This exists in the Boston area, under the name "ZipCar". I suspect this will only ever be effective for dense metropolitan areas, though, because you will be walking from your front door to the nearest car. Outside of this situation, it's just going to be too much of a pain to deal with the fact that cars will end up at people's houses all the time.
Obviously, there aren't going to be voting reforms to give us a third-party-friendly system so long as the power is held by two main parties opposed to these changes. Overcoming the system to have a non-main-party winner of enough elections to get control on a platform of voting reform is pretty unlikely. It seems to me that the only way to get there from here is to get one or both of the main parties to want a change to the system which, coincidentally or not, permits third parties. Any ideas of what motivation could be given to main parties to support reforms?
My thought is that a system in which a single party could run multiple candidates without those candidates splitting the vote might be beneficial to main parties as well as allowing third parties to get better showings. I bet an acceptance system with Bush, Kerry, and Dean (and others) would be a very interesting race.
As far as I can tell, IBM is still in the top ten in the Fortune 500. Microsoft is still down in the 40s. If IBM sees enough return on investment in OSS to keep contributing to it under the GPL, chances are that the GPL is just fine as a license for software that can make a company money. For that matter, you can buy Linux-based systems from Wal-Mart, at the top of the Fortune 500.
The idea that people only get paid for installation and not development is not at all borne out by the current stats; most of the contributions to the Linux kernel in the 2.6 series to date have come from people whose full-time job is Linux kernel development. The fact is that the way the GPL divides rights and responsibilities actually works just as well for making companies willing to contribute as it does for making developers willing to contribute, and the interested companies employ developers.
The GPL is actually a very strong system of property rights, unlike the flimsy ad hoc contracts that companies have traditionally attempted joint projects, which often break down, leading to acrimonious lawsuits. That's why OSS is becoming more acceptable to businesses of all sizes (and less "exceptible").
The older kids had each had a "home sign" which they developed with their parents; each different, without grammar, small vocabulary. From these they formed a pidgin at school, and the younger kids learned it as a creole. The novel thing is that there weren't any fully-formed languages that this creole was descended from. Usually, a pidgin is formed by a group of fluent speakers of different languages.
Aside from this data, it would be theoretically possible that all languages have common features because they are all in some way derived from languages that have those features. But these children weren't exposed to any language with UG, and developed a language with UG.
You need to write exploits in order to test whether you've actually fixed the bugs, and in order to determine whether the code is actually correct for some reason you're not seeing.
It is often the case that support for some functionality which is buggy in one implementation will be buggy in other implementations as well, so it is pretty common in general for a lot of similar bugs to turn up at the same time.
Re:Surely you must be joking Mr Feynman
on
Steel Bolt Hacking
·
· Score: 1
IIRC, you're confusing two stories. The time he got it on the first try was a regular office safe. The boss's safe he failed to crack, but someone else got it with the default combo. When he found out, he tried that on other safes and about a fifth were unchanged, but he didn't know about this trick until after the Manhatten Project.
He also had the nervous habit of figuring out the last and second-to-last digits on the combo for an unlocked safe when he was in someone's office. He then hid notes of most of people's combos in the lock of his safe, so he could open his co-workers' safes in ~6 tries.
This is based on a one-time-pad, which provides perfect encryption so long as you have a shared secret which nobody else has and you never reuse bits. The issue is, of course, to get the one-time-pad to the other end. This is an improvement over the original problem, because you don't care what the pad is, so long as both ends agree on it and nobody else knows it.
Quantum crypto works by having the two ends trade photons in such a way that half of them are successfully measured and the other half are garbage; after the measurement, they figure out which were successfully measured without revealing what they were measured to be. An attacker who measures the photons in the middle cannot reproduce them accurately, so the two ends find that they don't agree on the pad.
That is, you know that each bit was read only once. You don't know, for any given bit, if it was read by the right person, but you can, for each bit, randomly check that it was read by the right person or use it as part of the pad. The attacker doesn't know in advance which photons to pass through unmeasured and which to measure.
I somehow doubt that anyone will think they're watching a news broadcast when they see a movie, at least in theaters. For that matter, I bet they wouldn't really be able to fool anyone longer than it takes to switch to CNN and see that there's no special report.
Actually, voice is terrible for controlling anything that doesn't talk back, and pretty bad for anything without a large amount of common sense (i.e., unsolved AI problem). There just isn't enough information in speech to react at all appropriately to it without a very good understanding of context, and you generally can't express unscripted ideas without dialogue.
On the other hand, there's a lot of information currently available as speech which could be managed more usefully if transcribed automatically. I think the best use is a system which transcribes voice notes, which you can then clean up later (or just treat as rough notes anyway).
The reason there aren't any certified community distros (listed on the FSG page) is probably that community distros tend not to care much about how they are listed. I'd guess that Fedora Core is compliant with some version and just doesn't care about being listed. Most of the other ones, as people have noted, don't use RPM and are probably missing a bunch of things that are only important for strict conformance. (E.g., you're supposed to have a/lib/ld-lsb.so.2; actual programs seem mostly to use/lib/ld-linux.so.2)
Ideally, they'd have a test suite for systems available, and a distribution could claim to be compliant if it passed. Or, better, end users who are concerned could run the test suite themselves to find out.
For that matter, install scripts could include the test suite and check before installing whether your system seems plausible, with sufficient information to complain to your distro if it's not right.
Subject to sections 107 through 122, the owner of copyright under this title has the exclusive rights to do and to authorize any of the following:
(1) to reproduce the copyrighted work in copies or phonorecords;
(2) to prepare derivative works based upon the copyrighted work;
(3) to distribute copies or phonorecords of the copyrighted work to the public by sale or other transfer of ownership, or by rental, lease, or lending;
(4) in the case of literary, musical, dramatic, and choreographic works, pantomimes, and motion pictures and other audiovisual works, to perform the copyrighted work publicly;
(5) in the case of literary, musical, dramatic, and choreographic works, pantomimes, and pictorial, graphic, or sculptural works, including the individual images of a motion picture or other audiovisual work, to display the copyrighted work publicly; and
(6) in the case of sound recordings, to perform the copyrighted work publicly by means of a digital audio transmission.
Downloading is legal, because the copyright owner has no exclusive right to prohibit reception of their works. On the other hand, it may not be legal to write it to disk (which has been argued to be an act of copying and a violation of (1)). So they'd have to catch you writing it to disk in order to go after you for downloading music, which is even more impractical. Furthermore, it is unclear whether this is actually illegal or whether it is covered by fair use. Additionally, it would be difficult to demonstrate that "the effect of the use upon the potential market for or value of the copyrighted work" (107 (4)) is non-negligible for the individual act of making a copy for personal use. Chances are that the RIAA would not actually bring a lawsuit for downloading, simply due to the possibility that they could lose, thereby setting precedent that downloading music is legal.
The issue isn't really anything to do with shared libraries, per se, which are designed correctly. The issue is that packaging systems don't always have a way to deal with having multiple versions of the same package installed.
The actual dynamic linker is perfectly happy to have versions 3 and 4 of a library both installed, because it actually looks for libfoo.so.3 or libfoo.so.4. Actually, the libraries contain the information of what to look for; so libfoo.so.3.0.4 specifies libfoo.so.3, and so libfoo.so.3.0.5 will be accepted as a replacement, and libfoo.so.4 will be ignored. Of course, there are potential problems when libqt.so.3.2.0 isn't actually a suitable replacement for libqt.so.3.0.0, but claims to be, but that's due to people misusing the system, not the system failing to be sufficient to the problem.
IPv6 is presently useful for routing most of the traffic on the internet. The backbones are tunnelling IPv4 over IPv6 these days, because IPv6 routes better.
IPv4 addresses are mostly around 8 or 9 digits, and hard to remember. That's why people generally don't memorize them. DNS works almost exactly the same for IPv6 and IPv4.
It's trivial to get the firewalling benefits of NAT without actually doing NAT; people using NAT now will stop just translating the addresses and leave things the same as far as not passing unrequested inbound connections.
The best way to get rid of a tool for evil may well be to abuse it until it breaks. If possible, the quickest way to get rid of a bad law is to target its supporters.
We would obviously be opposed to an amendment to a law that states that it may only be applied by the immoral; the moral shouldn't add this in effect by way of self-control. I don't recall any comments to the effect that reverse engineering should be illegal, but that people who do it should be forgiven anyway.
What was just released was X11R6.8.0; this is the latest in a series going back to the original X. What this is is the X distribution, which includes a lot of things like clients, fonts, libraries, and so forth. For a long time, it didn't include an X server, because there wasn't an X server that they could distribute, because this distribution was supposed to be cross-platform, and there wasn't a portable X server.
What XFree86 would do is get the latest X distribution, add an X server for x86, and release the whole thing. XFree86 had its own version numbers, which were the version of the server. The version of the X distribution was reported in the release notes.
There was also some discussion of the X distribution changing license to something not compatible with the GPL, which meant that XFree86 took over from the X consortium as the trusted open source organization.
Three things happened: XFree86 started supporting non-x86 platforms, and became a reasonable thing to include in the portable distribution, the license-changing mumbling on the X distribution side ended (with a change of organization), and XFree86 had developer interaction problems and changed license. This meant that X.org, now in control of the X distribution, decided they had to take the standard back, took the XFree86 server, put it in the X distribution, and announced themselves open for development on the complete system, including the server and the libraries.
What you have is actually X11R6.6 with XFree86 4.3 added, packaged by the XFree86 folks. What was just releases was X11R6.8.0, which includes a new X server derived from the XFree86 one, but with further development at X.org.
X11R6 is the major version, which is to say that all X11R6 programs should be able to interoperate (although they may end up figuring out that they need unsupported features, and not actually work). X because they felt like it, 11 because that's when they got the protocol design down, R for "release", 6 is where they got the protocol contents down, and we're now up to.8.0 of things that didn't break X11R6.
It would be very interesting to see whether someone accustomed to reading blogs and IRC did better or worse than someone who only read published texts at understanding Chaucer. I would contend that the problem a lot of people have with Chaucer is that they have come to expect standardization in spelling which happened later. If that's true, then it shouldn't bother a blog reader nearly so much when Chaucer spells "scole", "ther", and "veray". Of "stil", "ful", and "wel", which are Chauser, and which are bloggers?
I have never had any use for any of the things Kids Need To Know, while I use the stuff I was interested in all the time. I'm not sure I learned anything in school in K or 4, and 1-3 I was at a Montessori school where they let me study what I wanted to. It seems that most people's lack of interest in something important by the time they get to high school stems from it being associated with something they don't need to know.
The problem with this is that revision control actually takes a certain amount of interaction, and the filesystem doesn't have the interface to interact with the user. The filesystem can't tell that some of the things that get written are auto-save backups, others are the user habitually hitting "save", others are the user exporting to PDF to see how it will actually look when printed, and a few are actual drafts of the document.
Furthermore, as soon as you start integrating changes, you need human help. When you add some text to your US version, and you pull those changes into the UK version, you'll obviously have to look over the new text to change the spelling. The filesystem is going to be hopeless at figuring out what tool is needed for this operation, and this isn't even a case where you're resolving conflicts.
The real solution is just to use a filesystem-external version control system, and integrate it with programs that manipulate documents (i.e., things the user thinks about, as opposed to arbitrary files). Ideally, have a standard API for integration with version control (with the commands "update from with ", "commit to ", and "revert to ", having everything else non-integrated) that applies across version control systems and document-manipulators.
The currently practical solution is to avoid binary formats like the plague and put all your documents in CVS (or SVN or arch). This actually works really well, and would probably work even better if diff and patch understood ZIP files (such as.sxw).
For a low range in the six figures this is true, but as you progress towards the richer and then the wealthy, the actual percentage of assets paid in taxes drops dramatically.
It's generally the acquisition or conversion of assets which is taxed, not the assets you have (with some exceptions for things whose existence is a burden on the community, like real estate). By percentage of assets, the rich will never pay much tax, because their turnover of assets is a small fraction of their holdings. If you start the year with $5K in the bank, make $20K, and spend $20K, you pay taxes on 80% of the money you had over the course of the year. If you start the year with $1B in the bank, make $1M, and spend $1M, you pay taxes on 0.1% of the money you had.
If assets were taxed instead of the conversion of assets, the rich wouldn't be able to stay rich, because the above billionaire, after a $1M year, would then pay $400M in taxes. Furthermore, these assets aren't generally liquid, and even those that are liquid are generally expected not to be moved; think how many banks would go under if, come April, the rich all had to take 40% of their balances out. If you tax a paper fortune, either you accept the paper as payment of taxes (and essentially nationalize industry) or you suck all the investment out of industry; neither is a particularly good idea.
The main unfair advantage the rich have is the income on their investments and the control over the entities they invest in, which they get simply because they had the money to make the investment, In order to have a fair but practical system, interest would be heavily taxed. Of course, it's tricky, because you have to do something for investments that lose money, but you have to watch out for investments designed to lose money as a tax shelter, and so forth. This would make it much more difficult to live entirely off your interest and not actually do anything and still stay rich.
If that's true, they should just do the GL part in userspace, and have the kernel portion (which they could then release entirely) just stream nVidia-specific commands to the hardware. There are plenty of good drivers already available which stream commands to hardware, so ATi wouldn't learn much (other than what the commands are, which isn't really more informative than probing the bus).
But nVidia's business model is to release as much of their driver code as they are legally permitted to (at least, that's what they say). Most likely, there's some patent licensing agreement involved which would mean that, even if nVidia were to release the source to their drivers, it wouldn't be legal for anyone else to do anything useful with it anyway. So nVidia would be doing the right thing by not inserting code of questionable legality into the kernel tree.
In any case, binary-only drivers aren't really nVidia's business model; they don't actually make any money on them, and they spend development effort on them. Their business model is selling hardware which uses proprietary techniques they've licensed from others. Either ATI is doing this or it isn't, but that's been decided long ago for all of the hardware that's been released.
Slashdot Mirror
This exists in the Boston area, under the name "ZipCar". I suspect this will only ever be effective for dense metropolitan areas, though, because you will be walking from your front door to the nearest car. Outside of this situation, it's just going to be too much of a pain to deal with the fact that cars will end up at people's houses all the time.
Obviously, there aren't going to be voting reforms to give us a third-party-friendly system so long as the power is held by two main parties opposed to these changes. Overcoming the system to have a non-main-party winner of enough elections to get control on a platform of voting reform is pretty unlikely. It seems to me that the only way to get there from here is to get one or both of the main parties to want a change to the system which, coincidentally or not, permits third parties. Any ideas of what motivation could be given to main parties to support reforms?
My thought is that a system in which a single party could run multiple candidates without those candidates splitting the vote might be beneficial to main parties as well as allowing third parties to get better showings. I bet an acceptance system with Bush, Kerry, and Dean (and others) would be a very interesting race.
As far as I can tell, IBM is still in the top ten in the Fortune 500. Microsoft is still down in the 40s. If IBM sees enough return on investment in OSS to keep contributing to it under the GPL, chances are that the GPL is just fine as a license for software that can make a company money. For that matter, you can buy Linux-based systems from Wal-Mart, at the top of the Fortune 500.
The idea that people only get paid for installation and not development is not at all borne out by the current stats; most of the contributions to the Linux kernel in the 2.6 series to date have come from people whose full-time job is Linux kernel development. The fact is that the way the GPL divides rights and responsibilities actually works just as well for making companies willing to contribute as it does for making developers willing to contribute, and the interested companies employ developers.
The GPL is actually a very strong system of property rights, unlike the flimsy ad hoc contracts that companies have traditionally attempted joint projects, which often break down, leading to acrimonious lawsuits. That's why OSS is becoming more acceptable to businesses of all sizes (and less "exceptible").
The older kids had each had a "home sign" which they developed with their parents; each different, without grammar, small vocabulary. From these they formed a pidgin at school, and the younger kids learned it as a creole. The novel thing is that there weren't any fully-formed languages that this creole was descended from. Usually, a pidgin is formed by a group of fluent speakers of different languages.
Aside from this data, it would be theoretically possible that all languages have common features because they are all in some way derived from languages that have those features. But these children weren't exposed to any language with UG, and developed a language with UG.
You need to write exploits in order to test whether you've actually fixed the bugs, and in order to determine whether the code is actually correct for some reason you're not seeing.
It is often the case that support for some functionality which is buggy in one implementation will be buggy in other implementations as well, so it is pretty common in general for a lot of similar bugs to turn up at the same time.
IIRC, you're confusing two stories. The time he got it on the first try was a regular office safe. The boss's safe he failed to crack, but someone else got it with the default combo. When he found out, he tried that on other safes and about a fifth were unchanged, but he didn't know about this trick until after the Manhatten Project.
He also had the nervous habit of figuring out the last and second-to-last digits on the combo for an unlocked safe when he was in someone's office. He then hid notes of most of people's combos in the lock of his safe, so he could open his co-workers' safes in ~6 tries.
This is based on a one-time-pad, which provides perfect encryption so long as you have a shared secret which nobody else has and you never reuse bits. The issue is, of course, to get the one-time-pad to the other end. This is an improvement over the original problem, because you don't care what the pad is, so long as both ends agree on it and nobody else knows it.
Quantum crypto works by having the two ends trade photons in such a way that half of them are successfully measured and the other half are garbage; after the measurement, they figure out which were successfully measured without revealing what they were measured to be. An attacker who measures the photons in the middle cannot reproduce them accurately, so the two ends find that they don't agree on the pad.
That is, you know that each bit was read only once. You don't know, for any given bit, if it was read by the right person, but you can, for each bit, randomly check that it was read by the right person or use it as part of the pad. The attacker doesn't know in advance which photons to pass through unmeasured and which to measure.
I somehow doubt that anyone will think they're watching a news broadcast when they see a movie, at least in theaters. For that matter, I bet they wouldn't really be able to fool anyone longer than it takes to switch to CNN and see that there's no special report.
Actually, voice is terrible for controlling anything that doesn't talk back, and pretty bad for anything without a large amount of common sense (i.e., unsolved AI problem). There just isn't enough information in speech to react at all appropriately to it without a very good understanding of context, and you generally can't express unscripted ideas without dialogue.
On the other hand, there's a lot of information currently available as speech which could be managed more usefully if transcribed automatically. I think the best use is a system which transcribes voice notes, which you can then clean up later (or just treat as rough notes anyway).
Remember to go to about:config and set "keyword:confusing_bugs" to false.
I bet Laughing Liquid just isn't the same.
The reason there aren't any certified community distros (listed on the FSG page) is probably that community distros tend not to care much about how they are listed. I'd guess that Fedora Core is compliant with some version and just doesn't care about being listed. Most of the other ones, as people have noted, don't use RPM and are probably missing a bunch of things that are only important for strict conformance. (E.g., you're supposed to have a /lib/ld-lsb.so.2; actual programs seem mostly to use /lib/ld-linux.so.2)
Ideally, they'd have a test suite for systems available, and a distribution could claim to be compliant if it passed. Or, better, end users who are concerned could run the test suite themselves to find out.
For that matter, install scripts could include the test suite and check before installing whether your system seems plausible, with sufficient information to complain to your distro if it's not right.
Downloading is legal, because the copyright owner has no exclusive right to prohibit reception of their works. On the other hand, it may not be legal to write it to disk (which has been argued to be an act of copying and a violation of (1)). So they'd have to catch you writing it to disk in order to go after you for downloading music, which is even more impractical. Furthermore, it is unclear whether this is actually illegal or whether it is covered by fair use. Additionally, it would be difficult to demonstrate that "the effect of the use upon the potential market for or value of the copyrighted work" (107 (4)) is non-negligible for the individual act of making a copy for personal use. Chances are that the RIAA would not actually bring a lawsuit for downloading, simply due to the possibility that they could lose, thereby setting precedent that downloading music is legal.
The issue isn't really anything to do with shared libraries, per se, which are designed correctly. The issue is that packaging systems don't always have a way to deal with having multiple versions of the same package installed.
The actual dynamic linker is perfectly happy to have versions 3 and 4 of a library both installed, because it actually looks for libfoo.so.3 or libfoo.so.4. Actually, the libraries contain the information of what to look for; so libfoo.so.3.0.4 specifies libfoo.so.3, and so libfoo.so.3.0.5 will be accepted as a replacement, and libfoo.so.4 will be ignored. Of course, there are potential problems when libqt.so.3.2.0 isn't actually a suitable replacement for libqt.so.3.0.0, but claims to be, but that's due to people misusing the system, not the system failing to be sufficient to the problem.
IPv6 is presently useful for routing most of the traffic on the internet. The backbones are tunnelling IPv4 over IPv6 these days, because IPv6 routes better.
IPv4 addresses are mostly around 8 or 9 digits, and hard to remember. That's why people generally don't memorize them. DNS works almost exactly the same for IPv6 and IPv4.
It's trivial to get the firewalling benefits of NAT without actually doing NAT; people using NAT now will stop just translating the addresses and leave things the same as far as not passing unrequested inbound connections.
The best way to get rid of a tool for evil may well be to abuse it until it breaks. If possible, the quickest way to get rid of a bad law is to target its supporters.
We would obviously be opposed to an amendment to a law that states that it may only be applied by the immoral; the moral shouldn't add this in effect by way of self-control. I don't recall any comments to the effect that reverse engineering should be illegal, but that people who do it should be forgiven anyway.
Oh, man. I'm definitely not going to see Fahrenheit 9/11 now...
What was just released was X11R6.8.0; this is the latest in a series going back to the original X. What this is is the X distribution, which includes a lot of things like clients, fonts, libraries, and so forth. For a long time, it didn't include an X server, because there wasn't an X server that they could distribute, because this distribution was supposed to be cross-platform, and there wasn't a portable X server.
.8.0 of things that didn't break X11R6.
What XFree86 would do is get the latest X distribution, add an X server for x86, and release the whole thing. XFree86 had its own version numbers, which were the version of the server. The version of the X distribution was reported in the release notes.
There was also some discussion of the X distribution changing license to something not compatible with the GPL, which meant that XFree86 took over from the X consortium as the trusted open source organization.
Three things happened: XFree86 started supporting non-x86 platforms, and became a reasonable thing to include in the portable distribution, the license-changing mumbling on the X distribution side ended (with a change of organization), and XFree86 had developer interaction problems and changed license. This meant that X.org, now in control of the X distribution, decided they had to take the standard back, took the XFree86 server, put it in the X distribution, and announced themselves open for development on the complete system, including the server and the libraries.
What you have is actually X11R6.6 with XFree86 4.3 added, packaged by the XFree86 folks. What was just releases was X11R6.8.0, which includes a new X server derived from the XFree86 one, but with further development at X.org.
X11R6 is the major version, which is to say that all X11R6 programs should be able to interoperate (although they may end up figuring out that they need unsupported features, and not actually work). X because they felt like it, 11 because that's when they got the protocol design down, R for "release", 6 is where they got the protocol contents down, and we're now up to
It would be very interesting to see whether someone accustomed to reading blogs and IRC did better or worse than someone who only read published texts at understanding Chaucer. I would contend that the problem a lot of people have with Chaucer is that they have come to expect standardization in spelling which happened later. If that's true, then it shouldn't bother a blog reader nearly so much when Chaucer spells "scole", "ther", and "veray". Of "stil", "ful", and "wel", which are Chauser, and which are bloggers?
I have never had any use for any of the things Kids Need To Know, while I use the stuff I was interested in all the time. I'm not sure I learned anything in school in K or 4, and 1-3 I was at a Montessori school where they let me study what I wanted to. It seems that most people's lack of interest in something important by the time they get to high school stems from it being associated with something they don't need to know.
The problem with this is that revision control actually takes a certain amount of interaction, and the filesystem doesn't have the interface to interact with the user. The filesystem can't tell that some of the things that get written are auto-save backups, others are the user habitually hitting "save", others are the user exporting to PDF to see how it will actually look when printed, and a few are actual drafts of the document.
.sxw).
Furthermore, as soon as you start integrating changes, you need human help. When you add some text to your US version, and you pull those changes into the UK version, you'll obviously have to look over the new text to change the spelling. The filesystem is going to be hopeless at figuring out what tool is needed for this operation, and this isn't even a case where you're resolving conflicts.
The real solution is just to use a filesystem-external version control system, and integrate it with programs that manipulate documents (i.e., things the user thinks about, as opposed to arbitrary files). Ideally, have a standard API for integration with version control (with the commands "update from with ", "commit to ", and "revert to ", having everything else non-integrated) that applies across version control systems and document-manipulators.
The currently practical solution is to avoid binary formats like the plague and put all your documents in CVS (or SVN or arch). This actually works really well, and would probably work even better if diff and patch understood ZIP files (such as
For a low range in the six figures this is true, but as you progress towards the richer and then the wealthy, the actual percentage of assets paid in taxes drops dramatically.
It's generally the acquisition or conversion of assets which is taxed, not the assets you have (with some exceptions for things whose existence is a burden on the community, like real estate). By percentage of assets, the rich will never pay much tax, because their turnover of assets is a small fraction of their holdings. If you start the year with $5K in the bank, make $20K, and spend $20K, you pay taxes on 80% of the money you had over the course of the year. If you start the year with $1B in the bank, make $1M, and spend $1M, you pay taxes on 0.1% of the money you had.
If assets were taxed instead of the conversion of assets, the rich wouldn't be able to stay rich, because the above billionaire, after a $1M year, would then pay $400M in taxes. Furthermore, these assets aren't generally liquid, and even those that are liquid are generally expected not to be moved; think how many banks would go under if, come April, the rich all had to take 40% of their balances out. If you tax a paper fortune, either you accept the paper as payment of taxes (and essentially nationalize industry) or you suck all the investment out of industry; neither is a particularly good idea.
The main unfair advantage the rich have is the income on their investments and the control over the entities they invest in, which they get simply because they had the money to make the investment, In order to have a fair but practical system, interest would be heavily taxed. Of course, it's tricky, because you have to do something for investments that lose money, but you have to watch out for investments designed to lose money as a tax shelter, and so forth. This would make it much more difficult to live entirely off your interest and not actually do anything and still stay rich.
If that's true, they should just do the GL part in userspace, and have the kernel portion (which they could then release entirely) just stream nVidia-specific commands to the hardware. There are plenty of good drivers already available which stream commands to hardware, so ATi wouldn't learn much (other than what the commands are, which isn't really more informative than probing the bus).
But nVidia's business model is to release as much of their driver code as they are legally permitted to (at least, that's what they say). Most likely, there's some patent licensing agreement involved which would mean that, even if nVidia were to release the source to their drivers, it wouldn't be legal for anyone else to do anything useful with it anyway. So nVidia would be doing the right thing by not inserting code of questionable legality into the kernel tree.
In any case, binary-only drivers aren't really nVidia's business model; they don't actually make any money on them, and they spend development effort on them. Their business model is selling hardware which uses proprietary techniques they've licensed from others. Either ATI is doing this or it isn't, but that's been decided long ago for all of the hardware that's been released.