Now I'm not a system administrator, but this sure sounds like Windows-centric thinking. Especially where you said "run an ssh script on the remote machine". You seem to think the only way to fix a machine is to run specialized programs *on* it.
Have you heard of rsync? rcp? ssh logins? How about nfs? How about centralized home directories? How about running an application that is stored on the network? 20 years ago a sysadmin could "remote administrate" a user without their machine even being turned on! And if it was on they could do everything from a terminal running on their own machine, without telling the user to wait and without the user getting to watch the mouse move around on their screen and open gui tools and accidentally type in the admin password into the clear field.
The only useful thing this provides (other than handholding for people brainwashed by Windows into believing you have to take over a machine to "administer" it) is imaging. Now I don't know what they are using, but at work on reboot the machine already checks and images itself from a central location, updating *both* Windows and Linux of a dual-boot system. Sounds like this already exists. Though if there was more intelligent non-Windows-centric design perhaps the BIOS could respond to a network or at least communicate over a serial line so the user does not have to reboot the machine. THAT would be an improvement!
Neither is theft. But there is a difference: plagarism. The person downloading Britney Spears usually does not try to convince people he wrote and performed the music himself, and sell it with no credit to Britney.
There are literally thousands of violations of the GPL that are similar to the RIAA downloads. I can get RPMs from many, many sites where there certainly is no lead to where I can get the source code. Your Linux-expert friend says "I got the damn kernel to compile correctly, here it is" and gives you a disk without the source and without their modifications to the makefile. Every bittorrent mirror holding a piece of a Linux distribution does not contain the matching source and has no links to the source. Technically all these people are in violation of the GPL.
There has been a sudden flood of people claiming "hypocrasy" at Slashdot. But I don't see it. Not as long as all the above are not being attacked as criminals.
There is a big difference between plagarizing and just making an unauthorized copy for your own use. Some people have pointed out that the sales are the difference, but I think there is still a big difference between ripping a Metallica album and selling it on the street, and making the same album but changing the cover and claiming it is your own music and your own band that recorded it (and somehow being convinced enough that people started thinking Metallica copied you or that the music really is original).
In any case neither of them is "theft". Downloading music is a copyright violation. Reusing GPL code is not only a copyright violation, but also plagarism.
Your analysis is good, but it does not explain why any OSS software exists at all.
The big thing you missed is that the OSS developer may have to write the code anyway. They may need the program working in order to do their job, or they may want to write the program to get the satisfaction of having done so, or to impress their peers, or to learn how to do something.
In such cases then marginal cost between "writing the code" and "writing OSS code" can literally approach zero. It can even be negative if the OSS version causes bug fixes and improvements to be sent to you.
Dual licensing was talked about extensivly above and is a very good way to make the business.
Lots of people think GPL==non commercial, but in fact it allows you to sell your software. The big problem is that anybody who gets it is allowed to do anything they want with it including sell it or give it to other people, making the apparent market cost zero.
However there is a difference, as what you stated has been tried many times in the past, and it was a failure compared to the enormous success of GPL code. This was to mark the code as "you may only use this code for non-commercial purposes". Obviously people disliked that but accepted the GPL. So the potential for selling it or making money from it is apparently valuable and making it very different.
There is a huge difference between downloading a copy of a Metallica album, and taking that copy and claiming you wrote and performed all the music and selling it for a profit.
Before you start thinking you are so clever, please locate somebody complaining about somebody *downloading* a piece of GPL code.
You are right. The correct term is "plagiarism". There is no need or reason to put the word "theft" in there at all, that was stupid on my part. As many people point out the owners of PearOS still possess their code and have not been deprived of it, so this is not theft.
I know you are trying to compare this to the "theft" of copyrighted music by downloaders. But there is a difference.
He didn't just download the GPL code and look at it, he is selling it and claiming he wrote it.
If you download Metallica, it's copyright infringement, but not theft. (actually to be fair, the copyright infringer is the one offering you the copy, the download site). If you download GPL code it's not anything, neither theft or copyright infringement (since that is a specific exception to copyright allowed by the GPL).
If you then took Metallica's songs and made a CD and convinced everybody you wrote and performed them in your garage and started selling copies printed with your own band name, it would be equivalent to what this is. That's is a theft of attribution, as well as a violation of copyright.
The software in a mixed device should not be patented. If the software were somehow part of the patent it must have no obvious use other than to drive the patented device. This means the software is useless and valueless without the patented device, so there is no reason to protect it!
The real threat of software patents is that they are used to lock up interfaces. No patent office would patent a new dimension of bolt threads so that a company can then make devices that require their own fasteners. ALL software patents are exactly like this and this is the real reason they should be disallowed.
I would agree, that is a strange statement. I think he is confusing the difficulty of "compiling" then with "installing" today. In 1986 there was *NO* portable "install" of any type whatsoever.
In 1986 making portable source code that could list a directory, which could be compiled without the user having to edit it, was virtually impossible. Managing to construct a piece of source code that would successfully do this without an error on any platform was probably an engineering feat equal to writing a whole new Unix style OS. "portable" software would at best come with a "edit the config.h file" and then "make -i ".
An "install" would have been virtually impossible and nobody even tried it. Literally every single file in/etc was named differently.
Getting source to compile on the different Linux platforms is pretty easy today. The biggest problems are those autoconf scripts, which are really legacies from trying to solve the 1986 problems. Install is now much more a problem of "I don't have this other thing installed yet", which although it is a huge problem it is insignificant compared to the impossibilty of being portable in 1986.
If Linux reaches a "tipping point" there will be tons of programs and movies for it. Everybody, even Microsoft, will start writing for it.
Too many people keep asking where the open-source free version of their Barbie Fashion Designer is. Hey, Microsoft is not writing that and including it with Windows, you know. How in the world, then, does Windows survive without that being written and included by Microsoft? It's because other people write it.
The fact that people have put so much hard work and time into making things like Open Office or the KDE or Gnome applications is proof that people are desperate to do anything to get rid of Microsoft and will literally do hundreds of times as much work as Microsoft has to do, in order to defeat them. If it was not for Microsoft's monopoly it would not be needed, as commercial software would be available for Linux or whatever the alternative system was.
Unfortunately the fact that Open Office was written has given rise to the greatest piece of FUD ever, spouted repeatedly right here by the astroturfers. That is that for some reason every application on Linux has to be open source and given away for free, and that without it you will not get any software for Linux.
Since I am personally working on an evil closed-source capitalistic piece of software for Linux, and making a good paycheck selling it right now, I consider this attitude pretty stupid and very misleading.
Neither Linux or Windows should be used for this. Maybe Linux without any GUI. But an embedded secure system would be a lot better. Like the old ATM's, they were progably an 8080 running exactly ONE program, not some "OS".
I was at Wells Fargo recently, and they could not print a receipt or statement for a deposit, I got a hand-written form, due to their systems being down, and in fact I still am unsure if they credited the deposit (I was depositing money for somebody else in their account). This was at their nifty new kiosk setups, I was there with about 50 other people who were waiting in line because of problems with the teller machines and they all seemed to think this was normal. I have also been at stores a LOT where you can't buy anything because the systems are down, once I waited for an hour with a huge crowd before giving up (PC Richardsons in NYC). This simply did not happen before about 1990, a store's computer system was always up. It is obvious from the screens that they are all running Windows XP (new colored window borders). The programmers can't even get rid of the damn taskbar despite these being single-use machines, so I would not blame only Microsoft for this horrible mess (I hardly use Windows at all and I know how to hide the taskbar). The fact is that our infrastructure is being designed by idiots who probably thing "running windows" is the natural way a computer works. It is all going to go to complete hell, this mess has only just begun.
MSDOS2 added a Unix style file i/o interface in parallel with the CP/M one and it would report the actual length of the file, which they managed to squeeze into the disk file system somehow.
Our text editor at that time had an option to not put the ^Z at the end of the file, if you were "only using modern programs". However there were still plenty of programs using the CP/M interface, some of them would even crash or read forever if there was no ^Z there.
I don't think that story could possibly be true. Do you know how *small* the code for CP/M is? I would think it would be impossible to insert any sort of "easter egg", especially if the source was available. Possibly into a run-time program like pip, but that would just prove they were emulating CP/M well enough to run the original pip atop it.
MSDOS 1 certainly hard-code what the various devices were. The only thing you could open by name was disk files.
MSDOS 2 had huge improvements becasue at the time they wanted to merge it with Xenix and make a Unix system out of it. It had named devices and opening them as files would connect you to the device drivers. I actually implemented some of these, including what I intended to be a graphical windowing system driven by printing to stdout, it was actually quite usable and powerful.
Unfortunately that level of device support is pretty trivial. The Linux drivers you are complaining about have many more interfaces such as being able to allocate memory and mess with other parts of the kernel. If the driver was limited to read-block and write-block like the MSDOS-2 drivers were, there is no question that they would be completely independent of the kernel.
You are certainly right that "viral" is often used to mean "what differs the GPL from other copyright exceptions".
The problem is that it is impossible to define "viral" as any actual part of the GPL and still make it something that uniquely describes it and does not cover all meaningful copyrights. A copyright exception that allowed you more rights if you are using a copy than if you are using the original is just silly, but that is the only way to make it not be "viral".
"Viral" is usually taken to mean it "infects" other work". This is not true of the GPL. If you use some GPL code to implement part of your program, you can still seperate it by removing that part, and your program is NOT "infected". The only true "viral" license would be NDA's that require you to not work on competitive products. These require signing contracts to be enforcable, so no GPL or EULA or anything else anybody would normally encouter counts.
If "the rules apply to the copy" is what you mean by "viral" than all traditionally-copyrighted work is "viral". If I make a copy of a book under some sort of fair-use rule and give it to you, that does not mean you are free to make unlimited copies from that. The copyright has "infected" the copy.
The problem you are describing is not with Windows or Linux. What you are describing is in fact exactly the lack of a "NX" bit. The Intel processors could not make memory readable and not be executable. Thus if you want to read the data on your stack then it was also possible to jump to it and execute it. The fact that Windows or Linux were unable to fix this problem is not their fault.
Possibly you are confused by 80286 segments, which could make memory readable without being executable (because you could only execute by loading the PC segment register, and the OS could apply totally different rules depending on which segment register was being loaded). However apparently the 80286 scheme has a lot of problems which is why neither Windows or Linux use it for virtual memory (I am not sure what the problems are but it is obvious nobody wanted to work with it).
Essentialy I write GPL software and I cannot use it in any commercial projects.
This is FALSE! You can do anything you want with code you write, including releasing it under dual licenses.
You obviously don't know what you are talking about. Read up on things. Making stupid and completely false statements like this defeats any argument you can make. Looking like idiots here is really killing any logical discussion of GPL alternatives. Personally I would like to see a well-specified LGPL-like thing, where you must release the original source code and patches to it, but can link that in any way into a program constructed of your own code. But as long as anybody arguing against the GPL looks like a loon this is not going to happen.
Actually GPL software could detect the existence of the library and use it if available. Plenty of people write GPL software that runs on Windows, despite the fact that it will not work without Microsoft's copyrighted and secret libraries.
I'm not sure if your argument works for only *one* type of advertisement. A company might well stop advertising somewhere else in order to buy these ads, so their costs do not increase.
Your arguments do apply to advertising in general. Companies are pretty much forced to pay for advertisements and this directly adds to the price of the products. It is likely everything would be cheaper by almost half if there were no advertisements. But even then, huge numbers of people are employed creating and distributing these advertisements, and products supported by these advertisements, so it is quite likely that purchasing power of the population would be reduced by the same amount.
I agree. This has nothing to do with patents. If this story is true, it seems Microsoft could easily win this case, and wants to spin it so it looks like patents are why it won, when they are irrelevant, they will win because it was a violation of their employment contract.
Thats "Bezier" curves. Actually TrueType uses "Quadric" curves, which are the same as a Bezier with the two handle points moved to the same point, this was done to speed up calculations. Even in modern hardware the conversion of curves to straight lines is done by the CPU, though it seems a likely direction the cards will go.
In any case this is not what is needed for hardware acceleration of fonts. The way hardware is likely to be used is the characters will be rasterized and stored at perhaps 128x128 and then it is the hardware's job to produce an antialiased scaled version of this image for every letter. Even present-day good hardware can do this, and use less resources than current font renderings. At sizes greater than 64x64 the hardware will probably be told to fill a triangulated version of the character that the CPU generates, this is actually plenty fast enough for the larger letters.
There are many indications that Longhorn is going to do this. I also recommend that Cairo and so on do this as well. One problem is if the idea of a font+size is encodied to strongly into the interface it makes this difficutl. The size should be completely seperated from the "current font".
Slashdot Mirror
Now I'm not a system administrator, but this sure sounds like Windows-centric thinking. Especially where you said "run an ssh script on the remote machine". You seem to think the only way to fix a machine is to run specialized programs *on* it.
Have you heard of rsync? rcp? ssh logins? How about nfs? How about centralized home directories? How about running an application that is stored on the network? 20 years ago a sysadmin could "remote administrate" a user without their machine even being turned on! And if it was on they could do everything from a terminal running on their own machine, without telling the user to wait and without the user getting to watch the mouse move around on their screen and open gui tools and accidentally type in the admin password into the clear field.
The only useful thing this provides (other than handholding for people brainwashed by Windows into believing you have to take over a machine to "administer" it) is imaging. Now I don't know what they are using, but at work on reboot the machine already checks and images itself from a central location, updating *both* Windows and Linux of a dual-boot system. Sounds like this already exists. Though if there was more intelligent non-Windows-centric design perhaps the BIOS could respond to a network or at least communicate over a serial line so the user does not have to reboot the machine. THAT would be an improvement!
Neither is theft. But there is a difference: plagarism. The person downloading Britney Spears usually does not try to convince people he wrote and performed the music himself, and sell it with no credit to Britney.
There are literally thousands of violations of the GPL that are similar to the RIAA downloads. I can get RPMs from many, many sites where there certainly is no lead to where I can get the source code. Your Linux-expert friend says "I got the damn kernel to compile correctly, here it is" and gives you a disk without the source and without their modifications to the makefile. Every bittorrent mirror holding a piece of a Linux distribution does not contain the matching source and has no links to the source. Technically all these people are in violation of the GPL.
There has been a sudden flood of people claiming "hypocrasy" at Slashdot. But I don't see it. Not as long as all the above are not being attacked as criminals.
There is a big difference between plagarizing and just making an unauthorized copy for your own use. Some people have pointed out that the sales are the difference, but I think there is still a big difference between ripping a Metallica album and selling it on the street, and making the same album but changing the cover and claiming it is your own music and your own band that recorded it (and somehow being convinced enough that people started thinking Metallica copied you or that the music really is original).
In any case neither of them is "theft". Downloading music is a copyright violation. Reusing GPL code is not only a copyright violation, but also plagarism.
Your analysis is good, but it does not explain why any OSS software exists at all.
The big thing you missed is that the OSS developer may have to write the code anyway. They may need the program working in order to do their job, or they may want to write the program to get the satisfaction of having done so, or to impress their peers, or to learn how to do something.
In such cases then marginal cost between "writing the code" and "writing OSS code" can literally approach zero. It can even be negative if the OSS version causes bug fixes and improvements to be sent to you.
Dual licensing was talked about extensivly above and is a very good way to make the business.
Lots of people think GPL==non commercial, but in fact it allows you to sell your software. The big problem is that anybody who gets it is allowed to do anything they want with it including sell it or give it to other people, making the apparent market cost zero.
However there is a difference, as what you stated has been tried many times in the past, and it was a failure compared to the enormous success of GPL code. This was to mark the code as "you may only use this code for non-commercial purposes". Obviously people disliked that but accepted the GPL. So the potential for selling it or making money from it is apparently valuable and making it very different.
There is a huge difference between downloading a copy of a Metallica album, and taking that copy and claiming you wrote and performed all the music and selling it for a profit.
Before you start thinking you are so clever, please locate somebody complaining about somebody *downloading* a piece of GPL code.
You are right. The correct term is "plagiarism". There is no need or reason to put the word "theft" in there at all, that was stupid on my part. As many people point out the owners of PearOS still possess their code and have not been deprived of it, so this is not theft.
I know you are trying to compare this to the "theft" of copyrighted music by downloaders. But there is a difference.
He didn't just download the GPL code and look at it, he is selling it and claiming he wrote it.
If you download Metallica, it's copyright infringement, but not theft. (actually to be fair, the copyright infringer is the one offering you the copy, the download site). If you download GPL code it's not anything, neither theft or copyright infringement (since that is a specific exception to copyright allowed by the GPL).
If you then took Metallica's songs and made a CD and convinced everybody you wrote and performed them in your garage and started selling copies printed with your own band name, it would be equivalent to what this is. That's is a theft of attribution, as well as a violation of copyright.
The software in a mixed device should not be patented. If the software were somehow part of the patent it must have no obvious use other than to drive the patented device. This means the software is useless and valueless without the patented device, so there is no reason to protect it!
The real threat of software patents is that they are used to lock up interfaces. No patent office would patent a new dimension of bolt threads so that a company can then make devices that require their own fasteners. ALL software patents are exactly like this and this is the real reason they should be disallowed.
I would agree, that is a strange statement. I think he is confusing the difficulty of "compiling" then with "installing" today. In 1986 there was *NO* portable "install" of any type whatsoever.
/etc was named differently.
In 1986 making portable source code that could list a directory, which could be compiled without the user having to edit it, was virtually impossible. Managing to construct a piece of source code that would successfully do this without an error on any platform was probably an engineering feat equal to writing a whole new Unix style OS. "portable" software would at best come with a "edit the config.h file" and then "make -i ".
An "install" would have been virtually impossible and nobody even tried it. Literally every single file in
Getting source to compile on the different Linux platforms is pretty easy today. The biggest problems are those autoconf scripts, which are really legacies from trying to solve the 1986 problems. Install is now much more a problem of "I don't have this other thing installed yet", which although it is a huge problem it is insignificant compared to the impossibilty of being portable in 1986.
If the program was innovative you would be posting "it's too hard to learn, why didn't they make it match MS Word", right?
If Linux reaches a "tipping point" there will be tons of programs and movies for it. Everybody, even Microsoft, will start writing for it.
Too many people keep asking where the open-source free version of their Barbie Fashion Designer is. Hey, Microsoft is not writing that and including it with Windows, you know. How in the world, then, does Windows survive without that being written and included by Microsoft? It's because other people write it.
The fact that people have put so much hard work and time into making things like Open Office or the KDE or Gnome applications is proof that people are desperate to do anything to get rid of Microsoft and will literally do hundreds of times as much work as Microsoft has to do, in order to defeat them. If it was not for Microsoft's monopoly it would not be needed, as commercial software would be available for Linux or whatever the alternative system was.
Unfortunately the fact that Open Office was written has given rise to the greatest piece of FUD ever, spouted repeatedly right here by the astroturfers. That is that for some reason every application on Linux has to be open source and given away for free, and that without it you will not get any software for Linux.
Since I am personally working on an evil closed-source capitalistic piece of software for Linux, and making a good paycheck selling it right now, I consider this attitude pretty stupid and very misleading.
Neither Linux or Windows should be used for this. Maybe Linux without any GUI. But an embedded secure system would be a lot better. Like the old ATM's, they were progably an 8080 running exactly ONE program, not some "OS".
I was at Wells Fargo recently, and they could not print a receipt or statement for a deposit, I got a hand-written form, due to their systems being down, and in fact I still am unsure if they credited the deposit (I was depositing money for somebody else in their account). This was at their nifty new kiosk setups, I was there with about 50 other people who were waiting in line because of problems with the teller machines and they all seemed to think this was normal. I have also been at stores a LOT where you can't buy anything because the systems are down, once I waited for an hour with a huge crowd before giving up (PC Richardsons in NYC). This simply did not happen before about 1990, a store's computer system was always up. It is obvious from the screens that they are all running Windows XP (new colored window borders). The programmers can't even get rid of the damn taskbar despite these being single-use machines, so I would not blame only Microsoft for this horrible mess (I hardly use Windows at all and I know how to hide the taskbar). The fact is that our infrastructure is being designed by idiots who probably thing "running windows" is the natural way a computer works. It is all going to go to complete hell, this mess has only just begun.
Roads
Air traffic control
Water
Sewage
Police
Fire department
Tax collection
Border patrol
Defense
War
MSDOS initially had the exact same problem.
MSDOS2 added a Unix style file i/o interface in parallel with the CP/M one and it would report the actual length of the file, which they managed to squeeze into the disk file system somehow.
Our text editor at that time had an option to not put the ^Z at the end of the file, if you were "only using modern programs". However there were still plenty of programs using the CP/M interface, some of them would even crash or read forever if there was no ^Z there.
I don't think that story could possibly be true. Do you know how *small* the code for CP/M is? I would think it would be impossible to insert any sort of "easter egg", especially if the source was available. Possibly into a run-time program like pip, but that would just prove they were emulating CP/M well enough to run the original pip atop it.
MSDOS 1 certainly hard-code what the various devices were. The only thing you could open by name was disk files.
MSDOS 2 had huge improvements becasue at the time they wanted to merge it with Xenix and make a Unix system out of it. It had named devices and opening them as files would connect you to the device drivers. I actually implemented some of these, including what I intended to be a graphical windowing system driven by printing to stdout, it was actually quite usable and powerful.
Unfortunately that level of device support is pretty trivial. The Linux drivers you are complaining about have many more interfaces such as being able to allocate memory and mess with other parts of the kernel. If the driver was limited to read-block and write-block like the MSDOS-2 drivers were, there is no question that they would be completely independent of the kernel.
You are certainly right that "viral" is often used to mean "what differs the GPL from other copyright exceptions".
The problem is that it is impossible to define "viral" as any actual part of the GPL and still make it something that uniquely describes it and does not cover all meaningful copyrights. A copyright exception that allowed you more rights if you are using a copy than if you are using the original is just silly, but that is the only way to make it not be "viral".
"Viral" is usually taken to mean it "infects" other work". This is not true of the GPL. If you use some GPL code to implement part of your program, you can still seperate it by removing that part, and your program is NOT "infected". The only true "viral" license would be NDA's that require you to not work on competitive products. These require signing contracts to be enforcable, so no GPL or EULA or anything else anybody would normally encouter counts.
If "the rules apply to the copy" is what you mean by "viral" than all traditionally-copyrighted work is "viral". If I make a copy of a book under some sort of fair-use rule and give it to you, that does not mean you are free to make unlimited copies from that. The copyright has "infected" the copy.
The problem you are describing is not with Windows or Linux. What you are describing is in fact exactly the lack of a "NX" bit. The Intel processors could not make memory readable and not be executable. Thus if you want to read the data on your stack then it was also possible to jump to it and execute it. The fact that Windows or Linux were unable to fix this problem is not their fault.
Possibly you are confused by 80286 segments, which could make memory readable without being executable (because you could only execute by loading the PC segment register, and the OS could apply totally different rules depending on which segment register was being loaded). However apparently the 80286 scheme has a lot of problems which is why neither Windows or Linux use it for virtual memory (I am not sure what the problems are but it is obvious nobody wanted to work with it).
Essentialy I write GPL software and I cannot use it in any commercial projects.
This is FALSE! You can do anything you want with code you write, including releasing it under dual licenses.
You obviously don't know what you are talking about. Read up on things. Making stupid and completely false statements like this defeats any argument you can make. Looking like idiots here is really killing any logical discussion of GPL alternatives. Personally I would like to see a well-specified LGPL-like thing, where you must release the original source code and patches to it, but can link that in any way into a program constructed of your own code. But as long as anybody arguing against the GPL looks like a loon this is not going to happen.
Actually GPL software could detect the existence of the library and use it if available. Plenty of people write GPL software that runs on Windows, despite the fact that it will not work without Microsoft's copyrighted and secret libraries.
I'm not sure if your argument works for only *one* type of advertisement. A company might well stop advertising somewhere else in order to buy these ads, so their costs do not increase.
Your arguments do apply to advertising in general. Companies are pretty much forced to pay for advertisements and this directly adds to the price of the products. It is likely everything would be cheaper by almost half if there were no advertisements. But even then, huge numbers of people are employed creating and distributing these advertisements, and products supported by these advertisements, so it is quite likely that purchasing power of the population would be reduced by the same amount.
I agree. This has nothing to do with patents. If this story is true, it seems Microsoft could easily win this case, and wants to spin it so it looks like patents are why it won, when they are irrelevant, they will win because it was a violation of their employment contract.
Thats "Bezier" curves. Actually TrueType uses "Quadric" curves, which are the same as a Bezier with the two handle points moved to the same point, this was done to speed up calculations. Even in modern hardware the conversion of curves to straight lines is done by the CPU, though it seems a likely direction the cards will go.
In any case this is not what is needed for hardware acceleration of fonts. The way hardware is likely to be used is the characters will be rasterized and stored at perhaps 128x128 and then it is the hardware's job to produce an antialiased scaled version of this image for every letter. Even present-day good hardware can do this, and use less resources than current font renderings. At sizes greater than 64x64 the hardware will probably be told to fill a triangulated version of the character that the CPU generates, this is actually plenty fast enough for the larger letters.
There are many indications that Longhorn is going to do this. I also recommend that Cairo and so on do this as well. One problem is if the idea of a font+size is encodied to strongly into the interface it makes this difficutl. The size should be completely seperated from the "current font".