The illustration in the article “Goodbye CRT” [November] shows a person carrying a piece of equipment, its cord dragging along behind him. Everything important I learned in kindergarten—among these rules, don't run with scissors or with your shoelaces untied. Later I learned, probably the hard way, not to walk with cords and ropes and strings dragging.

William J. Eccles

IEEE Life Member

Terre Haute, Ind.

The article has a great deal of useful information on TV displays, but it grossly distorts the technical facts on plasma displays. As an example, it states that plasma has limited longevity. In fact, the current life of a plasma display (60 000 hours) equals that of a CRT. Also, the measured data show that plasma has a longer life than the life of LCD TVs with fluorescent backlights. O’Donovan’s statement that “plasma’s phosphors exist in a hostile environment; the electron beam in a CRT is much kinder to phosphors” is not borne out by the measured data, which show that both technologies have the same equivalent long life. It’s an obvious cheap shot for O’Donovan to say that the 60 000-hour-life guarantee of the plasma manufacturers is “based on a few hundred hours of testing.”

The author’s claim that plasma is power hungry is equally fallacious. When typical TV images are displayed, the data averaged over a large number of TV sets show that plasma, LCD, and HDTV CRT TV sets all take the same power per unit in the screen area. This amazing fact is borne out by the data. Of course, small diagonal LCD TVs take less power than large diagonal plasma televisions, but the power is the same on a per-unit screen-area basis. Plasma is a power-on-demand technology that uses power when light is needed by the image. Since the average picture level (APL) of typical TV images is 20 percent or less, the power-per-unit area of the plasma TV can be equal to the LCD TV, since it uses fluorescent backlights that are on all of the time when even a very small number of pixels are lit. For certain low-intensity images such as those in a typical movie, the plasma TV takes considerably less power than the LCD TV.

The third factual error is the burn-in issue attributed to plasma. Early plasma TVs did have a burn-in problem, but this problem has been solved. The burn-in on today’s plasma TVs is no worse than the burn-in on a CRT. This was achieved through phosphor research and improved gas mixtures. By comparison, there was no mention of the LCD burn-in problem. My 37-inch LCD TV, which is less than three years old, has some permanent, ugly gray blotches in large areas of the display that substantially degrade the image. These blotches show up only at the lower-luminance gray levels, so I attribute them to a threshold voltage shift of the active-matrix thin-film-transistors. This is a serious LCD life issue that is rarely discussed by the press.

Finally, plasma TVs will continue to be strongly competitive in the future, because their manufacturing costs are lower than LCD TVs. This is primarily due to the lower materials cost for plasma, which has a structure more similar to the low-cost CRTs. The LCD device structure is more like a semiconductor than a CRT, so LCD TVs will continue to have higher manufacturing costs than plasma TVs. Price comparable plasma and LCD TVs today and you can see this difference. Making a larger LCD plant will not solve the major manufacturing cost problem for LCDs, which is the cost for materials.

Larry F. Weber

IEEE Fellow

New Paltz, N.Y.

Metcalfe’s Law

“Lots of luck” rightly describes my reaction to Robert Metcalfe's blog post in response to the “Metcalfe’s Law Is Wrong” [July]. The graceful and right thing to do would be to admit that the original equation cannot apply to larger and larger networks and that the proposed equation is better suited for that.

Partha Kaushik

The debate about Metcalfe’s Law is an important one, but the question of whether it is right or wrong is too simplistic. Are Newton’s laws of motion right or wrong? We now know that these laws are not applicable in some circumstances; nevertheless, they have superb predictive and explanatory power in many practical situations.

Instead, we should focus less on trying to find one law that applies to all networks for all sizes and at all times and more on understanding the different regimes of behavior (and points of inflection when regimes change) that networks may exhibit. For example, the value of a network and of the connections it enables are surely also a function of time as well as of size. Metcalfe himself suggests this in his response to the criticism of his law, although he relates time dependence to the costs of networks, not to the impact of their use.

A connection that may have very low value at one time may grow in importance as it is discovered and exploited, in much the same way as neural connections in the brain can be rewired. Other connections may decrease in value, just as some neural connections may be pruned over time. As a result, the total value of all the connections may change, even if the size of the network remains constant. Furthermore, as a network grows, perhaps (as Metcalfe also surmises) its value stops increasing once it has reached a certain size—or it even starts to decrease, thereby demonstrating diseconomies of scale. As has been said, “If something cannot go on forever, it will stop” (Herb Stein, circa 1980).

On the question of network costs, which Metcalfe links to Moore’s Law, note that the costs of connectivity are, perhaps unfortunately, not driven only by the costs and performance of semiconductors. Civil engineering costs, executive pay, and other phenomena, which contribute significantly to total network costs, are either flat or increasing. Nevertheless, the power of networks is increasing substantially. The value associated with this power does not seem to grow smoothly, but rather in fits and spurts, as network capabilities reach and exceed particular thresholds. For example, the increased value of first-generation 3G mobile networks over 2G at first was marginal at best, as was ISDN over dial-up. But once real speeds in several hundreds of kilobits per second can be achieved (as with cable modems and xDSL in fixed networks and emerging wireless HSDPA and EV-DO networks), then a new set of broadband services becomes economically feasible and operationally practical. So we should really be looking for points of discontinuity or thresholds as for continuous functions of value.

It is extremely unlikely that any one formula can be applied to all networks of all sizes at all times. It will be more fruitful to understand how to categorize networks so as to be able to apply different formulas to their valuation, depending on the category they are in, and to delineate the boundaries between these different categories, along dimensions such as scale and maturity.

In financial valuations, another scientific principle states that the act of observation changes what is being observed. The analogy in financial valuation is that once networks (or more generally businesses) start being valued according to specific metrics, these valuations will be affected as the people in these businesses start to “game” the metrics and thereby alter the expectations and hence the findings of the financial appraisers.

As for Metcalfe’s Law itself, its lasting value lies in the ways in which it has encouraged us to try to understand the value of networks, even if other approximations to the truth turn out to be better as we learn more (the value of Newton’s work was not diminished by Einstein’s theory of relativity). Metcalf’s Law may even be a reasonable approximation to the truth for networks of certain sizes at certain times. But as with any tool with far-reaching implications, it is dangerous when it is applied without understanding the scope of its applicability.

Martyn Roetter

IEEE Member

Boston

My guess is that the value of a network as the number of users grows should follow a sigmoid curve, the usual curve for describing the value of technological innovations, chemical reaction rates, economic development, and the like. Growth in value starts slowly, rises rapidly, and then reaches an inflection point, after which each additional user adds little. If you have 6 billion people and of those 100 are networked, the value is low, and going from 100 to 101 gets you only a small increase in value. If you have 10 or 100 million networked, the value per addition gets higher, because the network is starting to cover a useful section of the population. As most people become connected, it becomes less and less valuable for each user to do so.

The shape of the curve is obviously influenced by the combination of the various subcurves of the various subsets of communicating users. Except for a few spammers, very few people will ever send e-mails to more than a few hundred people. Cellphone usage plans that encourage friends and family imply that there are statistics to back the idea that most people communicate in a narrow circle, perhaps following a power law as they contact parties in broader and broader circles less and less frequently. For example, a person might frequently contact his or her workgroup but rarely contact the entire company or, even more rarely, everyone in the industry. The technology reaches a peak value when enough users have signed up for it.

People tend to forget the sigmoid curve, especially during the steeper portions of the ascent, and they can understand linearity and exponential growth but do not recognize that subsequent slowing and saturation are part of the pattern. Consider the predictions for the growth of such nations as China or India, which are based on the exploitation of their extremely limited urban and educated classes. Too few in those classes or industries can grow at all. But with just enough growth in some areas of these classes, industries can grow around them. If you keep trying to grow these industries, however, eventually you have to wait for the educational establishment to be restructured, along with the requisite restructuring of both cities and the countryside.

The difficult part of assessing technology is locating the spot where you are on the curve. Another example: we heard about the home computer all though the 1980s, but it wasn’t until the 1990s that it was widely adopted. Still, recognizing the existence of the sigmoid curve can provide us with some insight and give us clues as to what signs and portents might be relevant.

Seth Steinberg

Metcalfe’s Law states that the value of a network grows as the square of the number of connection points to the network. This is usually stated as the number of users where each user is assumed to have their own connection point. The law was popularized in 1993 by George Gilder, writer of “The Telecosm,” in the Gilder Technology Report, and chief cheerleader of the telecom/Internet revolution/bubble. Authors Briscoe, Odlyzko, and Tilly argue in the Spectrum article that the actual growth in value is n*log(n) and that the original formulation was bad because it directly led to the speculative excess of the telecom/Internet bubble.

Put baldly, Metcalfe’s Law says that if I have a network and you have a network, and we connect our networks together, they are worth much more than either network on its own or even the sum of the two networks. The more networks we connect, the more valuable the whole thing becomes. So the point of Metcalfe's Law is that there is a huge incentive for all networks to join together into one completely connected internetwork. This has come to pass, first for telephones and then for computers. Thus my position is that Metcalfe has been proven correct and that it is academic to argue whether the “value” (whatever that means) of the network grows quadratically or exponentially.

We need to understand the context when looking at Metcalfe and Gilder’s arguments. As Bob Metcalfe says in his blog entry, in 1980 when he devised Metcalfe's Law, he was just trying to sell the value of networks and create business for his company 3COM. This was at a time when an Ethernet card cost US $5000 and flinty-eyed accountants would argue to reduce the size of their network buy, while he would argue that they should increase it.

George Gilder is the person who foresaw a single interconnected Internet at a time when there was CompuServe, Prodigy, AOL, and thousands of local bulletin board systems. All of these were swept away by the Internet revolution except for AOL, which managed to ride the wave by co-opting it. So Gilder was correct as well, although he was eventually carried away by the force of his own argument like many who listened to him.

I posted this comment to my blog (http://bandb.blogspot.com/2006/11/metcalfes-law.html) on 5 November.

Richard Taylor

IEEE Member

Asking whether Metcalfe’s Law is right or wrong is like asking whether to “reach out and touch someone” is right or wrong. A successful marketing slogan is a promise, not a verifiable empirical statement. Both Metcalfe’s and Moore’s laws prove this: engineers—or more generally, entrepreneurs—are the best marketers.

The IEEE Annals of the History of Computing recently conducted an empirical examination of Moore’s Law. Surveying the historical record on the determinants of the law, Ethan Mollick shows how the law was adjusted to fit changing economic conditions, and specifically, the increase of foreign competition. These observations served as a rallying cry for a nascent industry and an ingenious method for addressing economic justifications that were “needed to continue technical advances.” Moore told Mollick that his original paper/prediction was “an attempt to show the cheapest way to produce microchips.”

Similarly, Metcalfe tells us that his is a “vision thing.” It helped him jump over a big hurdle: the first Ethernet card Metcalfe sold went for $5000 in 1980. He says he used the law to “convince early Ethernet adopters to try local area networks large enough to exhibit network effects,” in effect promising them that the value of their investment would grow as more people connected to the office network. Metcalfe’s Law encapsulates a brilliant marketing concept, engineered to get early adopters—and more important, their accountants—over the difficulty of calculating the return on investment for a new, expensive, unproven technology. It provided the ultimate promise: this technology gets more “valuable” the more you invest in it.

Moore and Metcalfe have done the world a great service by providing “scientific” justification for investing in new technologies. As Briscoe, Odlyzko, and Tilly argue, during the late 1990s Metcalfe’s Law—and I would argue, also Moore’s Law—were used as a justification for a “mad rush for growth.” But the excesses of the late 1990s were driven more by the old idea of the “New Economy" than by these laws. As Anthony B. Perkins and Michael C. Perkins, authors of The Internet Bubble, reminded us during the mad rush, and as Bernard Baruch said in 1932: “In the lamentable era of the ‘New Economics’ culminating in 1929... if we had all continuously repeated ‘two and two still make four,’ much of the evil might have been averted.” But regardless of the main culprit, the Internet bubble produced the backlash of “IT doesn’t matter,” which is more dangerous" (Briscoe’s term) or “evil” (Baruch’s) than the blind belief in the power of technology to improve our lives.

Blind faith supported by scientific justifications has powered the great, mostly U.S.-based, engine of innovation and growth for two centuries. As we enter a new era in which we will see this engine of growth distributed around the globe, I would like to offer a new law, the Marketing Law, to the future movers and shakers worldwide: the success of your idea will be proportional to the square root of the number of people repeating after you “two plus two equal five.”

Gil Press

IEEE Associate Member

Hopkinton, Mass.

The illustration in the article “Goodbye CRT” [November] shows a person carrying a piece of equipment, its cord dragging along behind him. Everything important I learned in kindergarten—among these rules, don't run with scissors or with your shoelaces untied. Later I learned, probably the hard way, not to walk with cords and ropes and strings dragging.

William J. Eccles

IEEE Life Member

Terre Haute, Ind.

The article has a great deal of useful information on TV displays, but it grossly distorts the technical facts on plasma displays. As an example, it states that plasma has limited longevity. In fact, the current life of a plasma display (60 000 hours) equals that of a CRT. Also, the measured data show that plasma has a longer life than the life of LCD TVs with fluorescent backlights. O’Donovan’s statement that “plasma’s phosphors exist in a hostile environment; the electron beam in a CRT is much kinder to phosphors” is not borne out by the measured data, which show that both technologies have the same equivalent long life. It’s an obvious cheap shot for O’Donovan to say that the 60 000-hour-life guarantee of the plasma manufacturers is “based on a few hundred hours of testing.”

The author’s claim that plasma is power hungry is equally fallacious. When typical TV images are displayed, the data averaged over a large number of TV sets show that plasma, LCD, and HDTV CRT TV sets all take the same power per unit in the screen area. This amazing fact is borne out by the data. Of course, small diagonal LCD TVs take less power than large diagonal plasma televisions, but the power is the same on a per-unit screen-area basis. Plasma is a power-on-demand technology that uses power when light is needed by the image. Since the average picture level (APL) of typical TV images is 20 percent or less, the power-per-unit area of the plasma TV can be equal to the LCD TV, since it uses fluorescent backlights that are on all of the time when even a very small number of pixels are lit. For certain low-intensity images such as those in a typical movie, the plasma TV takes considerably less power than the LCD TV.

The third factual error is the burn-in issue attributed to plasma. Early plasma TVs did have a burn-in problem, but this problem has been solved. The burn-in on today’s plasma TVs is no worse than the burn-in on a CRT. This was achieved through phosphor research and improved gas mixtures. By comparison, there was no mention of the LCD burn-in problem. My 37-inch LCD TV, which is less than three years old, has some permanent, ugly gray blotches in large areas of the display that substantially degrade the image. These blotches show up only at the lower-luminance gray levels, so I attribute them to a threshold voltage shift of the active-matrix thin-film-transistors. This is a serious LCD life issue that is rarely discussed by the press.

Finally, plasma TVs will continue to be strongly competitive in the future, because their manufacturing costs are lower than LCD TVs. This is primarily due to the lower materials cost for plasma, which has a structure more similar to the low-cost CRTs. The LCD device structure is more like a semiconductor than a CRT, so LCD TVs will continue to have higher manufacturing costs than plasma TVs. Price comparable plasma and LCD TVs today and you can see this difference. Making a larger LCD plant will not solve the major manufacturing cost problem for LCDs, which is the cost for materials.

Larry F. Weber

IEEE Fellow

New Paltz, N.Y.

Metcalfe’s Law

“Lots of luck” rightly describes my reaction to Robert Metcalfe's blog post in response to the “Metcalfe’s Law Is Wrong” [July]. The graceful and right thing to do would be to admit that the original equation cannot apply to larger and larger networks and that the proposed equation is better suited for that.

Partha Kaushik

The debate about Metcalfe’s Law is an important one, but the question of whether it is right or wrong is too simplistic. Are Newton’s laws of motion right or wrong? We now know that these laws are not applicable in some circumstances; nevertheless, they have superb predictive and explanatory power in many practical situations.

Instead, we should focus less on trying to find one law that applies to all networks for all sizes and at all times and more on understanding the different regimes of behavior (and points of inflection when regimes change) that networks may exhibit. For example, the value of a network and of the connections it enables are surely also a function of time as well as of size. Metcalfe himself suggests this in his response to the criticism of his law, although he relates time dependence to the costs of networks, not to the impact of their use.

A connection that may have very low value at one time may grow in importance as it is discovered and exploited, in much the same way as neural connections in the brain can be rewired. Other connections may decrease in value, just as some neural connections may be pruned over time. As a result, the total value of all the connections may change, even if the size of the network remains constant. Furthermore, as a network grows, perhaps (as Metcalfe also surmises) its value stops increasing once it has reached a certain size—or it even starts to decrease, thereby demonstrating diseconomies of scale. As has been said, “If something cannot go on forever, it will stop” (Herb Stein, circa 1980).

On the question of network costs, which Metcalfe links to Moore’s Law, note that the costs of connectivity are, perhaps unfortunately, not driven only by the costs and performance of semiconductors. Civil engineering costs, executive pay, and other phenomena, which contribute significantly to total network costs, are either flat or increasing. Nevertheless, the power of networks is increasing substantially. The value associated with this power does not seem to grow smoothly, but rather in fits and spurts, as network capabilities reach and exceed particular thresholds. For example, the increased value of first-generation 3G mobile networks over 2G at first was marginal at best, as was ISDN over dial-up. But once real speeds in several hundreds of kilobits per second can be achieved (as with cable modems and xDSL in fixed networks and emerging wireless HSDPA and EV-DO networks), then a new set of broadband services becomes economically feasible and operationally practical. So we should really be looking for points of discontinuity or thresholds as for continuous functions of value.

It is extremely unlikely that any one formula can be applied to all networks of all sizes at all times. It will be more fruitful to understand how to categorize networks so as to be able to apply different formulas to their valuation, depending on the category they are in, and to delineate the boundaries between these different categories, along dimensions such as scale and maturity.

In financial valuations, another scientific principle states that the act of observation changes what is being observed. The analogy in financial valuation is that once networks (or more generally businesses) start being valued according to specific metrics, these valuations will be affected as the people in these businesses start to “game” the metrics and thereby alter the expectations and hence the findings of the financial appraisers.

As for Metcalfe’s Law itself, its lasting value lies in the ways in which it has encouraged us to try to understand the value of networks, even if other approximations to the truth turn out to be better as we learn more (the value of Newton’s work was not diminished by Einstein’s theory of relativity). Metcalf’s Law may even be a reasonable approximation to the truth for networks of certain sizes at certain times. But as with any tool with far-reaching implications, it is dangerous when it is applied without understanding the scope of its applicability.

Martyn Roetter

IEEE Member

Boston

My guess is that the value of a network as the number of users grows should follow a sigmoid curve, the usual curve for describing the value of technological innovations, chemical reaction rates, economic development, and the like. Growth in value starts slowly, rises rapidly, and then reaches an inflection point, after which each additional user adds little. If you have 6 billion people and of those 100 are networked, the value is low, and going from 100 to 101 gets you only a small increase in value. If you have 10 or 100 million networked, the value per addition gets higher, because the network is starting to cover a useful section of the population. As most people become connected, it becomes less and less valuable for each user to do so.

The shape of the curve is obviously influenced by the combination of the various subcurves of the various subsets of communicating users. Except for a few spammers, very few people will ever send e-mails to more than a few hundred people. Cellphone usage plans that encourage friends and family imply that there are statistics to back the idea that most people communicate in a narrow circle, perhaps following a power law as they contact parties in broader and broader circles less and less frequently. For example, a person might frequently contact his or her workgroup but rarely contact the entire company or, even more rarely, everyone in the industry. The technology reaches a peak value when enough users have signed up for it.

People tend to forget the sigmoid curve, especially during the steeper portions of the ascent, and they can understand linearity and exponential growth but do not recognize that subsequent slowing and saturation are part of the pattern. Consider the predictions for the growth of such nations as China or India, which are based on the exploitation of their extremely limited urban and educated classes. Too few in those classes or industries can grow at all. But with just enough growth in some areas of these classes, industries can grow around them. If you keep trying to grow these industries, however, eventually you have to wait for the educational establishment to be restructured, along with the requisite restructuring of both cities and the countryside.

The difficult part of assessing technology is locating the spot where you are on the curve. Another example: we heard about the home computer all though the 1980s, but it wasn’t until the 1990s that it was widely adopted. Still, recognizing the existence of the sigmoid curve can provide us with some insight and give us clues as to what signs and portents might be relevant.

Seth Steinberg

Metcalfe’s Law states that the value of a network grows as the square of the number of connection points to the network. This is usually stated as the number of users where each user is assumed to have their own connection point. The law was popularized in 1993 by George Gilder, writer of “The Telecosm,” in the Gilder Technology Report, and chief cheerleader of the telecom/Internet revolution/bubble. Authors Briscoe, Odlyzko, and Tilly argue in the Spectrum article that the actual growth in value is n*log(n) and that the original formulation was bad because it directly led to the speculative excess of the telecom/Internet bubble.

Put baldly, Metcalfe’s Law says that if I have a network and you have a network, and we connect our networks together, they are worth much more than either network on its own or even the sum of the two networks. The more networks we connect, the more valuable the whole thing becomes. So the point of Metcalfe's Law is that there is a huge incentive for all networks to join together into one completely connected internetwork. This has come to pass, first for telephones and then for computers. Thus my position is that Metcalfe has been proven correct and that it is academic to argue whether the “value” (whatever that means) of the network grows quadratically or exponentially.

We need to understand the context when looking at Metcalfe and Gilder’s arguments. As Bob Metcalfe says in his blog entry, in 1980 when he devised Metcalfe's Law, he was just trying to sell the value of networks and create business for his company 3COM. This was at a time when an Ethernet card cost US $5000 and flinty-eyed accountants would argue to reduce the size of their network buy, while he would argue that they should increase it.

George Gilder is the person who foresaw a single interconnected Internet at a time when there was CompuServe, Prodigy, AOL, and thousands of local bulletin board systems. All of these were swept away by the Internet revolution except for AOL, which managed to ride the wave by co-opting it. So Gilder was correct as well, although he was eventually carried away by the force of his own argument like many who listened to him.

I posted this comment to my blog (http://bandb.blogspot.com/2006/11/metcalfes-law.html) on 5 November.

Richard Taylor

IEEE Member

Asking whether Metcalfe’s Law is right or wrong is like asking whether to “reach out and touch someone” is right or wrong. A successful marketing slogan is a promise, not a verifiable empirical statement. Both Metcalfe’s and Moore’s laws prove this: engineers—or more generally, entrepreneurs—are the best marketers.

The IEEE Annals of the History of Computing recently conducted an empirical examination of Moore’s Law. Surveying the historical record on the determinants of the law, Ethan Mollick shows how the law was adjusted to fit changing economic conditions, and specifically, the increase of foreign competition. These observations served as a rallying cry for a nascent industry and an ingenious method for addressing economic justifications that were “needed to continue technical advances.” Moore told Mollick that his original paper/prediction was “an attempt to show the cheapest way to produce microchips.”

Similarly, Metcalfe tells us that his is a “vision thing.” It helped him jump over a big hurdle: the first Ethernet card Metcalfe sold went for $5000 in 1980. He says he used the law to “convince early Ethernet adopters to try local area networks large enough to exhibit network effects,” in effect promising them that the value of their investment would grow as more people connected to the office network. Metcalfe’s Law encapsulates a brilliant marketing concept, engineered to get early adopters—and more important, their accountants—over the difficulty of calculating the return on investment for a new, expensive, unproven technology. It provided the ultimate promise: this technology gets more “valuable” the more you invest in it.

Moore and Metcalfe have done the world a great service by providing “scientific” justification for investing in new technologies. As Briscoe, Odlyzko, and Tilly argue, during the late 1990s Metcalfe’s Law—and I would argue, also Moore’s Law—were used as a justification for a “mad rush for growth.” But the excesses of the late 1990s were driven more by the old idea of the “New Economy" than by these laws. As Anthony B. Perkins and Michael C. Perkins, authors of The Internet Bubble, reminded us during the mad rush, and as Bernard Baruch said in 1932: “In the lamentable era of the ‘New Economics’ culminating in 1929... if we had all continuously repeated ‘two and two still make four,’ much of the evil might have been averted.” But regardless of the main culprit, the Internet bubble produced the backlash of “IT doesn’t matter,” which is more dangerous" (Briscoe’s term) or “evil” (Baruch’s) than the blind belief in the power of technology to improve our lives.

Blind faith supported by scientific justifications has powered the great, mostly U.S.-based, engine of innovation and growth for two centuries. As we enter a new era in which we will see this engine of growth distributed around the globe, I would like to offer a new law, the Marketing Law, to the future movers and shakers worldwide: the success of your idea will be proportional to the square root of the number of people repeating after you “two plus two equal five.”

Gil Press

IEEE Associate Member

Hopkinton, Mass.