In this case, the politics are the indirect reason. I think the writers (which were all part of the "FSF" cadre) are simply more knowledgeable about the GPL-style licensing than about BSD-style licensing. They have considerable experience and expertise drafting and enforcing the GPL and LGPL, but they have not been, for example, part of the AT&T-Berkeley BSD litigation. The discussion of copyright enforcement would have been more interesting if it included examples of successful enforcement of both the types of licenses.
The discussion of "copyright assignment" falls under the same heading: that's what the FSF does, and they are telling us what they learned about this kind of setup. Better they share their knowledge than leaving others to rediscover it.
I think the best way to view this document is as follows: "we are laywers, who have been helping FOSS projects. Here are some lessons we learned, in non-technical language. Read this before talking to your lawyer and you'll get more mileage out of him/her".
First, giving the FCC more discretionary authority is not a good thing to do. They are very receptive to lobbying (broadcast flag, mandatory DRM...) and industry corruption (employees that leave directly to cushy jobs in the industry they were supposedly regulating just recently). Secondly, I'm not sure where the Federal interest is in regulating businesses -- that the internet as a whole is international?
This is really a contract issue. If their TOS promise "unlimited bandwidth" then they should provide that. If the TOS say "we connect you to the internet" they should not be able to block random ports. And sending fake packets is already a computer crime (at least, if I sent fake packets to Comcast servers I would probably be charged with attempted DOS or something). So I would support a "contact terms mean what they mean" law -- not giving the FCC more discretion to help the industry to screw the customers.
MediaSentry claims to have verified that the names of the files they downloaded contained descriptions of music copyrighted by the plaintiffs. That they didn't bother to play the music and check the contents is very suspicious. It's true that the RIAA has been seeding fake torrents with damaged files and that filenames don't necessarily describe the content, so it's conceivable that the files didn't really contain what the names suggest, but I doubt that and I don't think you could sell this to a jury working at the "preponderance of the evidence" standard (I'm not even sure about "beyond reasonable doubt"). That said, it shows extreme incompetence on the part of MediaSentry. In my opinion this should be used to impeach their credibility: "if you didn't take this elementary step, what other important steps did you gloss over?" -- suggesting it's possible the files weren't song files is to "conspiracy theory"-like to work (the Thompson case reeked of such attempts).
Scantron voting has its own problems. What about voters who don't fill the bubbles accurately enough? Who fill two bubbles? What tolerance do you set the scanner to? You need to find a good division of labour between the computer (who is good at some things), the voter and the elections officials (which are human and good at other things).
Just like computers are better than people at scanning the ballots and tabulating the data, they are also better than people at filling the ovals (also known as "printing a filled ballot"). You will get a lot less scanning errors and invalid ballots if the voter gives the selections to a computer and then the computer prints the result out. For some voters it may be slower than hand-filling of the ballot, but for some (the disabled, the tech-savvy) it may be much faster. If you want, you can offer people the option: fill your own bubbles or have a computer fill them for you.
Moreover, a well-designed voting system should be 100% accurate in the counting of votes because of, not despite, the removal of humans from the counting process. The problem is that so far, no commercially available electronic voting system exists yet that has been well designed.
I think you are going a bit over the top. There is a trade-off between accuracy, flexibility, and openness here. There's no way to reach 100% accuracy in the counting of tens of millions of ballots, each containing selections for tens of races. If you allow yourself to also take into account the discrepancy between the voter's intent and the voter's markings on the ballot (I'm against it but many aren't) then "100% accuracy" is not even meaningful. What you should strive for is 1. maximum accuracy 2. known error rates. If you knew the error rate of the ballot-generating-and-counting system then you'd know at which point a thorough recount is warranted (assuming it had a lower error rate), and when you simply need to rerun the election (or draw cards).
By the way, it's true that ATMs are more reliable than voting machines, and that the banking system is more reliable than the election system -- and yet even there the system is not 100% reliable. A bank "lost" $20K belonging to a friend of mine through bad record-keeping on their part. It took weeks to get her money back. Once in a while banks will record a transaction wrong -- and each bank has a controlled system that they design and implement. Elections are run in parallel by many independent local authorities under many conflicting criteria and need to be more flexible (do we allos for write-in candidates? for people who are voting provisionally?). Yes, there is an accuracy price to pay for that, but since almost all races are not close, and when the race is close we shouldn't really care who wins in the end, it's not too much of a price to pay.
I also discuss this in another comment, but the problem really arises not when your candidate loses, but when your candidate loses narrowly. This is quite justifiable: the smaller the effect (difference of support between the candidates), the less likely it is that your detection system (election procedures) can measure it correctly.
There are two kinds of voter confidence issues: confidence that they system is free of biases, and confidence that, assuming it was free of biases, the system got the right result. It's true that electronic voting reduces confidence in the first property -- but I think the main driver for lack of voter confidence is their ignorance of the fact that even an unbiased system will get the "wrong" result some of the time. Since we lack an objective measure of the support of the candidates, there is of course no "right" result of the election beyond the actual results, but in the end I think that what happens is that when elections are close voters come face-to-face with what scientists have been facing for centuries under the name like "measurement error" and "scientific significance", they (the voters) tend to ascribe the problem to systematic bias rather than random error. It's true that less transparent systems make it easier for the voters to believe in conspiracy theories, but the underlying problem is lack of scientific thinking skills.
I'd predict that after a close election voters will react the same way regardless of the technology employed (or lack thereof).
In principle, the ballots being counted in public in front of everyone in the village will inspire more confidence than an obscure computer calculation. However, as the 2000 Florida debacle has demonstrated, hand counting has its own problems (e.g. error rates) which the voting public does not understand either. It seems to me that if the system artificially produces a landslide (e.g. via a winner-take-all-state-electoral-votes system), the public is happy that things went well. If the elections are close there is a lot of consternation and misunderstanding.
On the technical level, ballots that are both human- and machine-countable but generated automatically (so there is less room for voter marking errors], look best to me. If the voting machine prints the ballot out but keeps no record otherwise that would be best. But just wait for a close elections and the voters will express lack of confidence in the results. The problem is the following: if you are trying to measure a large effect, then you will get the right result no matter what method you use and everyone will be quite confident you got the right result. If you are trying to measure an effect which is just at the level of resolution for your detector (or worse, as in the Florida case, below the measurement error) then there is no way to be as confident that you got the right result.
Because there's more to a bus than the bandwidth. USB has a lot of overhead (it can be branched, hook many devices etc). SATA is dedicated for controlling storage. That's why we put cameras on the USB, hard-drives on a SATA bus, the network card on the PCI bus, video card on the VESA bus...
If I were RMS, I would forbid the packaging of any GNU code with a GPLv2 GNU/Linux.
Not only would this against RMS's free software philosophy, but the GPL (yes, even version 3!) expressly disclaims any limitation on the mere aggregation of software.
More to the point, this is much ado about nothing. Even if Mr. Torvalds "saw the light" and decided he wanted to move to GPL v3, this would be impossible in practical terms since Linux has no copyright escrow agent similar to the FSF for GNU. In other words, to move code licensed to Linux under GPL v2 (only) to GPL v3 requires re-licensing by the original author -- which you may never be able to find. So, you may safely assume that Linux will be GPL v2 until it is re-written from scratch.
If I were RMS, I would forbid the packaging of any GNU code with a GPLv2 GNU/Linux.
Not only would this against RMS's free software philosophy, but the GPL (yes, even version 3!) expressly disclaims any limitation on the mere packaging of software.
More to the point, this is much ado about nothing. Even if Mr. Torvalds "saw the light" and decided he wanted to move to GPL v3, this would be impossible in practical terms since Linux has no copyright escrow agent similar to the FSF for GNU. In other words, to move code licensed to Linux under GPL v2 (only) to GPL v3 requires re-licensing by the original author -- which you may never be able to find. So, you may safely assume that Linux will be GPL v2 until it is re-written from scratch.
You are right to point at the older story -- we need to make a distinction. The scientific point here is the use of satellite imagery to locate old cities. To social point is that Google Earth has made satellite images infinitely more accessible -- you don't need to be part of NASA anymore.
As you are surely aware, the US does not consider following treaty obligations a serious duty. Rather, when the US determines that a treaty obligation is against national interest, it just ignores it. Besides, given that the infinite copyright terms in the treaties were originally put there at the insistence of the US, the US can easily repudiate its earlier position and ask for everyone to reconsider.
Please take a moment to brush up on your history before you make such un-informed posts. If you join right now they tell you what the current behaviour is: "share" means with all contacts. That's an acceptable behaviour. Nothing wrong with it, execapt that the "sharing" feature is not new, and before it had a different behaviour: the shared feeds appeared on a "secret" URL, and you had control over who could see the feeds. So, if you "shared" feeds before, and didn't notice the new announcement, suddenly a lot of people were told of your share -- people who you may have not wanted to know. That is a serious privacy concern.
Given the abject disasters that Jackson's "The Two Towers" and "The Return of the King" were, is there any reason to expect that he'll do well with the Hobbit, not to speak of a "sequel" which will be mostly original? This is, of course, a matter of opinion, but it seemed to me that the weakest parts of there two films were the segments were Tolkien's material was explicitly rejected in favour of new material. The technical side will surely be amazing, but that is not quite reason enough to see a movie for me.
The lawsuit is Nigerian, but it's not so clearly a scam. It seems to be a claim that a keyboard layout (i.e. which key goes where) is a patentable design. Of course in most of the world keyboard layouts are standardized, denying us the fun of learning a new keyboard layout whenever we buy a new keyboard -- but perhaps this isn't the case there. If anything, I would suspect it to be a harassment tactic. I wonder if this Nigerian company has recently started a strategic partnership with a large American software company...
First of all, the EFF may has not tested your ISP. You may trust them that in general ISPs are sending spoofed packets, but still want to know whether your ISP is using the tactic. Beyond that, however, just because you trust them doesn't mean independent verification has no value. Results mean something different if you obtained them yourself. Also, as in regular science, independent confirmation of results gives more than that: more people conducting tests will also give better data.
I'm not sure why this story was filed under "privacy" rather than "technology". Nobody's freaked out by police helicopters, whether they are used to find traffic offenders, in police chases, or as observation posts for police raids. Using unmanned aircraft instead is a no-brainer. They are cheaper to operate, can stay up longer, and people don't die when they collide (though this incident was with civillian helicopters).
I see absolutely no downside about doing it the traditional way. Is there any reason to do it with machines in america, or you do it that way just because it's cooler?
The US, being an enormous country, has a many levels of government. Unlike many other countries, it runs all elections for all levels of government on a fixed date (some Tuesday in November), rather than spreading them around the year. Of course, not every position is up for election every year, but still this means that the "ballot" contains tens if not hundreds of separate elections, ranging all the way from the US President to the county water board and the town mayor, not to mention multiple "ballot initiative" (direct legislation). Each election (especially president, governor etc) can feature tens of candidates (most of them irrelevant). Printed ballots are thick booklets; both filling them correctly and manually reading them is a non-trivial operation. Also, manually tallying the votes in these hundreds of elections takes a lot of time.
This is not to say that this was not done manually in the past, but certainly using computers greatly simplifies the process. I think the best solution is to use computers to generate the ballot, but only use computer counts provisionally. That is, the voter will step up to a computer and will make selections, after which the computer will print a filled ballot that can be optically scanned. The computer will also tally the votes giving a quick result for most of the races. Nevertheless, the printed ballots should be considered the official votes, the ones to be used if a recount is necessary. In important races (President, Governor) it's probably better to automatically count the printed ballots and only use the computer counts for provisional results. Note that this also allows for people to manaually fill their ballots if they feel like it.
Ah, but he's not apologizing for the lawsuits -- he's apologizing for not releasing DRM-riddled restrictively-licensed music fast enough, which he thinks is what forced consumers to share music illegally. He's still behind the lawsuits (except when his own kids share music -- then it's a "family matter" best punished by the parents). He's warning the cell-phone companies that unless they allow limited sharing, consumers will find their own solutions, and not talking about tactics. The content industry (music, film etc) still seems to have no idea what the consumers want, or that the offering people what they want is usually much better than coercing them to buy what you want them to buy.
It actually goes the other way around: if the user's pictures are trapped on a photo-sharing site that only works on Windows, then the user is forever locked to buying future copies of Windows. Any data you have on MS-specific apps is continuously held hostage to ensure you keep paying the Microsoft tax.
First, given the current regulation scheme (the FDA in the US, for example) the distance between a fundamental discovery and an actual drug on the market is much greater in medicine than it is in technology -- Intel does not require approval from anyone to market their next-gen processor. Second, the current patent system makes making trivial improvements on existing drugs (hence extending monopoly protection) much more profitable than researching new drugs (high risk of failing to produce anything).
But even ignoring all these things, on a fundamental level biology is orders of magnitude more difficult than physics. We understand the physics of seminconductors and the mathematics of computation fairly well. We can simulate future processors ahead of time to see if a new cache design will improve performance or not. We have no idea how to simulate a biological system, and barely have quantitative models for event the simplest ones. Let's give it 100 years and try again.
I'd love to know how well these will sell -- perhaps we can arrange a follow-up in a few months?
More seriously, this falls under the "computers as appliances" paradigm. This "home computing appliance" cost less than many "home gaming" appliances, about as much as a cheap all-in-one sound system. It delivers basic internet functionality. The users shouldn't care what OS it runs anymore than they care what OS their printer runs. This is not to say that there can be a potential snag: users trying to install Microsoft-targeted software off the web. I hope there's a good version of wine installed.
Slashdot Mirror
In this case, the politics are the indirect reason. I think the writers (which were all part of the "FSF" cadre) are simply more knowledgeable about the GPL-style licensing than about BSD-style licensing. They have considerable experience and expertise drafting and enforcing the GPL and LGPL, but they have not been, for example, part of the AT&T-Berkeley BSD litigation. The discussion of copyright enforcement would have been more interesting if it included examples of successful enforcement of both the types of licenses.
The discussion of "copyright assignment" falls under the same heading: that's what the FSF does, and they are telling us what they learned about this kind of setup. Better they share their knowledge than leaving others to rediscover it.
I think the best way to view this document is as follows: "we are laywers, who have been helping FOSS projects. Here are some lessons we learned, in non-technical language. Read this before talking to your lawyer and you'll get more mileage out of him/her".
First, giving the FCC more discretionary authority is not a good thing to do. They are very receptive to lobbying (broadcast flag, mandatory DRM ...) and industry corruption (employees that leave directly to cushy jobs in the industry they were supposedly regulating just recently). Secondly, I'm not sure where the Federal interest is in regulating businesses -- that the internet as a whole is international?
This is really a contract issue. If their TOS promise "unlimited bandwidth" then they should provide that. If the TOS say "we connect you to the internet" they should not be able to block random ports. And sending fake packets is already a computer crime (at least, if I sent fake packets to Comcast servers I would probably be charged with attempted DOS or something). So I would support a "contact terms mean what they mean" law -- not giving the FCC more discretion to help the industry to screw the customers.
MediaSentry claims to have verified that the names of the files they downloaded contained descriptions of music copyrighted by the plaintiffs. That they didn't bother to play the music and check the contents is very suspicious. It's true that the RIAA has been seeding fake torrents with damaged files and that filenames don't necessarily describe the content, so it's conceivable that the files didn't really contain what the names suggest, but I doubt that and I don't think you could sell this to a jury working at the "preponderance of the evidence" standard (I'm not even sure about "beyond reasonable doubt"). That said, it shows extreme incompetence on the part of MediaSentry. In my opinion this should be used to impeach their credibility: "if you didn't take this elementary step, what other important steps did you gloss over?" -- suggesting it's possible the files weren't song files is to "conspiracy theory"-like to work (the Thompson case reeked of such attempts).
Scantron voting has its own problems. What about voters who don't fill the bubbles accurately enough? Who fill two bubbles? What tolerance do you set the scanner to? You need to find a good division of labour between the computer (who is good at some things), the voter and the elections officials (which are human and good at other things).
Just like computers are better than people at scanning the ballots and tabulating the data, they are also better than people at filling the ovals (also known as "printing a filled ballot"). You will get a lot less scanning errors and invalid ballots if the voter gives the selections to a computer and then the computer prints the result out. For some voters it may be slower than hand-filling of the ballot, but for some (the disabled, the tech-savvy) it may be much faster. If you want, you can offer people the option: fill your own bubbles or have a computer fill them for you.
I also discuss this in another comment, but the problem really arises not when your candidate loses, but when your candidate loses narrowly. This is quite justifiable: the smaller the effect (difference of support between the candidates), the less likely it is that your detection system (election procedures) can measure it correctly. There are two kinds of voter confidence issues: confidence that they system is free of biases, and confidence that, assuming it was free of biases, the system got the right result. It's true that electronic voting reduces confidence in the first property -- but I think the main driver for lack of voter confidence is their ignorance of the fact that even an unbiased system will get the "wrong" result some of the time. Since we lack an objective measure of the support of the candidates, there is of course no "right" result of the election beyond the actual results, but in the end I think that what happens is that when elections are close voters come face-to-face with what scientists have been facing for centuries under the name like "measurement error" and "scientific significance", they (the voters) tend to ascribe the problem to systematic bias rather than random error. It's true that less transparent systems make it easier for the voters to believe in conspiracy theories, but the underlying problem is lack of scientific thinking skills. I'd predict that after a close election voters will react the same way regardless of the technology employed (or lack thereof).
In principle, the ballots being counted in public in front of everyone in the village will inspire more confidence than an obscure computer calculation. However, as the 2000 Florida debacle has demonstrated, hand counting has its own problems (e.g. error rates) which the voting public does not understand either. It seems to me that if the system artificially produces a landslide (e.g. via a winner-take-all-state-electoral-votes system), the public is happy that things went well. If the elections are close there is a lot of consternation and misunderstanding. On the technical level, ballots that are both human- and machine-countable but generated automatically (so there is less room for voter marking errors], look best to me. If the voting machine prints the ballot out but keeps no record otherwise that would be best. But just wait for a close elections and the voters will express lack of confidence in the results. The problem is the following: if you are trying to measure a large effect, then you will get the right result no matter what method you use and everyone will be quite confident you got the right result. If you are trying to measure an effect which is just at the level of resolution for your detector (or worse, as in the Florida case, below the measurement error) then there is no way to be as confident that you got the right result.
Because there's more to a bus than the bandwidth. USB has a lot of overhead (it can be branched, hook many devices etc). SATA is dedicated for controlling storage. That's why we put cameras on the USB, hard-drives on a SATA bus, the network card on the PCI bus, video card on the VESA bus ...
More to the point, this is much ado about nothing. Even if Mr. Torvalds "saw the light" and decided he wanted to move to GPL v3, this would be impossible in practical terms since Linux has no copyright escrow agent similar to the FSF for GNU. In other words, to move code licensed to Linux under GPL v2 (only) to GPL v3 requires re-licensing by the original author -- which you may never be able to find. So, you may safely assume that Linux will be GPL v2 until it is re-written from scratch.
If I were RMS, I would forbid the packaging of any GNU code with a GPLv2 GNU/Linux. Not only would this against RMS's free software philosophy, but the GPL (yes, even version 3!) expressly disclaims any limitation on the mere packaging of software. More to the point, this is much ado about nothing. Even if Mr. Torvalds "saw the light" and decided he wanted to move to GPL v3, this would be impossible in practical terms since Linux has no copyright escrow agent similar to the FSF for GNU. In other words, to move code licensed to Linux under GPL v2 (only) to GPL v3 requires re-licensing by the original author -- which you may never be able to find. So, you may safely assume that Linux will be GPL v2 until it is re-written from scratch.
You are right to point at the older story -- we need to make a distinction. The scientific point here is the use of satellite imagery to locate old cities. To social point is that Google Earth has made satellite images infinitely more accessible -- you don't need to be part of NASA anymore.
Gilmore lost his suit; the Court of Appeals and the Supreme Court refused to intervene; see the coverage at Papers Please!.
As you are surely aware, the US does not consider following treaty obligations a serious duty. Rather, when the US determines that a treaty obligation is against national interest, it just ignores it. Besides, given that the infinite copyright terms in the treaties were originally put there at the insistence of the US, the US can easily repudiate its earlier position and ask for everyone to reconsider.
Please take a moment to brush up on your history before you make such un-informed posts. If you join right now they tell you what the current behaviour is: "share" means with all contacts. That's an acceptable behaviour. Nothing wrong with it, execapt that the "sharing" feature is not new, and before it had a different behaviour: the shared feeds appeared on a "secret" URL, and you had control over who could see the feeds. So, if you "shared" feeds before, and didn't notice the new announcement, suddenly a lot of people were told of your share -- people who you may have not wanted to know. That is a serious privacy concern.
Given the abject disasters that Jackson's "The Two Towers" and "The Return of the King" were, is there any reason to expect that he'll do well with the Hobbit, not to speak of a "sequel" which will be mostly original? This is, of course, a matter of opinion, but it seemed to me that the weakest parts of there two films were the segments were Tolkien's material was explicitly rejected in favour of new material. The technical side will surely be amazing, but that is not quite reason enough to see a movie for me.
SFLC is the Software Freedom Law Center. You can think of it as the militant arm of the Free Software Foundation (FSF), though one does not directly control the other. Its founder and main figure is Prof. Eben Moglen, formerly general counsel and board member of the FSF.
The lawsuit is Nigerian, but it's not so clearly a scam. It seems to be a claim that a keyboard layout (i.e. which key goes where) is a patentable design. Of course in most of the world keyboard layouts are standardized, denying us the fun of learning a new keyboard layout whenever we buy a new keyboard -- but perhaps this isn't the case there. If anything, I would suspect it to be a harassment tactic. I wonder if this Nigerian company has recently started a strategic partnership with a large American software company ...
First of all, the EFF may has not tested your ISP. You may trust them that in general ISPs are sending spoofed packets, but still want to know whether your ISP is using the tactic. Beyond that, however, just because you trust them doesn't mean independent verification has no value. Results mean something different if you obtained them yourself. Also, as in regular science, independent confirmation of results gives more than that: more people conducting tests will also give better data.
I'm not sure why this story was filed under "privacy" rather than "technology". Nobody's freaked out by police helicopters, whether they are used to find traffic offenders, in police chases, or as observation posts for police raids. Using unmanned aircraft instead is a no-brainer. They are cheaper to operate, can stay up longer, and people don't die when they collide (though this incident was with civillian helicopters).
The US, being an enormous country, has a many levels of government. Unlike many other countries, it runs all elections for all levels of government on a fixed date (some Tuesday in November), rather than spreading them around the year. Of course, not every position is up for election every year, but still this means that the "ballot" contains tens if not hundreds of separate elections, ranging all the way from the US President to the county water board and the town mayor, not to mention multiple "ballot initiative" (direct legislation). Each election (especially president, governor etc) can feature tens of candidates (most of them irrelevant). Printed ballots are thick booklets; both filling them correctly and manually reading them is a non-trivial operation. Also, manually tallying the votes in these hundreds of elections takes a lot of time.
This is not to say that this was not done manually in the past, but certainly using computers greatly simplifies the process. I think the best solution is to use computers to generate the ballot, but only use computer counts provisionally. That is, the voter will step up to a computer and will make selections, after which the computer will print a filled ballot that can be optically scanned. The computer will also tally the votes giving a quick result for most of the races. Nevertheless, the printed ballots should be considered the official votes, the ones to be used if a recount is necessary. In important races (President, Governor) it's probably better to automatically count the printed ballots and only use the computer counts for provisional results. Note that this also allows for people to manaually fill their ballots if they feel like it.
Ah, but he's not apologizing for the lawsuits -- he's apologizing for not releasing DRM-riddled restrictively-licensed music fast enough, which he thinks is what forced consumers to share music illegally. He's still behind the lawsuits (except when his own kids share music -- then it's a "family matter" best punished by the parents). He's warning the cell-phone companies that unless they allow limited sharing, consumers will find their own solutions, and not talking about tactics. The content industry (music, film etc) still seems to have no idea what the consumers want, or that the offering people what they want is usually much better than coercing them to buy what you want them to buy.
It actually goes the other way around: if the user's pictures are trapped on a photo-sharing site that only works on Windows, then the user is forever locked to buying future copies of Windows. Any data you have on MS-specific apps is continuously held hostage to ensure you keep paying the Microsoft tax.
First, given the current regulation scheme (the FDA in the US, for example) the distance between a fundamental discovery and an actual drug on the market is much greater in medicine than it is in technology -- Intel does not require approval from anyone to market their next-gen processor. Second, the current patent system makes making trivial improvements on existing drugs (hence extending monopoly protection) much more profitable than researching new drugs (high risk of failing to produce anything).
But even ignoring all these things, on a fundamental level biology is orders of magnitude more difficult than physics. We understand the physics of seminconductors and the mathematics of computation fairly well. We can simulate future processors ahead of time to see if a new cache design will improve performance or not. We have no idea how to simulate a biological system, and barely have quantitative models for event the simplest ones. Let's give it 100 years and try again.
I'd love to know how well these will sell -- perhaps we can arrange a follow-up in a few months?
More seriously, this falls under the "computers as appliances" paradigm. This "home computing appliance" cost less than many "home gaming" appliances, about as much as a cheap all-in-one sound system. It delivers basic internet functionality. The users shouldn't care what OS it runs anymore than they care what OS their printer runs. This is not to say that there can be a potential snag: users trying to install Microsoft-targeted software off the web. I hope there's a good version of wine installed.