Perhaps the math works out to a point but that doesn't make it sound. Why would MS put out a trillion license for something that probably will never go over a couple of billion (tops).
Because 10^12 / 10^38 still leaves them with a factor of 10^26 redundancy. Only 1 in 10^26 keys valid ought to be enough to prevent this attack from being feasible.
But clearly it isn't, so they've actually got more than 10^12 valid licenses, for some bizarre reason.
Sorry, that's their EULA. You have two choices when you purchase anything M$, return the package unopened for a full refund or use it. They do not and can not promise it will work and they are not responsible for the actions of others.
There's this little thing called an implied warranty of fitness for a particular purpose. When you buy something -- anything -- unless it has large letters on the outside of the box saying that it doesn't work, it comes with one. It states that, basically, if you use the product for the purpose for which it is marketed (i.e., with software, try to run it on a computer), it will perform that purpose to at least a basic level.
It is not legally possible for MS's EULA to disclaim this warranty, it's a basic right that you get when you buy something.
When you buy something that doesn't meet this warranty, you're entitled to a full refund. Whether you've opened the package or not.
The other response said it all, but here's another way of looking at it:
For a processor, the minimum clock speed required is
(rate of incoming data) * (# of instructions to process a unit of data) / (average number of instructions per clock cycle, aka IPC)
For a nicely pipelined hardware design, you could theoretically get away with a clock rate equal to the rate of incoming data, or even less, if you can process more than one unit of data per clock and have a separate, higher-clocked piece capturing the input.
Yes, but here's another way of looking at it. The post I was replying to suggested that the FPGA would be competitive with a GPU for some unspecified application. We must assume that it is an application that the GPU would be naturally good at -- that is, a serious number crunching application, because that's the only thing you'd usually consider using a GPU on.
For this kind of application, the limit is almost always incoming data rate. That's why I asked for a 500MHz FPGA -- I did so because 500MHz is approximately the clock rate where an FPGA (which would typically have a 64-bit wide memory interface) could theoretically approach the I/O bandwidth of a GPU (which typically has a 128-bit wide memory interface running between 250 and 300MHz).
Yes, I'm well aware of the architectural benefits of an FPGA over a processor, but note that for most real applications, the pipeline capacity of a GPU provides more than enough data processing and the memory interface will be the bottleneck.
that's because they don't need to clock that high....
That depends on your application. If we're talking about something that you would consider using a GPU for, then what you're talking about is something that uses a very fast memory interface (typically ~6 Gbyte/s throughput), does minimal processing (c. 8-16 floating point operations) on each word read and writes back to memory. That's what GPUs are designed for, so that's what the FPGA board would have to be competitive with to be better than a GPU.
Now, with a 64-bit memory interface (i.e., the widest memory interface most FPGAs will support, because you need to have some kind of phase shifter for each byte of the DDR interface, and most FPGAs only have 8 such devices), you can compete with that on an FPGA clocked at above 400MHz. If you know what you're doing, and have good enough memory, and you have a relatively modern and expensive FPGA. 500MHz would make it more comfortable.
But, the post I was responding to suggested that FPGA boards were a better & cheaper option for this kind of calculation than GPUs. So I ask again, where can I find an affordable FPGA development board that is capable of implementing this kind of specification?
Writing code to use FPU was also a complete pain in the ass, because you had to use assembly, with all the memory management and interrupt handling headaches inherent. FPUs from different vendors weren't guaranteed to have completely compatible instruction sets. Because it was such a pain in the ass, only highly special purpose applications made use of FPU code. [...] Finally, FPU was brought on die as a bog standard part of the CPU. At that point, FPU capabilities could be taken for granted and an explosion of applications requiring an FPU to achieve decent performance ensued (see, for istance, most games). And writing FPU code is now no longer any more difficult than declaring type float. The compiler handles all the tricky parts.
What compilers were you using that didn't do this during the 386/486 era? I mean, I was using MS QuickBASIC back in the 286 days, and *that* supported the FPU when it was available. So did TurboPASCAL 6, which I used to use as well.
they may all say that they're "3D" cards, but the cheap versions are way too slow in this area to be useful
Not really. Take, for instance, this incredibly cheap-ass card. Undeniably cheap, but it supports full DirectX 8.1 pixel shaders and has 4 pipelines @ 266MHz, which really is enough for many applications. It probably won't be running Doom3. But it'll cope with anything less demanding.
You can buy a decent FPGA development board and turn it into a DSP for the price of a high-end graphics card. It isn't a trivial project to get started with, but it might be easier than using a GPU. Plus, the skills and hardware from this project will take you much farther than GPU skills.
Really? I haven't seen PC-insertable FPGA dev boards that are capable of clocking anything like as high as a modern GPU (i.e. typically ~800MHz) for sub-$1000. If you can point me in the direction of a reasonably-priced FPGA dev board that I could implement a 500MHz system on, I'd really appreciate it, but so far I haven't seen one.
Current industry predictions suggest that by 2020 silicon devices will have shrunk to about 20 nanometres......after this... graphene will come into their own. And that gives scientists time to perfect the tricky fabrication methods...
I think if this is to be used in consumer products, market forces will tell them how long they have. Big leaps often come in short time spans. 13 years is a long time and it seems the longer we wait for something to come to market, the more likely it seems to be vapour ware. If this is pure research, they can take their time (and pure research is a good thing too).
I suspect, in fact, they're being hopelessly optmistic thinking they have 13 years until silicon transisters hit 20nm feature size. The last 5 years have seen 130->90->65. We're already in the gear-up to 45. Next 5 or 6 years, by this trend (which shows only one sign of slowing -- exponentially increasing costs) we should see 45->32->22. And given that apparently most transistors in 65nm chips are actually smaller than 65nm, that's probably the predicted "20nm limit". If there really is such a thing.
Congratulations! You've just taken a lossy audio format, and transcoded it into a totally different lossy audio format, with an unnecessary step in the middle provided by Apple. You have caused the quality to degrade significantly; most of the tones in the music will come through okay, but some will be completely trashed. Anything approximating a square wave (any kind of funk groove usually has some of this) will be utterly destroyed. Most of your highest highs will end up completely distorted as well.
This stupid argument about burning and re-ripping is, well, stupid. And yet someone brings it up every time this discussion happens.
You know why? Because what 99% of people want with DRM-free music is the ability to format shift it. Who wants AAC? Well, it's fine for people with iPods, but my player only does MP3 and WMA...
Absolutely. The music industry exists to fulfil a useful role, in the end, and that role isn't manufacturing records and arranging extraordinarily expensive publicity. Anyone could do that job. The job they are there to perform is to filter the music available according to the public's tastes. If they know that nobody will like some band (i.e. because they're totally rubbish), they aren't going to sign them.
The summary here implores me to download what I want from this apparently wonderful (but unfortunately slashdotted) site, but unless someone can direct me to something worth listening to, why would I waste my time listening to the ninety nine recordings of total dreck that Sturgeon's Law tells me I'll need to in order to find the one good one I'm looking for?
That said, you're assuming that they're also running a mail server on the same machine as the webserver. [...] I'm pretty sure your mail would bounce if no connection can be made to port 25.
Yes, but usually only after the relay server has spent a week trying it, in case the server has a temporary problem that's going to be fixed.
I think the problem is less that these people don't know what's going on here, but don't understand what was wrong with sitefinder. They think the entire objection to sitefinder was that it profitted from typos. But that wasn't it at all.
Here's a clue to just one of the issues for them: it should be up to the user what happens when an incorrect domain is entered. Sitefinder took that choice away from the user.
This sounds horrible. I find the idea of overriding another animal's free will very disturbing. The words "won't someone please think of the pidgeons!" come to mind, but we humans are animals after all. I would definitely not want this kind of mind control implemented with humans, and I don't want it implemented on any self-aware being.
This is almost certainly not taking over the pigeon's "free will" (if it ever had any such thing), at least not to any greater extent than a normal training program would.
When this kind of result has been achieved before (in rats), the method used is that a neuron is connected to a computer that can cause it to fire. The rat is then trained (through classical Pavlovian conditioning) to perform some specific action when that neuron fires.
This overrides an animal's free will to no greater extent than teaching a pet to perform tricks does. It's just more efficient. And almost certainly wouldn't work an anything approaching the intelligence of a human.
The only reason for this to be done in China is that in any civilised country the public will torch the lab doing this and they will be right to do so. In fact this will be one of the very few cases where I will happily side up with the animal rights people.
This is presumably how come State University of New York no longer has a biology lab. Wait. I missed that news. Perhaps it didn't happen.
A shame we will never know what this feels like for the pigeon. Is it really being forced to turn left against its will, or does the pigeon experience it as a sudden desire to turn left?
If it's anything like the way they did something similar to rats before, they do it by stimulating an impulse that has previously been trained to make the animal turn in a desired direction -- i.e., more like the sudden desire thing. They could, theoretically, break the training. In practice, I doubt it happens much (at least with rats -- I've kept the suckers. They're *stupid*).
Your point? I know perfectly well what "anthropomorphize" means (although I usually spell it with an s, rather than a z).
The point is that no anthropomorphisation is taking place in sentences like "the house enjoys views across the river" or whatever, because "enjoy" doesn't only refer to the human emotion, but has additional meanings also, which have existed for nearly 6 centuries. They may have been anthropomorphic then; they aren't any more.
Yes, windows supports multiple API's but the rest of the post is incorrect. Win16 was the API used in Windows 3.1.
I don't see the parent contradicting this anywhere.
Windows NT (First appeared in 1993) included Win32.
Nor this; he does appear to be wrong in saying it was designed for Win95, although Win95 was being worked in '93, because most sources seem to agree Win95 was the result of a backport of Win32 on top of Windows 3. So, yes, Win32 was almost certainly designed for NT, not ported to it.
The whole "NT" moniker was because of the microkernel architechture of the operating system.
There are multiple, conflicting explanations of the NT name. None that I've heard relate to it being a microkernel. In fact, it isn't a microkernel -- it runs at least some of its drivers in kernel mode (and, I believe, always has done).
If the arguments in these articles are actually interchangeable, then either:
1) These arguments have no rational basis, and both are BS (unjustified opinions).
2) There are serious factual errors in at least one article.
The arguments are rational, they just pick a side in an engineering tradeoff from which each side is just about as easy to justify -- higher degree of type safety vs flexible use of a small number of primitives. (Unusually, that's Windows playing the 'small number of primitives' card this time, whereas it's more usually a Unix philosophy) As with all such things, which side is best depends on a lot of factors, most notably the programmer's personality. And if this guy has changed his personality to match the Windows way of doing things over those three years of being exposed to Windows, I don't see any reason this should reflect on his reputation.
Do you have a reference to this being an absolute legal requirement? It could make, for instance, working as a porn website designer somewhat legally difficult if that's the case.
Or perhaps you're talking bollocks. I'm not sure.
There have been cases, certainly, where repeated viewing of pornographic material has heald to be sexual harassment. I'm not aware of any occasion a business has been prosecuted for isolated incidents. A warning is the appropriate action to take, not immediate dismissal.
That's an argument that you don't need to accept the EULA. It may or may not be true (note that the retailer doesn't necessarily have legal authority to grant you that license), but that doesn't mean EULAs aren't contracts. Here for example is a SCOTUS case where an EULA was treated as a contract.
Slashdot Mirror
Perhaps the math works out to a point but that doesn't make it sound. Why would MS put out a trillion license for something that probably will never go over a couple of billion (tops).
Because 10^12 / 10^38 still leaves them with a factor of 10^26 redundancy. Only 1 in 10^26 keys valid ought to be enough to prevent this attack from being feasible.
But clearly it isn't, so they've actually got more than 10^12 valid licenses, for some bizarre reason.
Sorry, that's their EULA. You have two choices when you purchase anything M$, return the package unopened for a full refund or use it. They do not and can not promise it will work and they are not responsible for the actions of others.
There's this little thing called an implied warranty of fitness for a particular purpose. When you buy something -- anything -- unless it has large letters on the outside of the box saying that it doesn't work, it comes with one. It states that, basically, if you use the product for the purpose for which it is marketed (i.e., with software, try to run it on a computer), it will perform that purpose to at least a basic level.
It is not legally possible for MS's EULA to disclaim this warranty, it's a basic right that you get when you buy something.
When you buy something that doesn't meet this warranty, you're entitled to a full refund. Whether you've opened the package or not.
The other response said it all, but here's another way of looking at it:
For a processor, the minimum clock speed required is
(rate of incoming data) * (# of instructions to process a unit of data) / (average number of instructions per clock cycle, aka IPC)
For a nicely pipelined hardware design, you could theoretically get away with a clock rate equal to the rate of incoming data, or even less, if you can process more than one unit of data per clock and have a separate, higher-clocked piece capturing the input.
Yes, but here's another way of looking at it. The post I was replying to suggested that the FPGA would be competitive with a GPU for some unspecified application. We must assume that it is an application that the GPU would be naturally good at -- that is, a serious number crunching application, because that's the only thing you'd usually consider using a GPU on.
For this kind of application, the limit is almost always incoming data rate. That's why I asked for a 500MHz FPGA -- I did so because 500MHz is approximately the clock rate where an FPGA (which would typically have a 64-bit wide memory interface) could theoretically approach the I/O bandwidth of a GPU (which typically has a 128-bit wide memory interface running between 250 and 300MHz).
Yes, I'm well aware of the architectural benefits of an FPGA over a processor, but note that for most real applications, the pipeline capacity of a GPU provides more than enough data processing and the memory interface will be the bottleneck.
that's because they don't need to clock that high....
That depends on your application. If we're talking about something that you would consider using a GPU for, then what you're talking about is something that uses a very fast memory interface (typically ~6 Gbyte/s throughput), does minimal processing (c. 8-16 floating point operations) on each word read and writes back to memory. That's what GPUs are designed for, so that's what the FPGA board would have to be competitive with to be better than a GPU.
Now, with a 64-bit memory interface (i.e., the widest memory interface most FPGAs will support, because you need to have some kind of phase shifter for each byte of the DDR interface, and most FPGAs only have 8 such devices), you can compete with that on an FPGA clocked at above 400MHz. If you know what you're doing, and have good enough memory, and you have a relatively modern and expensive FPGA. 500MHz would make it more comfortable.
But, the post I was responding to suggested that FPGA boards were a better & cheaper option for this kind of calculation than GPUs. So I ask again, where can I find an affordable FPGA development board that is capable of implementing this kind of specification?
Writing code to use FPU was also a complete pain in the ass, because you had to use assembly, with all the memory management and interrupt handling headaches inherent. FPUs from different vendors weren't guaranteed to have completely compatible instruction sets. Because it was such a pain in the ass, only highly special purpose applications made use of FPU code.
[...]
Finally, FPU was brought on die as a bog standard part of the CPU. At that point, FPU capabilities could be taken for granted and an explosion of applications requiring an FPU to achieve decent performance ensued (see, for istance, most games). And writing FPU code is now no longer any more difficult than declaring type float. The compiler handles all the tricky parts.
What compilers were you using that didn't do this during the 386/486 era? I mean, I was using MS QuickBASIC back in the 286 days, and *that* supported the FPU when it was available. So did TurboPASCAL 6, which I used to use as well.
they may all say that they're "3D" cards, but the cheap versions are way too slow in this area to be useful
Not really. Take, for instance, this incredibly cheap-ass card. Undeniably cheap, but it supports full DirectX 8.1 pixel shaders and has 4 pipelines @ 266MHz, which really is enough for many applications. It probably won't be running Doom3. But it'll cope with anything less demanding.
About 230W per card.
You can buy a decent FPGA development board and turn it into a DSP for the price of a high-end graphics card. It isn't a trivial project to get started with, but it might be easier than using a GPU. Plus, the skills and hardware from this project will take you much farther than GPU skills.
Really? I haven't seen PC-insertable FPGA dev boards that are capable of clocking anything like as high as a modern GPU (i.e. typically ~800MHz) for sub-$1000. If you can point me in the direction of a reasonably-priced FPGA dev board that I could implement a 500MHz system on, I'd really appreciate it, but so far I haven't seen one.
The traditional design is, I believe, a spherical section. Obviously there are a variety of different wok designs, and some may well be parabolic.
OK, just chasing up my references, seems hexagonal is correct. But still -- why ripply? It's supposed to be *flat*. That's *the point*. :)
I mean, the article's about a completely flat sheet of atoms joined in a structure with four edges from each node.
So, why are they showing a ripply surface made from a hexagonal structure, with three edges from each node?
Current industry predictions suggest that by 2020 silicon devices will have shrunk to about 20 nanometres...
I think if this is to be used in consumer products, market forces will tell them how long they have. Big leaps often come in short time spans. 13 years is a long time and it seems the longer we wait for something to come to market, the more likely it seems to be vapour ware. If this is pure research, they can take their time (and pure research is a good thing too).
I suspect, in fact, they're being hopelessly optmistic thinking they have 13 years until silicon transisters hit 20nm feature size. The last 5 years have seen 130->90->65. We're already in the gear-up to 45. Next 5 or 6 years, by this trend (which shows only one sign of slowing -- exponentially increasing costs) we should see 45->32->22. And given that apparently most transistors in 65nm chips are actually smaller than 65nm, that's probably the predicted "20nm limit". If there really is such a thing.
Congratulations! You've just taken a lossy audio format, and transcoded it into a totally different lossy audio format, with an unnecessary step in the middle provided by Apple. You have caused the quality to degrade significantly; most of the tones in the music will come through okay, but some will be completely trashed. Anything approximating a square wave (any kind of funk groove usually has some of this) will be utterly destroyed. Most of your highest highs will end up completely distorted as well.
This stupid argument about burning and re-ripping is, well, stupid. And yet someone brings it up every time this discussion happens.
You know why? Because what 99% of people want with DRM-free music is the ability to format shift it. Who wants AAC? Well, it's fine for people with iPods, but my player only does MP3 and WMA...
This will require a lossy->lossy step anyway.
Absolutely. The music industry exists to fulfil a useful role, in the end, and that role isn't manufacturing records and arranging extraordinarily expensive publicity. Anyone could do that job. The job they are there to perform is to filter the music available according to the public's tastes. If they know that nobody will like some band (i.e. because they're totally rubbish), they aren't going to sign them.
The summary here implores me to download what I want from this apparently wonderful (but unfortunately slashdotted) site, but unless someone can direct me to something worth listening to, why would I waste my time listening to the ninety nine recordings of total dreck that Sturgeon's Law tells me I'll need to in order to find the one good one I'm looking for?
That said, you're assuming that they're also running a mail server on the same machine as the webserver. [...] I'm pretty sure your mail would bounce if no connection can be made to port 25.
Yes, but usually only after the relay server has spent a week trying it, in case the server has a temporary problem that's going to be fixed.
I think the problem is less that these people don't know what's going on here, but don't understand what was wrong with sitefinder. They think the entire objection to sitefinder was that it profitted from typos. But that wasn't it at all.
Here's a clue to just one of the issues for them: it should be up to the user what happens when an incorrect domain is entered. Sitefinder took that choice away from the user.
The question is can you stop a pigeon A'ing and B'ing when you don't want it to?
This sounds horrible. I find the idea of overriding another animal's free will very disturbing. The words "won't someone please think of the pidgeons!" come to mind, but we humans are animals after all. I would definitely not want this kind of mind control implemented with humans, and I don't want it implemented on any self-aware being.
This is almost certainly not taking over the pigeon's "free will" (if it ever had any such thing), at least not to any greater extent than a normal training program would.
When this kind of result has been achieved before (in rats), the method used is that a neuron is connected to a computer that can cause it to fire. The rat is then trained (through classical Pavlovian conditioning) to perform some specific action when that neuron fires.
This overrides an animal's free will to no greater extent than teaching a pet to perform tricks does. It's just more efficient. And almost certainly wouldn't work an anything approaching the intelligence of a human.
The only reason for this to be done in China is that in any civilised country the public will torch the lab doing this and they will be right to do so. In fact this will be one of the very few cases where I will happily side up with the animal rights people.
This is presumably how come State University of New York no longer has a biology lab. Wait. I missed that news. Perhaps it didn't happen.
A shame we will never know what this feels like for the pigeon. Is it really being forced to turn left against its will, or does the pigeon experience it as a sudden desire to turn left?
If it's anything like the way they did something similar to rats before, they do it by stimulating an impulse that has previously been trained to make the animal turn in a desired direction -- i.e., more like the sudden desire thing. They could, theoretically, break the training. In practice, I doubt it happens much (at least with rats -- I've kept the suckers. They're *stupid*).
Your point? I know perfectly well what "anthropomorphize" means (although I usually spell it with an s, rather than a z).
The point is that no anthropomorphisation is taking place in sentences like "the house enjoys views across the river" or whatever, because "enjoy" doesn't only refer to the human emotion, but has additional meanings also, which have existed for nearly 6 centuries. They may have been anthropomorphic then; they aren't any more.
Yes, windows supports multiple API's but the rest of the post is incorrect. Win16 was the API used in Windows 3.1.
I don't see the parent contradicting this anywhere.
Windows NT (First appeared in 1993) included Win32.
Nor this; he does appear to be wrong in saying it was designed for Win95, although Win95 was being worked in '93, because most sources seem to agree Win95 was the result of a backport of Win32 on top of Windows 3. So, yes, Win32 was almost certainly designed for NT, not ported to it.
The whole "NT" moniker was because of the microkernel architechture of the operating system.
There are multiple, conflicting explanations of the NT name. None that I've heard relate to it being a microkernel. In fact, it isn't a microkernel -- it runs at least some of its drivers in kernel mode (and, I believe, always has done).
If the arguments in these articles are actually interchangeable, then either:
1) These arguments have no rational basis, and both are BS (unjustified opinions).
2) There are serious factual errors in at least one article.
The arguments are rational, they just pick a side in an engineering tradeoff from which each side is just about as easy to justify -- higher degree of type safety vs flexible use of a small number of primitives. (Unusually, that's Windows playing the 'small number of primitives' card this time, whereas it's more usually a Unix philosophy) As with all such things, which side is best depends on a lot of factors, most notably the programmer's personality. And if this guy has changed his personality to match the Windows way of doing things over those three years of being exposed to Windows, I don't see any reason this should reflect on his reputation.
Do you have a reference to this being an absolute legal requirement? It could make, for instance, working as a porn website designer somewhat legally difficult if that's the case.
Or perhaps you're talking bollocks. I'm not sure.
There have been cases, certainly, where repeated viewing of pornographic material has heald to be sexual harassment. I'm not aware of any occasion a business has been prosecuted for isolated incidents. A warning is the appropriate action to take, not immediate dismissal.
That's an argument that you don't need to accept the EULA. It may or may not be true (note that the retailer doesn't necessarily have legal authority to grant you that license), but that doesn't mean EULAs aren't contracts. Here for example is a SCOTUS case where an EULA was treated as a contract.