the country. And I posed the question to him, "Would you be prepared to bring in voice and data services in a community owned, community built access network that overbuilds Qwest territory?

He said, "In a heartbeat." I asked why. He said, "It’s about as profitable a business as we could ever be in, to bring in that service and be the service provider." That suggested to me -- and those conversations were about a week apart -- where the puck is going. That the service provider new business model is to bring services into -- and the ironic thing is into those territories.

Will they be the service provider into their former dark side colleagues' serving area? Seems to me in that very small sample, that there is a willingness to build these business models.

The Century guy says I don’t have to deploy my capital. My ROI hurdle is much much lower. Let the municipality do it and finance a revenue bond. I’ll make my money delivering and differentiating services in that community. That is a really salable, sustainable business model.

Isenberg: Until it becomes competitive.

Victor Blake: Observations about DSL: I actually very much agree that it can be difficult deploy. Conditions are pretty horrible out there. On the other hand even when I was at AOL, we were doing pilots for DSL. As the manager of architecture I was a candidate to be a pilot if I chose to be. I asked about DSL in my service area knowing that there is a PairGain multiplexer running HDSL about 40 ft. from my driveway, and it successfully carries a T-1 with multiplexed telephone lines on top of it every day and therefore I completely qualify for HDSL, which is more stringent than ADSL.

I was told that DSL can’t work where you are. And I said sure it can, the tech on the street showed me while he was configuring the PairGain box. And still I was told by a variety of Bell Atlantic people that the equipment wasn’t there. They denied for months. I invited them to come to my house to see it. Maybe there is truth that there are technical complexities. I think it would be less rude if they said, you know what Victor, it will work there, but it’s not cost effective for us to deploy it.

Matt Oristano: Along the lines of not true and DSL. For a long time telephone companies very happily provided dry copper pair to alarm companies. As soon as people figured out they could use that to do home grown DSL, the telephone companies all of a sudden started saying we can’t find any more dry copper pairs, and they started terminating the tarriffs. Viscous gamesmanship in the regulatory area.

Adina Levin: I would like to hear David Reed explain what he means about having a model for capitalism that’s about deferred reward and not about sheer property rights.

Reed: This is really flaky. I run into danger trying to pretend to be an economist so . . .

Isenberg: That's OK. You stayed at the Nautilus Motel last night.

Reed: If you ask really what capital is, I think everybody, even regular economists would say capital is about the ability to defer -- capital is an accumulation of money, which is just a deferred reward. Somebody who has capital has excess something and they are saying I’m willing to take that off the market and give it to you because in the long run I’ll get more back. Money is a great vehicle for time travel. Stored value, that’s the right term.

So capital is really useful when you have to make decisions that happen over a period of time. Capital is completely useless in a situation where manna is raining down from heaven. You don’t need capital to allocate the manna. There is more than you need. If you have more manna than people you don’t have an allocation problem you have a disposal problem. We don’t normally live in that kind of economics although I would suggest that the manna is a very good analogy.

The other thing is that I think in the world of markets. Markets are great allocators. They tend to allocate things -- with the exception of broken markets -- they tend to allocate things to the highest possible use because everybody trades off value. I believe that works. I’m not a socialist because of that. I think most of the time you can make that work.

But the interesting question is what do you trade? Does it have to be something that behaves like property, which is something that has a strong conservation law? Or can you trade other things? And we trade other things all the time in the same money system we trade services, we trade options. Options are not property in any real sense.

Isenberg: Services have a conservation law.

Reed: No. Labor may have a conservation law. The amount of service you can provide with a certain amount of labor is very architecture dependent. So if you have me try to design this building you are wasting your money, whereas if you have me design your network it might be useful.

I think there is an interesting market for communications capability. But if you really want to create a market that works and rewards the right things you ought to focus on not trying to map it into property, but into what it behaves like. There are a bunch of things that tend to be closer to that. A futures market is not about property, it’s about claims, contingent claims, which are different than property. It’s a much richer economic market. The interesting thing about futures markets is that they can cope with more futures than actual goods. That doesn’t mean I know how to make a futures market. I think that may be a better place to look.

The second thing is that when you are talking about structure, the corporation itself may hold assets, but the value of a corporation is different than [property]. The structure of a corporation and the ways of doing business have value that are not a simple property -- you can't divide it and get half the value.

When you are talking about a communications system, that’s an assemblage of parts with an architecture that behaves more like a corporation than it behaves like a field of wheat. A field of wheat you can imagine buying and selling a la Hazlett. But a corporation you buy it’s from a very different world. It’s called a market for corporate control. It’s not a market for corporate property -- it’s a market for corporate control and that’s a much richer market. What I’m saying is that there are things you can buy and sell in this space, I’m not sure what they are.

There is a broad analogy from mathematics that I hope will sort of point out that low level architecture can have big effects. Imagine the world as a two dimensional grid. a simple world, and you are measuring the distance between two points. Some mathematicians make these kind of mistakes and then they get corrected, but if you measure the distance between two points, if you could only cut vertically or horizontally the distance between two points is always equal to the sum of the height difference and the horizontal difference. No matter how many points you put in that space, no matter how densely you space the grid you can never get diagonal lines and therefore you always get this distance function that's rather strange.

If you admit diagonal lines, you end up with a very different scaling plot. And a very different system behavior. And we see this in peers all the time because people do design bad architectures like that -- the rectilinear communications model -- but that doesn’t mean we have to force a rectilinear model on any currently non-rectilinear world. So that is my metaphor of the day.

Isenberg: Does that answer your question? I hope not. Let’s break and have some lunch, take a swim, the water is warm, and reconvene at 2 p.m.

(return to top)
page last modified: July 10, 2002

Copyright 2001 isen.com, inc.