Here at KSU we take a hybrid approach to public labs. Probably because of some kind donations by Sun Microsystems. I'd say its about half MS and half miscellaneous (Solaris, Debian, minx..) on the desktops, but most if not all of our mainframes and servers are UNIX based.
Its important to note that the solaris labs are treated like lepper colonies. Sure its nice for me when I need to do something real quick because I know where theres at least 12 open computers guarenteed, but that goes out the window if I need to print, watch a tegrity lecture (online lecture recording and viewing system that uses Java + MS extensions for video) or do something in Rose. Honestly, I don't blame them. The Sun workstations are pretty damn old SPARC 5's. Plus they don't run trillian and solitaire (little do the grad students know/care that gaim is installed on the central server).
But in general I think the mix approach is best, allthough if they're not gonna make MINIX boostrap then they outta trash it and put in something usable. A lot of the windows boxes are just faster, and we have a nice win32 based X windows site liscence. You can learn about UNIX without having a UNIX desktop, believe it or not.
The one thing you really have to realize here is that your father's Computer Science is not your Computer Science. In the 70s most undergraduate programs were closer to Accounting With Computers; after all, the biggest company out there was International Business Machines, and most the people hiring were either state and federal departments or insurance companies. Courses like office automation (I believe my department still has a course like that on the books, albeit grad level). If you want to see a glimpse of what CS professors were thinking coming out of the 70s, take a look at "Computer education in the 1980s, a somber view" by William Mitchell published in SIGCSE in 1980. Not only does he predict the teachers will not be prepared for the microcomputer when it hits their classrooms (which was probably an easy prediction) but also proposes that teaching must become an emphasis in order to preserve the field. He appears to have been wrong on that count but I can't say for how much longer. With salaries dropping for new CS graduates I can see quite a few students deciding that as interesting as that they'd better get something different. I don't blame them.
Fortunately I can say that CS is no longer a IBM seminar, in part thanks to acreditation, another part due to the maturiation of the field.
Molyneux wants to make big games with a huge budget, but is finding that its not that profitable. Rather than find a more appropriate design, the UK government should fund the development? Oh yea, DOA Extreme Vollyball was definately a pinnacle of modern expression, and a clear example of why they should fund this new expansion of the arts. Seriously people, there's another group of developers in the UK thats not only smaller than Molyneux but also successfuly. Introversion has released a successful game to the PC market without support of a publisher for distrobution and marketing. Their game Uplink is a nice combination of gameplay that doesn't require a 30+ minute investment of time but still lets you make progress in the game.
The companies Molyneux cited all suffered from liscencing exposure. Crawfish made several quality GBA games. Unfortunately they were all ports. The bad deal with ports is while the liscence does come with an instant fanbase, it raises the standards for your first game (can't put off features like multiplayer in a street fighter game, even though you haven't figured a way around the slow bus speed yet), and even if you do pull it off to critical acclaim, the liscener will probably do it in house instead, ala Goldeneye/Rare.
On the other hand, I'd love to see cheaper games, and if the UK is willing to foot the bill on it, I'm game!
My understanding is that 64bit mostly slows things down because 64bit memory addressing requires more logic than 32bit addressing. Doubling the bits won't make the resolution draw faster or make the clock run faster. The main advantages appear to be longer word lengths and larger RAM access (32bit is limited to 4GB). I don't think that the limiting factor in your decription is nessecarily bitness, but CPU throughput.
Atari Jaguar, "Do The Math." 32 + 32 = 64. Gotta love marketing people and journalists who push meaningless numbers. For the record, the Jaguar wasn't 64bit in the same manner the Opteron is.
Well the n64 certainly didn't have 64bit graphics. Basically the n64 had a lot of 64 bit wide registers, because there wasn't even enough RAM to max out 32bit addressing. 64 bit graphics... wtf were you thinking, "moron"?
There's just too much information to be displayed. Its nice for showing things like how the directory structure evolved within the kernel, or how quickly dependencies grew but you can't tell one file from another, and the contents are far more important than the directory structure. In short, nothing can feasibly replace a rejection with a short explaination and request for resubmission.
Its important to remember that addressing concerns of interactivity during high loads actually reduces performance. The situation being addressed is not several interactive tasks on a slow system (for which I assume that there's no real solution other than faster hardware). The tesing conditions are XMMS or xine while building something with make.
Another fact to consider is that linux is a multiuser OS. You probably don't want your research compiles slowed down because some jackass is running a hacked up mozilla over remote X. Or consider two computational physics research projects running on the same department mainframe.
There are other reasons against static assignment of interactivity boosts, like degraded overall performance, imperfect information, and an innacturate definition of interactivity.
Charging on outgoing packets is an idea with some interesting side effects. If you start charging per byte, then all of a sudden theres an increased incentive by the ISP to encourage use rather than discourage. Of course there's also a new incentive to keep it legal, as the ISPs then profit off of piracy networks. So what kind of data would both be legal (at least at the federal level) and bandwidth consuming?
Perhaps a glance at another real world communication network where the burden lies on the sender: the US Postal Office. Good ol' USPS makes a good deal of revenue off their "bulk rate" which is basically junk mail. In fact, without junk mail the post office probably couldn't turn a profit.
If a sender burden internet was the norm, its often argued that spam mail would trickle to a halt as the burden would easily tip the scale of profitibility. I argue that rates would not be set so prohibitively, although I do not have the time to research the cost benefit weights.
I've read the law, and I don't see what you're saying. In fact, if I recall correctly, the amount of distrobution and the profit made both factor into punishment, but a zero revenue pirate is still guilty, and may be subject to criminal law.
Its not impossible to design a game cd that can handle bootCD'ing or installing. All thats desperately needed is a quality NTFS linux system (for swap and maybe persistant data like saves).
The real question is who wants to be bothered by booting. On one hand you have the "hardcore" gamers who essentially turn their multitasking OS off and focus exclusively on the game. On the other hand you have what I call "solitaire" players who simply want a short break. Neither want to be bothered by a time intensive reboot, to simply reboot back when they're done. But your hardcore gamer might go for it if it results in extra performance on their computationally challenging game. Of course, that isn't the case, a custom linux compile would probably benefit over a mega module kernel, not to mention the actual software.
About the only thing I can think of to motivate a bootCD game is a hardware independent and hassle free demo, which ironically enough, won't happen without an NTFS system (for those windows systems without FAT or more 3d card support (ATi is pathetic and nv apparenlty doesn't allow free redistrobution).
Maybe you didn't play enough of that GT3. My roommate and brother both have it so we've tinkered around with it. Looking at that list its like you didnt even play the goddamned game. Every single one of those things mentioned is a factor, but you're gonna have to spend some in game money for parts. You can adjust camber angle, toe angle, gear ratios, downforce angles (if the car has them), tire type, etc.
The game does start off very arcady in that you don't need to (I don't think you really can) modify the car to race. But once you get out of the amateur division, its gonna take some fine tuning (or maybe just an escudo). It starts displaying tire wear when you run into 5 lap races, complete with tire heat and grip curve modelling.
The real missing attribute here is damage. But a hardcore racer player like yourself should know that manufacturers are reluctant to let their shiny car models indicate a dent or scratch. Thats why most of your fancy sim games in the list are either custom car, fictional car or unliscenced.
About that... it seems that people want real cars over damage models. For some crazy reason, car manufacturers want games to give the illusion that you can't wreck their cars when driving at insane speeds. So its mostly been an either or situation with damage and liscenced cars. Its slightly changing and pretty much obvious and easy to implement, but look at all the games with damage modelling. Then look at all the games with real cars. You should then notice that all the good sellers were liscenced not damage based. Maybe its just that people don't want to worry about wrecking their car, but I think that if you could combine the two it would do very well. I mean, fzero and wipeout both have damage and sell well. (Speaking of fzero, have you seen the latest videos? holy heat effects!).
I thought about mentioning freenet but decided against it, since the system is insanely complex and in general not for making files available to everyone. It does achieve a great deal of security in these transactions but in my experience, actually using it for filesharing doesn't work well.
Should every citizen be expected to have the knowledge of a medical degree, mechanic and engineer? Hopefully not.
Of course there should be reasonable limits on liscencure, and your examples highlight the system's abuse. But what are some good guidelines for what gets liscenced and what doesn't?
I suggest: 1) The government itself has an interest in hiring qualified individuals. 2) A likelyhood of making poor long term choices, like prescribing antibiotics left and right. 3) Mistakes result in high societal costs, and these mistakes have commonly known solutions. In other words, there is a high correlation between passing a test and not making high risk mistakes.
There is another concern, however. We don't want to impose too great a cost on the economy through liscensure, so perhaps instead of liscence to practice we want liscence to practice with the government. This often occurs in Engineering fields, though some employers look for liscensure as a means of certification. What kinds of criteria should make a liscence manditory for practice? Perhaps:
1) The work is unreversable; once done cannot be undone
Of course these are just suggestions, and as such require both further inspection from myself and from others. We must be careful, we're not looking to justify the status quo, but to develop a new status quo.
Government liscencing programs are done to ensure the quality of technical government employees, despite your anecdotal complaints. When Kansas builds a highway, someone needs to design it, which includes all the overpasses. And in proper Confuscian manner, these liscenced are conferred by tests are written and standardized, not a general seal of approval by your local Doctor Approver General. These tests are usually designed by a professional organization in conjunction with the government. Its like a government approved certification, and prevents Governor BillyBob from appointing his idiot son to design the new capitol (which would need be evacuated when discovered that the dome is prone to collapsing).
Outside of the medical field, liscencure is only required by the government if you want to work for the government. Some employers often desire liscenced engineers, even though they don't contract to the government.
As far as I can tell, the real reason doctors are expensive is good old Malpractice Insurance. Insurance is expensive, but losing your practice with 20k left to go on your school debt is more so. The insurance company knows its shit. They reward the most court defendable medical processes with rate reductions. Of course, this often means more doctor involvement (which drives up costs), but your practice really really fears malpractice, far more so than the cost of hiring another doctor.
Our Government is errected to establish the peace needed for prosperity, not to oppress and tax. If you have any specific problems with government liscensure that aren't paranoid ramblings about the Man, please let us hear them. These things are done on a State to State basis, though many times there is a common theme.
For those who haven't standardized on MS tools, at least (and probably many who have). A friend of mine who works for the central computing services has designed a system of filesharing blocking that unfortunately works far too well. He's made it into his employer's newsletter and then into the actual campus newspaper.
Why? Because before, filesharing was filling the Internet pipe. It was awful. People sharing movies across a whole 20Mbps meant instant plateau the minute the university woke up at 8am every day. The solution was to simply cut off napster (this particular one managed to solve itself, as did eMesh). Once again I can browse the internet without insane response times. And the university doesn't have to pay for emergency bandwidth type things. We've since doubled the pipe size at no extra cost as a result of contract renegotiations, but I don't think its possible to know how much would be used in absence of filters or pipe limitations.
The only downside to this is a chilling effect. By shutting down access to file sharing programs, you cut access to all files reguardless of copyright and distrobution desires. Wanted to give your cool guitar solo piece to the world? Well you better have a nice hosting service set up when word gets around that it doesn't suck, because there's no avoiding flashcrowds anywhere but p2p.
The answer is something like encrypted gnutella. Then the man from oofle cannot filter on ip because there would be no central location, cannot filter on port because we're abusing a standard port and cannot filter by signature because the packets have been obscured at worst, encrypted at best.
The difference being that a significant amount of work was done in acadamia on a method that is not "patently" obvious. But way to conjure a straw man against Microsoft, though. The only demonstrable example of a Microsoft patent that I can remember is when they patented a version of CSS while attending a session on standardizing CSS. It doesn't matter who it is, its the action that sucks. Always will be. And until Microsoft stops being the man of egregious action, I don't think/. is going to stop posting about it.
I realize that VA has the editors slaving over a new backend system designed to improve connectedness with bullshit like journals and "friend lists" but I took the time to read that article and you know what? I recognized this and sure enough, its from 1995. I think I saw this one back on slashdot a long long time ago, but even then it was probably a dupe.
Withdrawl of troops from their holy cities perhaps? Bin Laden's catalyst to American hatred seems to be military occupation of Mecca and Medina. Something about driving them from the moeques I believe. But its not like holy scripture can't be twisted to serve worldly means.
Its not my fault linux runs on 2 out of 3 of the major consoles commercially available. But thanks for the time tip, I'll be sure to use it in the future.
Out of the six billion people on the planet, only 3 percent can afford one. Of those that can afford one, half decide they actually want one. Combine that half with the lonely few in cyber cafes and markets and you have the world's top spenders in one place, perfect for advertisers.
Slashdot Mirror
Here at KSU we take a hybrid approach to public labs. Probably because of some kind donations by Sun Microsystems. I'd say its about half MS and half miscellaneous (Solaris, Debian, minx..) on the desktops, but most if not all of our mainframes and servers are UNIX based.
Its important to note that the solaris labs are treated like lepper colonies. Sure its nice for me when I need to do something real quick because I know where theres at least 12 open computers guarenteed, but that goes out the window if I need to print, watch a tegrity lecture (online lecture recording and viewing system that uses Java + MS extensions for video) or do something in Rose. Honestly, I don't blame them. The Sun workstations are pretty damn old SPARC 5's. Plus they don't run trillian and solitaire (little do the grad students know/care that gaim is installed on the central server).
But in general I think the mix approach is best, allthough if they're not gonna make MINIX boostrap then they outta trash it and put in something usable. A lot of the windows boxes are just faster, and we have a nice win32 based X windows site liscence. You can learn about UNIX without having a UNIX desktop, believe it or not.
The one thing you really have to realize here is that your father's Computer Science is not your Computer Science. In the 70s most undergraduate programs were closer to Accounting With Computers; after all, the biggest company out there was International Business Machines, and most the people hiring were either state and federal departments or insurance companies. Courses like office automation (I believe my department still has a course like that on the books, albeit grad level). If you want to see a glimpse of what CS professors were thinking coming out of the 70s, take a look at "Computer education in the 1980s, a somber view" by William Mitchell published in SIGCSE in 1980. Not only does he predict the teachers will not be prepared for the microcomputer when it hits their classrooms (which was probably an easy prediction) but also proposes that teaching must become an emphasis in order to preserve the field. He appears to have been wrong on that count but I can't say for how much longer. With salaries dropping for new CS graduates I can see quite a few students deciding that as interesting as that they'd better get something different. I don't blame them. Fortunately I can say that CS is no longer a IBM seminar, in part thanks to acreditation, another part due to the maturiation of the field.
Molyneux wants to make big games with a huge budget, but is finding that its not that profitable. Rather than find a more appropriate design, the UK government should fund the development? Oh yea, DOA Extreme Vollyball was definately a pinnacle of modern expression, and a clear example of why they should fund this new expansion of the arts. Seriously people, there's another group of developers in the UK thats not only smaller than Molyneux but also successfuly. Introversion has released a successful game to the PC market without support of a publisher for distrobution and marketing. Their game Uplink is a nice combination of gameplay that doesn't require a 30+ minute investment of time but still lets you make progress in the game.
The companies Molyneux cited all suffered from liscencing exposure. Crawfish made several quality GBA games. Unfortunately they were all ports. The bad deal with ports is while the liscence does come with an instant fanbase, it raises the standards for your first game (can't put off features like multiplayer in a street fighter game, even though you haven't figured a way around the slow bus speed yet), and even if you do pull it off to critical acclaim, the liscener will probably do it in house instead, ala Goldeneye/Rare.
On the other hand, I'd love to see cheaper games, and if the UK is willing to foot the bill on it, I'm game!
Yea, except you gotta actually have games for it to sell, and those come from developers, not just the manufacturing plant.
My understanding is that 64bit mostly slows things down because 64bit memory addressing requires more logic than 32bit addressing. Doubling the bits won't make the resolution draw faster or make the clock run faster. The main advantages appear to be longer word lengths and larger RAM access (32bit is limited to 4GB). I don't think that the limiting factor in your decription is nessecarily bitness, but CPU throughput.
Atari Jaguar, "Do The Math." 32 + 32 = 64. Gotta love marketing people and journalists who push meaningless numbers. For the record, the Jaguar wasn't 64bit in the same manner the Opteron is.
Well the n64 certainly didn't have 64bit graphics. Basically the n64 had a lot of 64 bit wide registers, because there wasn't even enough RAM to max out 32bit addressing. 64 bit graphics... wtf were you thinking, "moron"?
There's just too much information to be displayed. Its nice for showing things like how the directory structure evolved within the kernel, or how quickly dependencies grew but you can't tell one file from another, and the contents are far more important than the directory structure. In short, nothing can feasibly replace a rejection with a short explaination and request for resubmission.
Its important to remember that addressing concerns of interactivity during high loads actually reduces performance. The situation being addressed is not several interactive tasks on a slow system (for which I assume that there's no real solution other than faster hardware). The tesing conditions are XMMS or xine while building something with make.
Another fact to consider is that linux is a multiuser OS. You probably don't want your research compiles slowed down because some jackass is running a hacked up mozilla over remote X. Or consider two computational physics research projects running on the same department mainframe.
There are other reasons against static assignment of interactivity boosts, like degraded overall performance, imperfect information, and an innacturate definition of interactivity.
Charging on outgoing packets is an idea with some interesting side effects. If you start charging per byte, then all of a sudden theres an increased incentive by the ISP to encourage use rather than discourage. Of course there's also a new incentive to keep it legal, as the ISPs then profit off of piracy networks. So what kind of data would both be legal (at least at the federal level) and bandwidth consuming?
Perhaps a glance at another real world communication network where the burden lies on the sender: the US Postal Office. Good ol' USPS makes a good deal of revenue off their "bulk rate" which is basically junk mail. In fact, without junk mail the post office probably couldn't turn a profit.
If a sender burden internet was the norm, its often argued that spam mail would trickle to a halt as the burden would easily tip the scale of profitibility. I argue that rates would not be set so prohibitively, although I do not have the time to research the cost benefit weights.
I've read the law, and I don't see what you're saying. In fact, if I recall correctly, the amount of distrobution and the profit made both factor into punishment, but a zero revenue pirate is still guilty, and may be subject to criminal law.
Its not impossible to design a game cd that can handle bootCD'ing or installing. All thats desperately needed is a quality NTFS linux system (for swap and maybe persistant data like saves).
The real question is who wants to be bothered by booting. On one hand you have the "hardcore" gamers who essentially turn their multitasking OS off and focus exclusively on the game. On the other hand you have what I call "solitaire" players who simply want a short break. Neither want to be bothered by a time intensive reboot, to simply reboot back when they're done. But your hardcore gamer might go for it if it results in extra performance on their computationally challenging game. Of course, that isn't the case, a custom linux compile would probably benefit over a mega module kernel, not to mention the actual software.
About the only thing I can think of to motivate a bootCD game is a hardware independent and hassle free demo, which ironically enough, won't happen without an NTFS system (for those windows systems without FAT or more 3d card support (ATi is pathetic and nv apparenlty doesn't allow free redistrobution).
You know, you don't need objects to get polymorphism. And in statically checked languages like ML you get to know what the function is returning.
Maybe you didn't play enough of that GT3. My roommate and brother both have it so we've tinkered around with it. Looking at that list its like you didnt even play the goddamned game. Every single one of those things mentioned is a factor, but you're gonna have to spend some in game money for parts. You can adjust camber angle, toe angle, gear ratios, downforce angles (if the car has them), tire type, etc.
The game does start off very arcady in that you don't need to (I don't think you really can) modify the car to race. But once you get out of the amateur division, its gonna take some fine tuning (or maybe just an escudo). It starts displaying tire wear when you run into 5 lap races, complete with tire heat and grip curve modelling.
The real missing attribute here is damage. But a hardcore racer player like yourself should know that manufacturers are reluctant to let their shiny car models indicate a dent or scratch. Thats why most of your fancy sim games in the list are either custom car, fictional car or unliscenced.
And of course, its Gran Tourismo, not Grand.
About that... it seems that people want real cars over damage models. For some crazy reason, car manufacturers want games to give the illusion that you can't wreck their cars when driving at insane speeds. So its mostly been an either or situation with damage and liscenced cars. Its slightly changing and pretty much obvious and easy to implement, but look at all the games with damage modelling. Then look at all the games with real cars. You should then notice that all the good sellers were liscenced not damage based. Maybe its just that people don't want to worry about wrecking their car, but I think that if you could combine the two it would do very well. I mean, fzero and wipeout both have damage and sell well. (Speaking of fzero, have you seen the latest videos? holy heat effects!).
Come up with an interesting software idea, design the software and sell it. Preferrably to someone with a lot of money like MS.
I thought about mentioning freenet but decided against it, since the system is insanely complex and in general not for making files available to everyone. It does achieve a great deal of security in these transactions but in my experience, actually using it for filesharing doesn't work well.
Should every citizen be expected to have the knowledge of a medical degree, mechanic and engineer? Hopefully not.
Of course there should be reasonable limits on liscencure, and your examples highlight the system's abuse. But what are some good guidelines for what gets liscenced and what doesn't?
I suggest:
1) The government itself has an interest in hiring qualified individuals.
2) A likelyhood of making poor long term choices, like prescribing antibiotics left and right.
3) Mistakes result in high societal costs, and these mistakes have commonly known solutions. In other words, there is a high correlation between passing a test and not making high risk mistakes.
There is another concern, however. We don't want to impose too great a cost on the economy through liscensure, so perhaps instead of liscence to practice we want liscence to practice with the government. This often occurs in Engineering fields, though some employers look for liscensure as a means of certification. What kinds of criteria should make a liscence manditory for practice? Perhaps:
1) The work is unreversable; once done cannot be undone
Of course these are just suggestions, and as such require both further inspection from myself and from others. We must be careful, we're not looking to justify the status quo, but to develop a new status quo.
Government liscencing programs are done to ensure the quality of technical government employees, despite your anecdotal complaints. When Kansas builds a highway, someone needs to design it, which includes all the overpasses. And in proper Confuscian manner, these liscenced are conferred by tests are written and standardized, not a general seal of approval by your local Doctor Approver General. These tests are usually designed by a professional organization in conjunction with the government. Its like a government approved certification, and prevents Governor BillyBob from appointing his idiot son to design the new capitol (which would need be evacuated when discovered that the dome is prone to collapsing).
Outside of the medical field, liscencure is only required by the government if you want to work for the government. Some employers often desire liscenced engineers, even though they don't contract to the government.
As far as I can tell, the real reason doctors are expensive is good old Malpractice Insurance. Insurance is expensive, but losing your practice with 20k left to go on your school debt is more so. The insurance company knows its shit. They reward the most court defendable medical processes with rate reductions. Of course, this often means more doctor involvement (which drives up costs), but your practice really really fears malpractice, far more so than the cost of hiring another doctor.
Our Government is errected to establish the peace needed for prosperity, not to oppress and tax. If you have any specific problems with government liscensure that aren't paranoid ramblings about the Man, please let us hear them. These things are done on a State to State basis, though many times there is a common theme.
Why? Because before, filesharing was filling the Internet pipe. It was awful. People sharing movies across a whole 20Mbps meant instant plateau the minute the university woke up at 8am every day. The solution was to simply cut off napster (this particular one managed to solve itself, as did eMesh). Once again I can browse the internet without insane response times. And the university doesn't have to pay for emergency bandwidth type things. We've since doubled the pipe size at no extra cost as a result of contract renegotiations, but I don't think its possible to know how much would be used in absence of filters or pipe limitations.
The only downside to this is a chilling effect. By shutting down access to file sharing programs, you cut access to all files reguardless of copyright and distrobution desires. Wanted to give your cool guitar solo piece to the world? Well you better have a nice hosting service set up when word gets around that it doesn't suck, because there's no avoiding flashcrowds anywhere but p2p.
The answer is something like encrypted gnutella. Then the man from oofle cannot filter on ip because there would be no central location, cannot filter on port because we're abusing a standard port and cannot filter by signature because the packets have been obscured at worst, encrypted at best.
The difference being that a significant amount of work was done in acadamia on a method that is not "patently" obvious. But way to conjure a straw man against Microsoft, though. The only demonstrable example of a Microsoft patent that I can remember is when they patented a version of CSS while attending a session on standardizing CSS. It doesn't matter who it is, its the action that sucks. Always will be. And until Microsoft stops being the man of egregious action, I don't think /. is going to stop posting about it.
I realize that VA has the editors slaving over a new backend system designed to improve connectedness with bullshit like journals and "friend lists" but I took the time to read that article and you know what? I recognized this and sure enough, its from 1995. I think I saw this one back on slashdot a long long time ago, but even then it was probably a dupe.
Withdrawl of troops from their holy cities perhaps? Bin Laden's catalyst to American hatred seems to be military occupation of Mecca and Medina. Something about driving them from the moeques I believe. But its not like holy scripture can't be twisted to serve worldly means.
Its not my fault linux runs on 2 out of 3 of the major consoles commercially available. But thanks for the time tip, I'll be sure to use it in the future.
Out of the six billion people on the planet, only 3 percent can afford one. Of those that can afford one, half decide they actually want one. Combine that half with the lonely few in cyber cafes and markets and you have the world's top spenders in one place, perfect for advertisers.