If the opposite images are filtered out, your eyes will cross automatically, because your brain is convinced that they're just focusing on an object that's closer.
The real problem is that this technique doesn't allow you to have any objects appear to be outside of a wedge with corners at your eyes and the center of the screen; anything outside of this wedge have one eye or the other looking at the wrong side of the screen. So this is great for objects which seem to hit the user in the face, but is not generally useful. (Actually, it can't get too close to the user's face, either, because that puts the image off the side of the screen, but the user won't look that close anyway)
If all you're doing is anonymous ftp, you might as well just serve the files by http instead of ftp, and use apache. It's generally more stable software, a more manageable protocol, and the firewall rules are simpler.
All we have in this story is the list of packages whose authenticity they haven't been able to confirm yet (I'd guess that they'll get the MD5SUMs from whoever is trusted to upload them in the first place).
Someone commented that the attack was through a known wu-ftp (not GNU software) exploit. I've often wondered why people run ftp servers instead of http ones to distribute stuff; http supports the same operations in a simpler way, and apache is much more cleanly written than wu-ftp.
Insurance companies don't make any money on others' suffering. They lose money on others' suffering and make money on others' expecting to suffer and not actually suffering.
Somebody ought to organize a day on which every interested copyright holder claims that various DMCA-using companies are violating their copyrights on their websites, and see if the public can get them thrown off the net. My guess is that http://www.theESA.com/piracy.html is a copy of somebody's short story about pirates, and http://www.theESA.com/contact.html might be a collection of my personal correspondence.
XFree86 (and X in general) actually comes with substantial documentation for all of this stuff, although it needs to get formatted during the build process. What's mainly lacking, I think, is a sufficiently detailed overview of the API. There's nothing that actually tells you where to look for the clipboard, how to tell things to the window manager, and even things like the best way to get a client-side color array to appear as an image.
It sounds to me like the biggest need that you could help with is documentation of the new features in 1.8.5, especially if you can write documentation that will make sense to people doing the books for small businesses (which seems to be one of the new target markets).
It's actually perfectly reasonable that they didn't find the code in the kernel before they released it. Nobody reads the whole kernel source, and they were in the middle of a lot of upheaval (plus the whole dot-com bubble); part of the advantage of open source is that the license allows you to redistribute stuff you didn't write and you don't have to read through the whole thing, either. If you did, no distribution would get anywhere.
The issue is that they didn't stop distributing the kernel when they discovered code they suspected of being theirs in it. At that point, they could determine that somewhere in the licensing chain, someone had incorporated code they did not own, and the license on the work as a whole was not valid. But SCO continued to distribute the code, an action which could only be legal if either they accepted the validity of the license they received or if they decided to license the code in question under the GPL. Attacking the validity of the GPL doesn't help their case any, because nothing else gives them the right to distribute Linux. Nobody (else) seriously believes that the GPL relinquishes the copyright holder's rights.
I don't think the market would take MS buying SCO licenses for Linux as particularly encouraging for SCO. And it would be too obvious if they talked about a 100 or 50. How many companies in the Fortune 50 are using Linux but not relying on IBM to provide them with a complete solution and take care of everything? It's much more plausible that a company in the 100-500 range would be doing their own Linux deployment and might get sufficiently worried about SCO to buy a license.
The consumer has already come to think of "mp3" as short for compressed digital music. This doesn't mean that Vorbis doesn't have a chance, though. Once the industry has accepted it, consumers will use it, even if they don't realize that their "mp3"s aren't actually mp3 at all. People will download and play Ogg files without knowing the technical details. People already don't know the difference between avi, wmv, and mpg, and really don't know that there are tons of different sorts of mpegs; there's no reason audio won't be the same, with nobody understanding or caring what format they're using, so long as it works, and always calling it "mp3" regardless of what it is.
You clearly haven't looked at the Windows API. At least having consistent and recognizable spelling is an occasional topic of conversation in the Linux development process. (The changes that are going in involve things like "foo_flavor_t *foo_flavour".)
Actually, they aren't trying to be unbiased. If they were, they would say "Windows is better for some things", rather than "Windows could be better for some things". In fact, they haven't done a study at all, and are therefore not qualified to claim that Linux or Windows is good for anything, let alone that one or the other is better for a particular thing. What they are saying is that the embedded market is sufficiently broad that no single study can cover it, that accurate results either way couldn't show nearly this big an effect, and that the methodology in the part of the paper they could critique was badly flawed.
It simply says that the previous paper was marketting fluff; it doesn't say anything about the truth or falsehood of the conclusions of that paper. This is therefore unbiased, although not necessarily particularly useful.
At the level of individual subatomic particles, there is no entropy. Entropy measures the randomness is the system, inverse of the correlations between things, but with a single particle, there's nothing to correlate or be uncorrelated. It takes three particles to have some state that's more likely by chance (2 one way, 1 the other) than another state (all three matching); with fewer, there's no higher entropy state, so entropy can't increase.
Alternatively, you could say that a system like a single electron has already reached maximum entropy. Of course, there are subatomic particles which have not (like neutrons), and nobody really knows about protons.
Entropy never decreases, and may increase if possible. But there are conditions in which it stays the same, either because there is no intermediate state of the correct entropy or because it was already maximized.
'umount -f' does this for NFS mounts. It really ought to apply to any device (like it does in OSX). Obviously, the kernel has to deal with the media disappearing due to physical disconnection (particularly with USB), so reporting to userspace that the filesystem is suddenly gone is nothing new; the kernel should be able to plan for this condition if you tell it to. For that matter, linux isn't that clever about processes doing I/O on missing filesystems. Really, they ought to simply get I/O errors, rather than getting stuck in D, which isn't really helpful.
That's why extra features use new flags, like, for example, -z, which isn't part of traditional tar. A --detect option to determine the right compression program (using file) wouldn't hurt usage in scripts.
Actually, what would make more sense is to have a standard commandline script that deals with a number of combinations correctly and uses the standard conventions (unlike tar) for specifying the files.
The traditional model is to sell the consoles at a loss, and make up the money by selling or licensing games. If the consoles aren't selling, and the games are selling, that would indicate that enough people have the consoles.
What is important to the survival of a system is not that people are buying the console, it's that people are buying the games, because the games are the interesting part, and because a devoted player will buy games regularly, but the console only once.
With respect to people playing for fun, the situation is the same as with collectable card games. People playing for fun have some small chance of randomly coming across a valuable item. But for game balance, you make the rare items distinctive and interesting, but not significantly more powerful than mid-range items. That way, a player who gets a rare item can build a character around it, but the players who never get any interesting rare items can do perfectly well with the items they get.
Nobody knows what the Linux kernel is supposedly infringing, but it's almost certainly not the IP that came from Novell. It's almost certainly something that was developed for Monteray, which had nothing to do with Novell except that one thing it used was UNIX.
This isn't really an attempt to get people to pay for online content. It's actually an attempt to get people to pay online for magazine content. The content in question is not freely available, and it's already proven to the customer. If you pick up a magazine at the supermarket and turn to the "letters to the editor", there will be a number than start, "In your article in the July issue...". If the letter is interesting and you didn't pick up this magazine last month, you may be interested to read this article. You'll probably have a lot of trouble finding it anywhere convenient. These people have it and will show it to you if you've subscribed.
The general problem with online content is that, when you identify that it's worth paying for, you're done with it. The main way to escape this is to either make the content freely available, with some perks for subscribing, or to sell content that the user already knows to be worthwhile. Slashdot gives you everything, with a delay if you're not paying. There are comic strip sites that let you see dialy strips for free, but the old ones require a subscription.
To use a multi-billion dollar example, people are quite willing to pay for books on Borders; it makes sense that they'll pay for magazines on KeepContent. In both cases, they have some expectation in advance of liking the content, and they're used to paying for the same content in a different context.
Emminent Domain is so... 20th century? Constitutional? These days, if someone is in the way of national security, they get declared a material witness; the courts don't even get a chance to get involved. Really pissing off the US government would be really bad for SCO, because SCO's only product at the moment is PR, and that's a bit hard from solitary confinement.
I suspect that most SCO investors are short-term investors. They don't care if SCO has a case or not, because their goal is to hold SCO stock while it goes up. Chances are that SCO stock will drop a lot when SCO is about to need to prove something. Until then, the market is betting that SCO will release good FUD soon, enabling people to profit off the fluctuations.
I'm sure they know things the market doesn't know. What we know is sufficient to explain their behavior...
Of course, a substantial portion of the market probably knows, too, and is just betting that they'll be able to get out in time. In fact, it's pretty likely that the stock won't go down much until their court date.
It's not like there's a shortage of slots (particularly in UCS-32); the problem is getting the character set codified to the point of being suitable for inclusion, and generating fonts which can be used. The people who do this work for fictional scripts are generally more interested in the fictional scripts than in the general issue. If they weren't working on Tengwar, they'd be translating things into Quenya or installing light fixtures or something, not working on Berber.
Getting hobbyists involved in Unicode helps the adoption of UTF-8 and font management and rendering technology. Chances are that more people who add support for non-iso8859 character sets to their programs will be doing so in order to support Tengwar than Berber, even though Berber is probably more important.
Slashdot Mirror
If the opposite images are filtered out, your eyes will cross automatically, because your brain is convinced that they're just focusing on an object that's closer.
The real problem is that this technique doesn't allow you to have any objects appear to be outside of a wedge with corners at your eyes and the center of the screen; anything outside of this wedge have one eye or the other looking at the wrong side of the screen. So this is great for objects which seem to hit the user in the face, but is not generally useful. (Actually, it can't get too close to the user's face, either, because that puts the image off the side of the screen, but the user won't look that close anyway)
I presume that they got a shell with the wu-ftp hole, and root with ptrace. The ptrace exploit is a local exploit.
If all you're doing is anonymous ftp, you might as well just serve the files by http instead of ftp, and use apache. It's generally more stable software, a more manageable protocol, and the firewall rules are simpler.
All we have in this story is the list of packages whose authenticity they haven't been able to confirm yet (I'd guess that they'll get the MD5SUMs from whoever is trusted to upload them in the first place).
Someone commented that the attack was through a known wu-ftp (not GNU software) exploit. I've often wondered why people run ftp servers instead of http ones to distribute stuff; http supports the same operations in a simpler way, and apache is much more cleanly written than wu-ftp.
Insurance companies don't make any money on others' suffering. They lose money on others' suffering and make money on others' expecting to suffer and not actually suffering.
Somebody ought to organize a day on which every interested copyright holder claims that various DMCA-using companies are violating their copyrights on their websites, and see if the public can get them thrown off the net. My guess is that http://www.theESA.com/piracy.html is a copy of somebody's short story about pirates, and http://www.theESA.com/contact.html might be a collection of my personal correspondence.
XFree86 (and X in general) actually comes with substantial documentation for all of this stuff, although it needs to get formatted during the build process. What's mainly lacking, I think, is a sufficiently detailed overview of the API. There's nothing that actually tells you where to look for the clipboard, how to tell things to the window manager, and even things like the best way to get a client-side color array to appear as an image.
It sounds to me like the biggest need that you could help with is documentation of the new features in 1.8.5, especially if you can write documentation that will make sense to people doing the books for small businesses (which seems to be one of the new target markets).
It's actually perfectly reasonable that they didn't find the code in the kernel before they released it. Nobody reads the whole kernel source, and they were in the middle of a lot of upheaval (plus the whole dot-com bubble); part of the advantage of open source is that the license allows you to redistribute stuff you didn't write and you don't have to read through the whole thing, either. If you did, no distribution would get anywhere.
The issue is that they didn't stop distributing the kernel when they discovered code they suspected of being theirs in it. At that point, they could determine that somewhere in the licensing chain, someone had incorporated code they did not own, and the license on the work as a whole was not valid. But SCO continued to distribute the code, an action which could only be legal if either they accepted the validity of the license they received or if they decided to license the code in question under the GPL. Attacking the validity of the GPL doesn't help their case any, because nothing else gives them the right to distribute Linux. Nobody (else) seriously believes that the GPL relinquishes the copyright holder's rights.
I don't think the market would take MS buying SCO licenses for Linux as particularly encouraging for SCO. And it would be too obvious if they talked about a 100 or 50. How many companies in the Fortune 50 are using Linux but not relying on IBM to provide them with a complete solution and take care of everything? It's much more plausible that a company in the 100-500 range would be doing their own Linux deployment and might get sufficiently worried about SCO to buy a license.
The consumer has already come to think of "mp3" as short for compressed digital music. This doesn't mean that Vorbis doesn't have a chance, though. Once the industry has accepted it, consumers will use it, even if they don't realize that their "mp3"s aren't actually mp3 at all. People will download and play Ogg files without knowing the technical details. People already don't know the difference between avi, wmv, and mpg, and really don't know that there are tons of different sorts of mpegs; there's no reason audio won't be the same, with nobody understanding or caring what format they're using, so long as it works, and always calling it "mp3" regardless of what it is.
You clearly haven't looked at the Windows API. At least having consistent and recognizable spelling is an occasional topic of conversation in the Linux development process. (The changes that are going in involve things like "foo_flavor_t *foo_flavour".)
Actually, they aren't trying to be unbiased. If they were, they would say "Windows is better for some things", rather than "Windows could be better for some things". In fact, they haven't done a study at all, and are therefore not qualified to claim that Linux or Windows is good for anything, let alone that one or the other is better for a particular thing. What they are saying is that the embedded market is sufficiently broad that no single study can cover it, that accurate results either way couldn't show nearly this big an effect, and that the methodology in the part of the paper they could critique was badly flawed.
It simply says that the previous paper was marketting fluff; it doesn't say anything about the truth or falsehood of the conclusions of that paper. This is therefore unbiased, although not necessarily particularly useful.
At the level of individual subatomic particles, there is no entropy. Entropy measures the randomness is the system, inverse of the correlations between things, but with a single particle, there's nothing to correlate or be uncorrelated. It takes three particles to have some state that's more likely by chance (2 one way, 1 the other) than another state (all three matching); with fewer, there's no higher entropy state, so entropy can't increase.
Alternatively, you could say that a system like a single electron has already reached maximum entropy. Of course, there are subatomic particles which have not (like neutrons), and nobody really knows about protons.
Entropy never decreases, and may increase if possible. But there are conditions in which it stays the same, either because there is no intermediate state of the correct entropy or because it was already maximized.
'umount -f' does this for NFS mounts. It really ought to apply to any device (like it does in OSX). Obviously, the kernel has to deal with the media disappearing due to physical disconnection (particularly with USB), so reporting to userspace that the filesystem is suddenly gone is nothing new; the kernel should be able to plan for this condition if you tell it to. For that matter, linux isn't that clever about processes doing I/O on missing filesystems. Really, they ought to simply get I/O errors, rather than getting stuck in D, which isn't really helpful.
That's why extra features use new flags, like, for example, -z, which isn't part of traditional tar. A --detect option to determine the right compression program (using file) wouldn't hurt usage in scripts.
Actually, what would make more sense is to have a standard commandline script that deals with a number of combinations correctly and uses the standard conventions (unlike tar) for specifying the files.
The traditional model is to sell the consoles at a loss, and make up the money by selling or licensing games. If the consoles aren't selling, and the games are selling, that would indicate that enough people have the consoles.
What is important to the survival of a system is not that people are buying the console, it's that people are buying the games, because the games are the interesting part, and because a devoted player will buy games regularly, but the console only once.
With respect to people playing for fun, the situation is the same as with collectable card games. People playing for fun have some small chance of randomly coming across a valuable item. But for game balance, you make the rare items distinctive and interesting, but not significantly more powerful than mid-range items. That way, a player who gets a rare item can build a character around it, but the players who never get any interesting rare items can do perfectly well with the items they get.
Nobody knows what the Linux kernel is supposedly infringing, but it's almost certainly not the IP that came from Novell. It's almost certainly something that was developed for Monteray, which had nothing to do with Novell except that one thing it used was UNIX.
This isn't really an attempt to get people to pay for online content. It's actually an attempt to get people to pay online for magazine content. The content in question is not freely available, and it's already proven to the customer. If you pick up a magazine at the supermarket and turn to the "letters to the editor", there will be a number than start, "In your article in the July issue...". If the letter is interesting and you didn't pick up this magazine last month, you may be interested to read this article. You'll probably have a lot of trouble finding it anywhere convenient. These people have it and will show it to you if you've subscribed.
The general problem with online content is that, when you identify that it's worth paying for, you're done with it. The main way to escape this is to either make the content freely available, with some perks for subscribing, or to sell content that the user already knows to be worthwhile. Slashdot gives you everything, with a delay if you're not paying. There are comic strip sites that let you see dialy strips for free, but the old ones require a subscription.
To use a multi-billion dollar example, people are quite willing to pay for books on Borders; it makes sense that they'll pay for magazines on KeepContent. In both cases, they have some expectation in advance of liking the content, and they're used to paying for the same content in a different context.
Emminent Domain is so... 20th century? Constitutional? These days, if someone is in the way of national security, they get declared a material witness; the courts don't even get a chance to get involved. Really pissing off the US government would be really bad for SCO, because SCO's only product at the moment is PR, and that's a bit hard from solitary confinement.
I suspect that most SCO investors are short-term investors. They don't care if SCO has a case or not, because their goal is to hold SCO stock while it goes up. Chances are that SCO stock will drop a lot when SCO is about to need to prove something. Until then, the market is betting that SCO will release good FUD soon, enabling people to profit off the fluctuations.
You've clearly never travelled to a gun platform in the North Sea. It probably means half-drowned and hypothermic.
I'm sure they know things we don't know
I'm sure they know things the market doesn't know. What we know is sufficient to explain their behavior...
Of course, a substantial portion of the market probably knows, too, and is just betting that they'll be able to get out in time. In fact, it's pretty likely that the stock won't go down much until their court date.
It's not like there's a shortage of slots (particularly in UCS-32); the problem is getting the character set codified to the point of being suitable for inclusion, and generating fonts which can be used. The people who do this work for fictional scripts are generally more interested in the fictional scripts than in the general issue. If they weren't working on Tengwar, they'd be translating things into Quenya or installing light fixtures or something, not working on Berber.
Getting hobbyists involved in Unicode helps the adoption of UTF-8 and font management and rendering technology. Chances are that more people who add support for non-iso8859 character sets to their programs will be doing so in order to support Tengwar than Berber, even though Berber is probably more important.