long *lpActiveTurdPolish; HRESULT h = GetPolisher(&lpActiveTurdPolish); if (h == S_OK) { ((void STDMETHODCALLTYPE (*)(PROGLANG*))lpActiveTurdPolish)(visual_basic); }
386 segmented memory maps directly to addresses in the 32bit logical address space. That means even the large granularity segments don't help (granularity flag in segment descriptor set). You need to be able to map the 4GB logical address space to places in the larger physical address space. That means using the paging extensions from the pentium-pro (if somebody can correct me on which processor introduced it, I'll accept that since I'm not bothered enough to go through my intel architecture manuals).
As bersl points out, there are other advantages. But I'd like to correct a common misconception - the 32 bit PC chips from (I think) Pentium pro onward from intel, and some time later from the other pc chip manufacturers, support much more than 4GB even though they only support 32 bit addresses in the CPU registers. 32 bits limit any given process on an OS such as Linux or Windows to 4GB of addresses at any one time. Although how much of that can be mapped to real storage is OS and OS version dependent. The AMD 64 initialy provides a virtual memory address space of 2^18 GB, but that is not directly related to the amount of physical memory that can be operated by an OS. apparently that 2^18 (262144) GB may be extended to 2^34 (17179869184) GB in later models.
No, he is pushing *for* a stronger, united Europe. It is you who is against it.
Europe is not united by directives. That only makes a law over its people. Europe is united by driving for the good of its people and that is not happening, that goal is opposed by the proposed constitution. This directive, for example, was passed by a weak, divided europe that couldn't get enough MEPs to vote against the strong, united USA.
After discussion with the patent attourney, I think there should be a time-limited copyright on the choice of mathematical algorithm by limiting the set of algorithms according to engineering constraints. So if I work from the same constraints, I can produce and use the same algorithm. If I have the same constraints and copy their algorithm it would be a breach of copyright since I hadn't put in the investment that they had into figuring it out. If my constraint is to be compatible with the data their program produces, then I work out an algorithm that turns out to be the same, I have clearly put in my own investment. If the producer publishes their mathematical workings out, then I should be free to copy the idea without doing the workings out myself.
The lawyer was right that the determination of which algorithm can require a significant investment, but patents are an unreasonable idiom. They are like a rich landowner purchasing a lordship. life+some years copyright for art (limited to duplication beyond educational purposes and allowing derivatives where the original art does not contribute to the artistic value of the derivative - only the choice of that art representing something). And the limited time frame for copyright on engineering concept where the engineering concept is not *necessary* to enter a market (eg safety critical feature). The engineering concept can also be independently developed from the same constraints like two peices of art can be independently produced. The source code and structure is an artistic form, and would be covered by the normal copyright while the algorithm is an engineering concept made to meet engineering constraints using costly mathematical labour.
I think that is the core of patents (the engineering concept), and the whole point is to protect investments in effort from an investment free competition. Patents don't do that, while a time limited copyright *does*.
ah crap. That's posted with plain text and the arsing thing got formatted all shitty!
;; hw.s ;; assemble with as -o hw.o hw.s ;; then link with ld -o hw hw.o ;; then run with./hw ;; ;; does not account for interrupted or ;; incomplete write, needs more code for that, ;; can't be arsed:)
.globl _start .text
_start: movl $4,%eax movl $1,%ebx movl $msg,%ecx movl $msg_end - msg,%edx int $0x80 movl $1,%eax movl $0,%ebx int $0x80.data
msg:.ascii "Hello, World!\n" msg_end:
;; This is an attempt to get past the (ironically) lame slashdot lameness filter
That doesn't assemble on Linux.;; hw.s;; assemble with as -o hw.o hw.s;; then link with ld -o hw hw.o;; then run with./hw;;;; does not account for interrupted or;; incomplete write, needs more code for that,;; can't be arsed:).globl _start.text
_start: movl $4,%eax
movl $1,%ebx
movl $msg,%ecx
movl $msg_end - msg,%edx
int $0x80
movl $1,%eax
movl $0,%ebx
int $0x80.data msg:.ascii "Hello, World!\n" msg_end:;; This is an attempt to get past the (ironically) lame slashdot lameness filter
I don't agree, if I want my co-worker to have something I've got, I sometimes first think about the co-worker, sometimes the thing, but rarely the give.
Other times I have a result scheduled for production, so I want to select a pre-named product from my list of tasks, then select the objects to work with, then do the work to produce it.
Other times I just want to play with a new tool, so I want to select the tool and get to see what things I can do with it before selecting an object to use it on.
FUSE should make this clean an safe so that users can be permitted to mount loopback/network volumes as they wish on directories they own. At the moment this is restricted due to security issues - although mounting of floppies isn't yet.
Specifically, he committed an API to get a cairo drawable for a window or something (not looked at the details). So its just to let GTK using app devels use cairo calls from their apps for advanced drawing on canvas's and such. So far GTK doesn't actually use cairo. Though apparently GTK head will start using cairo for fonts shortly. Hopefully a theme engine will appear for using cairo for widgets (imagine SVG widgets rendered with a librsvg for cairo:) shaped buttons and stuff !
This reply has a lot of long sentences, ironically in an attempt to be clear:)
> "A method of implementing a psychoacoustic model for sound, comprising..."
No, it wouldn't be a method of implementing, a method of implementing would be "typing on the keyboard while sat at a computer, then running the compiler". It would be something like:
"Claim 1: A method of approximating the information in a sound comprising a set of scoring relations such that the method *can* be implemented in a computational system using few enough basic operations that so much data can be processed in so much time on some baseline hardware, and that the resulting approximation is a strong local or global minima wrt compressibility in the domain of some class of known lossless compression techniques.
Claim 2: The scoring relations.
etc..." (or however a patent should be written).
The implementation of a known psychoacoustic model is trivial. That's the point. Putting any known fact into software is trivial.
The difficult thing is how the information encoded in a sound can be approximated to a more sparse version that can be fed to some known lossless compression technique such that the data is an easy subject for the compression technique and that the sparse approximation sounds disproportionately better than one might expect given the resulting size.
It is finding the set of scores upon the factors that form the information in a sound and how well a known compression algorithm can encode it that is difficult. It is thus *that* that is costly and not protected by copyright, and it is that effort that should be protected by patents.
I will below describe my understanding of what the different IP protections are intended to protect (this will not use terms from statutes).
Patents protect the investment in the discovery of scientific and engineering knowledge where the incentive to discover the knowledge is that it can be used in business, and when it can be discovered by study of your products instead of putting in the effort to research as you did originally.
Copyrights protect the effort in the production of a non-physical "information product" from virtually costless duplication.
Trade secrets protect the investment in the discovery of scientific or engineering knowledge where it cannot be discovered by study of your products.
In software: For shipped products, the structure and design could be discovered by study of a product, so patents would seem to apply. For non-shipped products (eg, Amazon backend), trade secrets would seem to apply. The development of the structure and design is, however, quite a mechanical task, so there is very little investment in that aspect and patents wouldn't apply after all. The investment is in making the code maintainable and bug free, and that aspect of software is covered to extremely powerful effect by copyright.
> On the contrary, software patents are a huge boon to software companies that want to compete with Microsoft. How can a small startup ever get venture capital if Microsoft can eat their technology for lunch?
If Microsoft did not use the code, then Microsoft had every right to implement a similar product that did things in the same way, unless they had signed an agreement not to in order to get the product in earlier software. Such a patent should not apply even to Microsoft. In fact, drive compression is relatively easy. Although depending on the defined semantics of the filesystem interface of the OS, it may be provably impossible in general. In either case, patents shouldn't apply because Microsoft should be allowed to implement an easy idea, not prohibited because somebody made a business case for doing the obvious thing first. If they are to be prohibited, it should be because they are a monopoly and will implement such simple ideas at lower cost employing fewer people without passing the benefits on to allow those they didn't employ to find work elsewhere
An algorithm is mathematics, no more no less. If an algorithm produces a sequence of numbers describing a useful way to move something, then it is an obvious thing to implement in software. It is then obvious to run the software on a computer with an output device that translates the number into a movement. Thus the possible novel/inventive steps are to realise that there does exist an algorithm that describes the sequence, that you can construct the algorithm, and what the algorithm is.
Those are all logical things, and not patentable. There is one other thing that could be considered to be the inventive step, it is to determine the reasonable sampling rate for the sequence and what the sequence is. That is an engineering feat, and fits the idea of patents well.
I will use a concrete example for this: mp3's
It is obvious that the compression waves for a sound can be described by a sequence of numbers. In this case the common options for sample rate are trivial to determine as useful choices. The transformation into and back from the frequency domain is straightforward maths, the framing is an obvious requirement and pretty trivial. The entropy encoding is logical maths and thus should not be patentable. That leaves one thing: the psycoacoustic model. That requires a large amount of experimentation to determine how many bits to assign to various parts of the bitstream, and what data can be thrown away. That is the novel/engineering component of the system, and that is the part that is reasonable to get a patent on. And you would get the patent on using one specific relation between the coefficients of the frequencies and a reasonable approximation that can be effectively losslessly compressed with the (respectively) obvious and mathematically based bitcoding methods that mp3 has.
So mp3 wouldn't be patented (that concept doesn't even make any sense), the advance fraunhofer audio perception model would be patented. LAME could patent theirs too.
That the examiners don't understand this nature of computation is part of the cause of the problems. That policy and law-makers consult only those with something to gain (and thus pay them off with those gains) is the real cause.
Patent lawyers and major corporations lobbied for software and business method patents, and patent lawyers and major corporations benefit. The rest of us cannot set up our small companies to join in. Microsoft became huge without software patents and they wouldn't have if there were software patents back then.
There is no reason that large corporations should get to purchase monopoly control of an easy idea. The decision on who has the right to use a simple idea must *not* be influenced by the ability to pay.
I don't believe the judge in the case you mention ruled that the USPTO had to start granting such patents as a matter of policy, just that that sort of patent should be upheld if they were granted. So they could have waited till they got a prior art database, and had examiners who new that all software is easy, and that running software on a computer is obvious. Software is only faulty or feature incomplete because it takes time to create a given piece of software correctly and completely. Just like paintings and books, etc, that is protected by copyright, as it should be.
And besides, that just means that judge is an idiot, and the opinion of an idiot must never be respected, so stop arguing that its just tough and that the decision has been made. Like all the other patent lawyers (as I gather you are from another of your posts) you are trying to maintain your income and giving advice on the meaning of the ruling based on how it will improve your income for people to believe it. I was under the impression that that was illegal.
This can be changed and strong opposition is what will change it. I understand that you specifically want it to not be changed, but as a legal practitioner you should not be giving incorrect information about the limitations of the influence slashdot readers can have.
16 year olds find 16 year olds attractive because there are attractive 16 year olds, people are not uglier at 16 because they were born in a later year.
If you are born in say 1980, and somebody is born in 1984, then when you are 20, there is a good chance that other person (at 16) is sexually attractive, just as they would have been if you were born in 1984 too. It doesn't matter if you fancy him/her, it doesn't matter if you kiss each other. Just don't fucking fuck, abuse, assault or harrass them.
'course ages of consent are different in different countries. Whether a 16-year old is capable of entering a meaningful sexual relationship with, say, a 20 year old seems to be a matter of debate. There could be a difference between intent to shag for fun for one partner, and normal teenage experimentation for the other. That can cause problems, but then two 16 year olds can have very different attitudes towards sex, and somebody is going to feel hurt then too. Its really a matter of whether a persons actions could be considered to be torture or likely to cause distress and unreasonable feelings of self shame in the other. I say unreasonable because I could feel ashamed if I pull an ugly bird in a nightclub but that is my own fault:)
But laws tend to be black and white so just act according to the most limited of the letter of the law and your own moral values.
5MB hard drive?! You were lucky. We had no hard disk and had to load DOS commands into a RAM disk just to play a game or run a word processor. Then the computer screen would take 3 minutes to warm up and only did green and black. We didn't have a mouse either, we had to pretend with a lump of coal. And we were glad!
What sort of screwdriver do you need? Flathead, Crosshead, or Philips? And what size?
And when you get another screw, will you be needing a second screwdriver, or would you rather replace the existing screwdriver with one that can do both?
> While you almost certainly know a whole lot more about this than I do, I would think that OpenGL with control over pixel shaders will let you do just about anything you really need.
Yes, but not necessarily with good performance. The data may need to be in early. Then you've got all the decisions on how early, how do you decide, is the decision on where it is stored in video RAM made on the board (could be costly), or in the driver. Should probably be in the driver. That means you've got details of the hardware implementation in the driver. Any other way and you've got costly, slow, hot-running hardware. You cannot simply slap a mid-high level graphics API in the hardware and expect to create a domestic product.
OpenGL is only fairly low level when considered as running on a single CPU system (or SMP). As soon as you've got a NUMA system, OpenGL abstracts away a whole lot of communication details and buffering issues.
Slashdot Mirror
you mean:
long *lpActiveTurdPolish;
HRESULT h = GetPolisher(&lpActiveTurdPolish);
if (h == S_OK) {
((void STDMETHODCALLTYPE (*)(PROGLANG*))lpActiveTurdPolish)(visual_basic);
}
They managed to polish win32 and COM too.
386 segmented memory maps directly to addresses in the 32bit logical address space. That means even the large granularity segments don't help (granularity flag in segment descriptor set). You need to be able to map the 4GB logical address space to places in the larger physical address space. That means using the paging extensions from the pentium-pro (if somebody can correct me on which processor introduced it, I'll accept that since I'm not bothered enough to go through my intel architecture manuals).
As bersl points out, there are other advantages. But I'd like to correct a common misconception - the 32 bit PC chips from (I think) Pentium pro onward from intel, and some time later from the other pc chip manufacturers, support much more than 4GB even though they only support 32 bit addresses in the CPU registers. 32 bits limit any given process on an OS such as Linux or Windows to 4GB of addresses at any one time. Although how much of that can be mapped to real storage is OS and OS version dependent. The AMD 64 initialy provides a virtual memory address space of 2^18 GB, but that is not directly related to the amount of physical memory that can be operated by an OS. apparently that 2^18 (262144) GB may be extended to 2^34 (17179869184) GB in later models.
The sea moved several miles inland. How is that only vertical movement?
No, he is pushing *for* a stronger, united Europe. It is you who is against it.
Europe is not united by directives. That only makes a law over its people. Europe is united by driving for the good of its people and that is not happening, that goal is opposed by the proposed constitution. This directive, for example, was passed by a weak, divided europe that couldn't get enough MEPs to vote against the strong, united USA.
> the GPL is not "fair" to potential users who cannot accomodate this "viral" philosophy
It's perfectly fair, they can just ignore the code and write their own version, same as the person who created the GPL licensed code had to do.
"it uses Adobe's Portable Document Format (PDF) standard which is a superset of Adobe Postscript"
Not to mention that Postscript is a turing complete programming language and PDF isn't, so there is no way it can be a superset.
Shame the KDE download rate is virtually non-existent... Come on you guys!
My wristwatch is only... Goddam!
The directors would also have to declare bancruptcy when the shareholders go after them for not looking after their interests.
Yeah,
After discussion with the patent attourney, I think there should be a time-limited copyright on the choice of mathematical algorithm by limiting the set of algorithms according to engineering constraints. So if I work from the same constraints, I can produce and use the same algorithm. If I have the same constraints and copy their algorithm it would be a breach of copyright since I hadn't put in the investment that they had into figuring it out. If my constraint is to be compatible with the data their program produces, then I work out an algorithm that turns out to be the same, I have clearly put in my own investment. If the producer publishes their mathematical workings out, then I should be free to copy the idea without doing the workings out myself.
The lawyer was right that the determination of which algorithm can require a significant investment, but patents are an unreasonable idiom. They are like a rich landowner purchasing a lordship. life+some years copyright for art (limited to duplication beyond educational purposes and allowing derivatives where the original art does not contribute to the artistic value of the derivative - only the choice of that art representing something). And the limited time frame for copyright on engineering concept where the engineering concept is not *necessary* to enter a market (eg safety critical feature). The engineering concept can also be independently developed from the same constraints like two peices of art can be independently produced. The source code and structure is an artistic form, and would be covered by the normal copyright while the algorithm is an engineering concept made to meet engineering constraints using costly mathematical labour.
I think that is the core of patents (the engineering concept), and the whole point is to protect investments in effort from an investment free competition. Patents don't do that, while a time limited copyright *does*.
That doesn't assemble on Linux. ;; hw.s ;; assemble with as -o hw.o hw.s ;; then link with ld -o hw hw.o ;; then run with ./hw ;; ;; does not account for interrupted or ;; incomplete write, needs more code for that, ;; can't be arsed :) .globl _start .text
.data .ascii "Hello, World!\n" ;; This is an attempt to get past the (ironically) lame slashdot lameness filter
_start: movl $4,%eax
movl $1,%ebx
movl $msg,%ecx
movl $msg_end - msg,%edx
int $0x80
movl $1,%eax
movl $0,%ebx
int $0x80
msg:
msg_end:
I don't agree, if I want my co-worker to have something I've got, I sometimes first think about the co-worker, sometimes the thing, but rarely the give.
Other times I have a result scheduled for production, so I want to select a pre-named product from my list of tasks, then select the objects to work with, then do the work to produce it.
Other times I just want to play with a new tool, so I want to select the tool and get to see what things I can do with it before selecting an object to use it on.
FUSE should make this clean an safe so that users can be permitted to mount loopback/network volumes as they wish on directories they own. At the moment this is restricted due to security issues - although mounting of floppies isn't yet.
yeah, the "Audio" tab should be "Sound", and the "Video" tab should be "TV"
Mine does, and DVDs too. My parents computer has Windows XP, it doesn't play DVDs even though it has a DVD drive.
Specifically, he committed an API to get a cairo drawable for a window or something (not looked at the details). So its just to let GTK using app devels use cairo calls from their apps for advanced drawing on canvas's and such. So far GTK doesn't actually use cairo. Though apparently GTK head will start using cairo for fonts shortly. Hopefully a theme engine will appear for using cairo for widgets (imagine SVG widgets rendered with a librsvg for cairo :) shaped buttons and stuff !
This reply has a lot of long sentences, ironically in an attempt to be clear :)
> "A method of implementing a psychoacoustic model for sound, comprising..."
No, it wouldn't be a method of implementing, a method of implementing would be "typing on the keyboard while sat at a computer, then running the compiler". It would be something like:
"Claim 1: A method of approximating the information in a sound comprising a set of scoring relations such that the method *can* be implemented in a computational system using few enough basic operations that so much data can be processed in so much time on some baseline hardware, and that the resulting approximation is a strong local or global minima wrt compressibility in the domain of some class of known lossless compression techniques.
Claim 2: The scoring relations.
etc..." (or however a patent should be written).
The implementation of a known psychoacoustic model is trivial. That's the point. Putting any known fact into software is trivial.
The difficult thing is how the information encoded in a sound can be approximated to a more sparse version that can be fed to some known lossless compression technique such that the data is an easy subject for the compression technique and that the sparse approximation sounds disproportionately better than one might expect given the resulting size.
It is finding the set of scores upon the factors that form the information in a sound and how well a known compression algorithm can encode it that is difficult. It is thus *that* that is costly and not protected by copyright, and it is that effort that should be protected by patents.
I will below describe my understanding of what the different IP protections are intended to protect (this will not use terms from statutes).
Patents protect the investment in the discovery of scientific and engineering knowledge where the incentive to discover the knowledge is that it can be used in business, and when it can be discovered by study of your products instead of putting in the effort to research as you did originally.
Copyrights protect the effort in the production of a non-physical "information product" from virtually costless duplication.
Trade secrets protect the investment in the discovery of scientific or engineering knowledge where it cannot be discovered by study of your products.
In software: For shipped products, the structure and design could be discovered by study of a product, so patents would seem to apply. For non-shipped products (eg, Amazon backend), trade secrets would seem to apply. The development of the structure and design is, however, quite a mechanical task, so there is very little investment in that aspect and patents wouldn't apply after all. The investment is in making the code maintainable and bug free, and that aspect of software is covered to extremely powerful effect by copyright.
> On the contrary, software patents are a huge boon to software companies that want to compete with Microsoft. How can a small startup ever get venture capital if Microsoft can eat their technology for lunch?
If Microsoft did not use the code, then Microsoft had every right to implement a similar product that did things in the same way, unless they had signed an agreement not to in order to get the product in earlier software. Such a patent should not apply even to Microsoft. In fact, drive compression is relatively easy. Although depending on the defined semantics of the filesystem interface of the OS, it may be provably impossible in general. In either case, patents shouldn't apply because Microsoft should be allowed to implement an easy idea, not prohibited because somebody made a business case for doing the obvious thing first. If they are to be prohibited, it should be because they are a monopoly and will implement such simple ideas at lower cost employing fewer people without passing the benefits on to allow those they didn't employ to find work elsewhere
An algorithm is mathematics, no more no less. If an algorithm produces a sequence of numbers describing a useful way to move something, then it is an obvious thing to implement in software. It is then obvious to run the software on a computer with an output device that translates the number into a movement. Thus the possible novel/inventive steps are to realise that there does exist an algorithm that describes the sequence, that you can construct the algorithm, and what the algorithm is.
Those are all logical things, and not patentable. There is one other thing that could be considered to be the inventive step, it is to determine the reasonable sampling rate for the sequence and what the sequence is. That is an engineering feat, and fits the idea of patents well.
I will use a concrete example for this: mp3's
It is obvious that the compression waves for a sound can be described by a sequence of numbers. In this case the common options for sample rate are trivial to determine as useful choices. The transformation into and back from the frequency domain is straightforward maths, the framing is an obvious requirement and pretty trivial. The entropy encoding is logical maths and thus should not be patentable. That leaves one thing: the psycoacoustic model. That requires a large amount of experimentation to determine how many bits to assign to various parts of the bitstream, and what data can be thrown away. That is the novel/engineering component of the system, and that is the part that is reasonable to get a patent on. And you would get the patent on using one specific relation between the coefficients of the frequencies and a reasonable approximation that can be effectively losslessly compressed with the (respectively) obvious and mathematically based bitcoding methods that mp3 has.
So mp3 wouldn't be patented (that concept doesn't even make any sense), the advance fraunhofer audio perception model would be patented. LAME could patent theirs too.
That the examiners don't understand this nature of computation is part of the cause of the problems. That policy and law-makers consult only those with something to gain (and thus pay them off with those gains) is the real cause.
Patent lawyers and major corporations lobbied for software and business method patents, and patent lawyers and major corporations benefit. The rest of us cannot set up our small companies to join in. Microsoft became huge without software patents and they wouldn't have if there were software patents back then.
There is no reason that large corporations should get to purchase monopoly control of an easy idea. The decision on who has the right to use a simple idea must *not* be influenced by the ability to pay.
I don't believe the judge in the case you mention ruled that the USPTO had to start granting such patents as a matter of policy, just that that sort of patent should be upheld if they were granted. So they could have waited till they got a prior art database, and had examiners who new that all software is easy, and that running software on a computer is obvious. Software is only faulty or feature incomplete because it takes time to create a given piece of software correctly and completely. Just like paintings and books, etc, that is protected by copyright, as it should be.
And besides, that just means that judge is an idiot, and the opinion of an idiot must never be respected, so stop arguing that its just tough and that the decision has been made. Like all the other patent lawyers (as I gather you are from another of your posts) you are trying to maintain your income and giving advice on the meaning of the ruling based on how it will improve your income for people to believe it. I was under the impression that that was illegal.
This can be changed and strong opposition is what will change it. I understand that you specifically want it to not be changed, but as a legal practitioner you should not be giving incorrect information about the limitations of the influence slashdot readers can have.
16 year olds find 16 year olds attractive because there are attractive 16 year olds, people are not uglier at 16 because they were born in a later year.
:)
If you are born in say 1980, and somebody is born in 1984, then when you are 20, there is a good chance that other person (at 16) is sexually attractive, just as they would have been if you were born in 1984 too. It doesn't matter if you fancy him/her, it doesn't matter if you kiss each other. Just don't fucking fuck, abuse, assault or harrass them.
'course ages of consent are different in different countries. Whether a 16-year old is capable of entering a meaningful sexual relationship with, say, a 20 year old seems to be a matter of debate. There could be a difference between intent to shag for fun for one partner, and normal teenage experimentation for the other. That can cause problems, but then two 16 year olds can have very different attitudes towards sex, and somebody is going to feel hurt then too. Its really a matter of whether a persons actions could be considered to be torture or likely to cause distress and unreasonable feelings of self shame in the other. I say unreasonable because I could feel ashamed if I pull an ugly bird in a nightclub but that is my own fault
But laws tend to be black and white so just act according to the most limited of the letter of the law and your own moral values.
5MB hard drive?! You were lucky. We had no hard disk and had to load DOS commands into a RAM disk just to play a game or run a word processor. Then the computer screen would take 3 minutes to warm up and only did green and black. We didn't have a mouse either, we had to pretend with a lump of coal. And we were glad!
What sort of screwdriver do you need? Flathead, Crosshead, or Philips? And what size?
And when you get another screw, will you be needing a second screwdriver, or would you rather replace the existing screwdriver with one that can do both?
I think there's a word for that... Extortion or something.
> While you almost certainly know a whole lot more about this than I do, I would think that OpenGL with control over pixel shaders will let you do just about anything you really need.
Yes, but not necessarily with good performance. The data may need to be in early. Then you've got all the decisions on how early, how do you decide, is the decision on where it is stored in video RAM made on the board (could be costly), or in the driver. Should probably be in the driver. That means you've got details of the hardware implementation in the driver. Any other way and you've got costly, slow, hot-running hardware. You cannot simply slap a mid-high level graphics API in the hardware and expect to create a domestic product.
OpenGL is only fairly low level when considered as running on a single CPU system (or SMP). As soon as you've got a NUMA system, OpenGL abstracts away a whole lot of communication details and buffering issues.