And here is a recent demo of real world performance. Compared to SSB the encoded voice is more artificial sounding, but there is no background noise (hiss and clicks) and it uses less than half the bandwidth to transmit. There is more info and a large playlist of demo/tutorial videos on David Rowe's blog (the creator of codec2).
While all the posts here so far are in favor of this move, it is a very bad thing, and not just for the publishers that depend on ad revenue. If my browser has requested data from the internet, by default the ISP's job is to faithfully forward those requests and the responses to me, not to selectively block, modify, or even inspect the packets I have sent. To do otherwise is a violation of network neutrality.
This is bad because it can be abused by the ISP to serve their goals, and not that of the user. For example, in this case the founder of Free, Xavier Niel, is also a partial owner of the newspaper Le Monde, and by some reports ads are not being blocked on that site, while they are on others. Other accounts give different results with ad blocking, so that may not be intentional, but regardless it is a good hypothetical example of why this can be a very bad idea. It is one thing if the ISP offers additional services that the user can opt-in to use, but very different if they require users to opt-out (many of whom may not even know/understand that the ISP is modifying their traffic).
Your work at Samsung involves making EFL a great library for designing touch-based interfaces for use in a future Linux-based smartphone platform, presumably Tizen. But every time I've heard about Tizen in the press Samsung has made a big deal about HTML5 being the development platform. How do these two development platforms play together? Also can you provide any information about when we can expect to see the first Tizen phones hitting the US?
I hadn't seen it at the time of posting either, but have now. It did have longer action scenes than I was expecting, but I didn't have trouble keeping up with it at 24fps (the 3D 48fps showings around here have been sold out).
Jackson has stated that they used post processing to do the 48 to 24 down-conversion rather than just dropping frames. And many (most?) films these days perform color correction in post processing. I would expect that anyone who is making the jump to 48fps digital won't be shy of using modern digital post processing.
Agreed. Which is why I don't think the Hobbit was the best film to introduce high framerate. Transformers or any other action flicks with "jumbled mess of motion that you can't follow" scenes would have done a much better job of showcasing the technology.
What this means is there's no practical (the film industry definition of practical) way of getting more motion blur than your frame rate and shutter angle allows.
Post processing by averaging frames is definitely practical, especially as these were shot on digital cameras to begin with. Of course, you loose most of the benefit of the higher frame rate if you do that, but it is entirely possible if they decide they don't like the raw results.
In the first case, the camera is not panning, but just filming the scenario as it is, and projector playing it back at the filmed rate. Thus viewing the projection is the same experience as looking at the scene in real life, to within the fidelity of the playback. Notably, there is no depth (or a poor simulation of depth with forced focus), but apart from that higher fidelity should be more realistic. The viewer's eyes will be jumping around the big screen and blinking just like normal so there is absolutely no reason to try to "simulate" that; you have the real-life effect already occurring. Same with motion blur; the eye will supply the same amount of blur that it does in real life, so there is no reason to simulate it, beyond compensating for too *low* of a frame rate, which requires a longer integration time to avoid appearing choppy.
And yet it is exactly this sort of scene that was causing people to deride 48fps as being "soap opera like". They talked about how watching the Hobbits slowly walk down the hill towards them looked epic in 24fps, and looked like a documentary in 48fps. It destroyed the suspension of disbelief for them, and made them think they were looking at actors not Hobbits. That has nothing to faking limitations of human vision. It is completely psychological; whether that psychological effect is inherent in the medium or the result of prior conditioning is debatable, though.
The second scenario is where the camera is panning, and thus forcing visual motion on the user even though they didn't initiate it. This is identical to being smoothly flown around a scene, and how "realistic" it is will depend on whether that would actually happen in real life. In situations where it is realistic my argument above would apply; the eye will be looking around the moving scene just like it would be when looking out a train window.
On the other hand, in situations where panning is being used to simulate human motion, I would argue that 48fps could allow the filmmaker to have more realistic view changes if they want them. Low rate 24fps forces the director to have slow gradual pans less they create a choppy or blurry mess as a result in the limitations of the rate. However, as you pointed out, the eye doesn't work that way. It jumps around, taking time to settle and focus each time. If you tried to do that at 24fps the viewer would get lost, unable to follow the transitions. In large part this is because in real life they are controlling the transitions so they know in advance where the view is changing to, but to a lesser extend this is due to the limitations of the frame rate. Faster frame rates will allow for more abrupt translations that are still possible to follow.
And should the HFH and ACLU and all the other newsletters I subscribe to be blocked as spam as well? They send far more than 400k emails a month. Email is more convenient than RSS or worse Twitter, and is newsletters are a perfectly legitimate use of the medium.
Because those 31 64-core piledriver machines won't be able to push the same amount of IO as 1000 2-core Atom machines. These things aren't for compute intensive tasks. Intel's own advertising comparing to Xeons show the Atoms having twice the performance-per-watt for scale-out tasks, but half the performance-per-watt for compute intensive tasks. It is about providing another option to better match the processor to the task. And it is here today while 64-bit ARM is still a year into the future.
No he's right. On the desktop Safari and Chrome both use heavily modified versions of webkit for rendering, and they use completely different javascript engines. On iOS Chrome is forced to use the exact same webkit as Safari and a crappy interpreted javascript engine, rather than V8 (it's own JIT engine) or even Nitro (Safari's JIT engine). Chrome is prevented from changing anything that matters for the browser so it really is just a skin of UIWebView.
What if I do like the contract, but they break it? Then I go to court to resolve the matter. Oh, wait I can't, I have to go to arbitration, where the result is already determined against me.
If a contract isn't enforceable by law than it isn't a contract anymore. It is a bundle of official looking lies with no weight. Such a thing is unconscionable. It violates the very foundation of market based societies of law, and is tantamount to fraud.
I thought NASA was ordered to be completely open and no information was to be considered sensitive.
While very little of NASA's work is classified, the vast majority of their technical work is covered by ITAR and export control laws, and has to be protected from dissemination outside of the US. Export control can be very over-reaching, and needs to have a major overhaul, however some of the restrictions are on things that could easily be militarized.
Some 300 million US Americans manage to generate the carbon footprint of a Billion Chinese, while 500 million Europeans can hardly hold a candle to the US in terms of carbon emissions.
Well the reality is that 200 million Chinese manage to generate the carbon footprint of 300 million US Americans while the other 1.1 billion Chinese generate very little. Both the US and China need to get their shit together, while India should be commended for being able to ramp up their economy without generating so many greenhouse gases. For bargaining purposes, a more fair arrangement is to agree to a limit of X tons/person + Y tons/GDP.
Nick states here that their code doesn't link against any symbols marked GPL only, and according to the lkml itself, this defines what is permitted for non-GPL code.
It's much more murky than that. In the end what matters is whether the proprietary work is considered a derivative work as interpreted by the courts in various countries. Different people, including different Linux contributors have different opinions on what constitutes a derivative work. Some believe that anything that links against GPL code is a derivative work. In fact this was/is the belief if the GPL authors, otherwise the LGPL wouldn't have been needed. Others don't have a problem with proprietary drivers. It turns out that the original author's intent is fairly important in determining whether infringement is willful, and to a lesser extent whether infringement occurred at all. So EXPORT_SYMBOL_GPL was created to allow individual contributors to express their intent that only GPL compatible code cannot be linked against those symbols. From the legal arguments I've read this makes the question of whether infringement was willful or not pretty clear cut, but it doesn't have nearly as much bearing on whether infringement occurred or not. For more details, read some of Alan Cox's posts on this matter (in this thread and elsewhere).
That said, your interpretation of what those headers mean is very common, and has created a lot of confusion and misinformation, hence my subject title.
The issue here is whether or not Rising Tide backported other's contributions from their GPL version which would be a violation of the GPL.
Except RTS is in full agreement that they are using modules containing code committed by third parties, including those that Andy has pointed out. They have not disputed this at all. What is in dispute is whether combining their proprietary module with the GPL kernel and other GPL modules and is a GPL violation. They are pushing the boundary of the "using APIs is not a derivative work" argument to it's limits.
First off this is nothing like the Oracle case. That was a case about reimplementing APIs, and has nothing to do with linking against someone-else's code that provides APIs. Secondly, it is a stretch to say that the RTS SCSI target is just including APIs. It is using all sorts of internal kernel functions that go far beyond what most reasonable programmers would consider to be an API to the kernel. If you interpret things that liberally, then any proprietary modifications to a GPL application would be allowed by just bundling up the list of functions you happen to use and calling it an API.
If you read the list, it is clear that no one is disputing the facts of what happened; no one is making unfounded assumptions. Rising Tide Systems has a SCSI module that they have written entirely from scratch. In the process of writing this they pushed much of it into the kernel, so much in fact that one of their employees became the Linux SCSI targets maintainer. They have kept some of it back and are shipping a modified kernel containing their code to customers without providing the source code.
They believe that this is allowed, and that their code is not a derivative work of the kernel. Basically RTS is saying: If NVIDIA can have proprietary drivers, why can't we have proprietary kernel subsystems? The other side believes that what NVIDIA is doing likely is a GPL violation, and furthermore some of the technicalities that NVIDIA claims make it legal don't apply to what RTS is doing.
This issue has been ticking time bomb ready to go off. It is entirely possible that some, all, or none of the proprietary drivers written are a violation of the GPL. It all depends on the courts interpretation of derivative work, and no one knows for certain (although some arguments have stronger precedent than others). Furthermore, it is too late to add GPL linking exception to Linux's license to clarify this (one way or the other), because there are too contributors at this point to come to an agreement. So it will remain murky as mud until someone finally sues someone over the issue. Sounds like this may actually be happening. But even then, the ruling may end up depending on the nature of the proprietary extension, and thus remain fuzzy.
I wonder if this means that some others who left in the recent past, like J Allard or Ray Ozzie will be coming back. The rumors were that Sinofsky vigorously opposed their plans, and they left after Balmer decided to back Sinofsky's way rather than them.
The vast majority of murders today don't have any eye witnesses, and yet many of those cases still get solved and the perpetrator caught. Furthermore, it really isn't hard to sneak out behind someone and shoot someone today, without even being seen by the victim. So an invisibility cloak will only make it slightly easier to kill someone, and won't make it any harder to catch them. Not much of a game changer.
So the person that gets more popular votes and electoral votes wins the election
No that's not what happens. For example, suppose that in state A each of 20 districts is 45% Republican and 55% Democrat so all 20 electoral votes go to Democrats. Then in state B each of 15 districts is 65% Republican and 35% Democrat, so all 15 votes go to the Republicans. Then the Democrat wins the electoral vote by 20 to 15, even though the Republican had 54% of the popular vote. This is despite the fact that the electoral votes in this example were even distributed purely based on population (each district is the exact same size).
Splitting large elections into a series of smaller winner takes all elections does not give the same result as popular vote, because it suppresses the minority vote in each district. And that's before you get into bigger practical problems like gerrymandering. If you want popular vote then go with direct popular vote. If you want to maintain the current handicap for smaller states that the electoral congress provides (I think it's good), then require all states to allocate their electoral votes in proportion to the total vote within the entire state. Don't split things up into even smaller winner-takes-all elections.
In case you were tempted to RTFA, don't. You have to click through two ad-laden pages, and there really isn't any more information than in the summary.
I won't speak for Tesla, but SpaceX does not depend on government money. The Falcon 1 was created entirely with private funding, which includes capital investments to build their entire vertically integrated production facilities (they don't contract anything), some launch facilities, and design, construction, and multiple test flights of an entirely new design of rocket. The Falcon 9 was mostly NASA funding, but it built heavily on the Falcon 1 design, and was thus less expensive to design and test than the Falcon 1 (even without including the huge facilities investments mentioned before). Furthermore, SpaceX already had financing to develop Falcon 9 when they won the NASA contract. The contract allowed them to divert that money into the Dragon Capsule instead, the majority of which is thus privately funded.
So without government funding, they would be about where they are with Falcon 1/9, but just getting started with Dragon. Government money sped them up a bit, but they aren't even close to being dependent on that funding.
Slashdot Mirror
And here is a recent demo of real world performance. Compared to SSB the encoded voice is more artificial sounding, but there is no background noise (hiss and clicks) and it uses less than half the bandwidth to transmit. There is more info and a large playlist of demo/tutorial videos on David Rowe's blog (the creator of codec2).
While all the posts here so far are in favor of this move, it is a very bad thing, and not just for the publishers that depend on ad revenue. If my browser has requested data from the internet, by default the ISP's job is to faithfully forward those requests and the responses to me, not to selectively block, modify, or even inspect the packets I have sent. To do otherwise is a violation of network neutrality.
This is bad because it can be abused by the ISP to serve their goals, and not that of the user. For example, in this case the founder of Free, Xavier Niel, is also a partial owner of the newspaper Le Monde, and by some reports ads are not being blocked on that site, while they are on others. Other accounts give different results with ad blocking, so that may not be intentional, but regardless it is a good hypothetical example of why this can be a very bad idea. It is one thing if the ISP offers additional services that the user can opt-in to use, but very different if they require users to opt-out (many of whom may not even know/understand that the ISP is modifying their traffic).
Your work at Samsung involves making EFL a great library for designing touch-based interfaces for use in a future Linux-based smartphone platform, presumably Tizen. But every time I've heard about Tizen in the press Samsung has made a big deal about HTML5 being the development platform. How do these two development platforms play together? Also can you provide any information about when we can expect to see the first Tizen phones hitting the US?
I hadn't seen it at the time of posting either, but have now. It did have longer action scenes than I was expecting, but I didn't have trouble keeping up with it at 24fps (the 3D 48fps showings around here have been sold out).
Jackson has stated that they used post processing to do the 48 to 24 down-conversion rather than just dropping frames. And many (most?) films these days perform color correction in post processing. I would expect that anyone who is making the jump to 48fps digital won't be shy of using modern digital post processing.
Agreed. Which is why I don't think the Hobbit was the best film to introduce high framerate. Transformers or any other action flicks with "jumbled mess of motion that you can't follow" scenes would have done a much better job of showcasing the technology.
What this means is there's no practical (the film industry definition of practical) way of getting more motion blur than your frame rate and shutter angle allows.
Post processing by averaging frames is definitely practical, especially as these were shot on digital cameras to begin with. Of course, you loose most of the benefit of the higher frame rate if you do that, but it is entirely possible if they decide they don't like the raw results.
Lets consider two scenarios here.
In the first case, the camera is not panning, but just filming the scenario as it is, and projector playing it back at the filmed rate. Thus viewing the projection is the same experience as looking at the scene in real life, to within the fidelity of the playback. Notably, there is no depth (or a poor simulation of depth with forced focus), but apart from that higher fidelity should be more realistic. The viewer's eyes will be jumping around the big screen and blinking just like normal so there is absolutely no reason to try to "simulate" that; you have the real-life effect already occurring. Same with motion blur; the eye will supply the same amount of blur that it does in real life, so there is no reason to simulate it, beyond compensating for too *low* of a frame rate, which requires a longer integration time to avoid appearing choppy.
And yet it is exactly this sort of scene that was causing people to deride 48fps as being "soap opera like". They talked about how watching the Hobbits slowly walk down the hill towards them looked epic in 24fps, and looked like a documentary in 48fps. It destroyed the suspension of disbelief for them, and made them think they were looking at actors not Hobbits. That has nothing to faking limitations of human vision. It is completely psychological; whether that psychological effect is inherent in the medium or the result of prior conditioning is debatable, though.
The second scenario is where the camera is panning, and thus forcing visual motion on the user even though they didn't initiate it. This is identical to being smoothly flown around a scene, and how "realistic" it is will depend on whether that would actually happen in real life. In situations where it is realistic my argument above would apply; the eye will be looking around the moving scene just like it would be when looking out a train window.
On the other hand, in situations where panning is being used to simulate human motion, I would argue that 48fps could allow the filmmaker to have more realistic view changes if they want them. Low rate 24fps forces the director to have slow gradual pans less they create a choppy or blurry mess as a result in the limitations of the rate. However, as you pointed out, the eye doesn't work that way. It jumps around, taking time to settle and focus each time. If you tried to do that at 24fps the viewer would get lost, unable to follow the transitions. In large part this is because in real life they are controlling the transitions so they know in advance where the view is changing to, but to a lesser extend this is due to the limitations of the frame rate. Faster frame rates will allow for more abrupt translations that are still possible to follow.
And should the HFH and ACLU and all the other newsletters I subscribe to be blocked as spam as well? They send far more than 400k emails a month. Email is more convenient than RSS or worse Twitter, and is newsletters are a perfectly legitimate use of the medium.
Because those 31 64-core piledriver machines won't be able to push the same amount of IO as 1000 2-core Atom machines.
These things aren't for compute intensive tasks. Intel's own advertising comparing to Xeons show the Atoms having twice the performance-per-watt for scale-out tasks, but half the performance-per-watt for compute intensive tasks. It is about providing another option to better match the processor to the task. And it is here today while 64-bit ARM is still a year into the future.
No he's right. On the desktop Safari and Chrome both use heavily modified versions of webkit for rendering, and they use completely different javascript engines. On iOS Chrome is forced to use the exact same webkit as Safari and a crappy interpreted javascript engine, rather than V8 (it's own JIT engine) or even Nitro (Safari's JIT engine). Chrome is prevented from changing anything that matters for the browser so it really is just a skin of UIWebView.
If this happens, it will be using BGA pads which is fairly difficult to solder at home.
What if I do like the contract, but they break it? Then I go to court to resolve the matter. Oh, wait I can't, I have to go to arbitration, where the result is already determined against me.
If a contract isn't enforceable by law than it isn't a contract anymore. It is a bundle of official looking lies with no weight. Such a thing is unconscionable. It violates the very foundation of market based societies of law, and is tantamount to fraud.
I thought NASA was ordered to be completely open and no information was to be considered sensitive.
While very little of NASA's work is classified, the vast majority of their technical work is covered by ITAR and export control laws, and has to be protected from dissemination outside of the US. Export control can be very over-reaching, and needs to have a major overhaul, however some of the restrictions are on things that could easily be militarized.
Some 300 million US Americans manage to generate the carbon footprint of a Billion Chinese, while 500 million Europeans can hardly hold a candle to the US in terms of carbon emissions.
Well the reality is that 200 million Chinese manage to generate the carbon footprint of 300 million US Americans while the other 1.1 billion Chinese generate very little. Both the US and China need to get their shit together, while India should be commended for being able to ramp up their economy without generating so many greenhouse gases. For bargaining purposes, a more fair arrangement is to agree to a limit of X tons/person + Y tons/GDP.
Nick states here that their code doesn't link against any symbols marked GPL only, and according to the lkml itself, this defines what is permitted for non-GPL code.
It's much more murky than that. In the end what matters is whether the proprietary work is considered a derivative work as interpreted by the courts in various countries. Different people, including different Linux contributors have different opinions on what constitutes a derivative work. Some believe that anything that links against GPL code is a derivative work. In fact this was/is the belief if the GPL authors, otherwise the LGPL wouldn't have been needed. Others don't have a problem with proprietary drivers. It turns out that the original author's intent is fairly important in determining whether infringement is willful, and to a lesser extent whether infringement occurred at all. So EXPORT_SYMBOL_GPL was created to allow individual contributors to express their intent that only GPL compatible code cannot be linked against those symbols. From the legal arguments I've read this makes the question of whether infringement was willful or not pretty clear cut, but it doesn't have nearly as much bearing on whether infringement occurred or not. For more details, read some of Alan Cox's posts on this matter (in this thread and elsewhere).
That said, your interpretation of what those headers mean is very common, and has created a lot of confusion and misinformation, hence my subject title.
The issue here is whether or not Rising Tide backported other's contributions from their GPL version which would be a violation of the GPL.
Except RTS is in full agreement that they are using modules containing code committed by third parties, including those that Andy has pointed out. They have not disputed this at all. What is in dispute is whether combining their proprietary module with the GPL kernel and other GPL modules and is a GPL violation. They are pushing the boundary of the "using APIs is not a derivative work" argument to it's limits.
First off this is nothing like the Oracle case. That was a case about reimplementing APIs, and has nothing to do with linking against someone-else's code that provides APIs. Secondly, it is a stretch to say that the RTS SCSI target is just including APIs. It is using all sorts of internal kernel functions that go far beyond what most reasonable programmers would consider to be an API to the kernel. If you interpret things that liberally, then any proprietary modifications to a GPL application would be allowed by just bundling up the list of functions you happen to use and calling it an API.
If you read the list, it is clear that no one is disputing the facts of what happened; no one is making unfounded assumptions. Rising Tide Systems has a SCSI module that they have written entirely from scratch. In the process of writing this they pushed much of it into the kernel, so much in fact that one of their employees became the Linux SCSI targets maintainer. They have kept some of it back and are shipping a modified kernel containing their code to customers without providing the source code.
They believe that this is allowed, and that their code is not a derivative work of the kernel. Basically RTS is saying: If NVIDIA can have proprietary drivers, why can't we have proprietary kernel subsystems? The other side believes that what NVIDIA is doing likely is a GPL violation, and furthermore some of the technicalities that NVIDIA claims make it legal don't apply to what RTS is doing.
This issue has been ticking time bomb ready to go off. It is entirely possible that some, all, or none of the proprietary drivers written are a violation of the GPL. It all depends on the courts interpretation of derivative work, and no one knows for certain (although some arguments have stronger precedent than others). Furthermore, it is too late to add GPL linking exception to Linux's license to clarify this (one way or the other), because there are too contributors at this point to come to an agreement. So it will remain murky as mud until someone finally sues someone over the issue. Sounds like this may actually be happening. But even then, the ruling may end up depending on the nature of the proprietary extension, and thus remain fuzzy.
I wonder if this means that some others who left in the recent past, like J Allard or Ray Ozzie will be coming back. The rumors were that Sinofsky vigorously opposed their plans, and they left after Balmer decided to back Sinofsky's way rather than them.
The vast majority of murders today don't have any eye witnesses, and yet many of those cases still get solved and the perpetrator caught. Furthermore, it really isn't hard to sneak out behind someone and shoot someone today, without even being seen by the victim. So an invisibility cloak will only make it slightly easier to kill someone, and won't make it any harder to catch them. Not much of a game changer.
How does the winner take all approach we have now NOT suppress the minority vote?
I'm not saying that. I'm saying your suggestion has the exact same flaws as the current system, and won't make people feel any less disenfranchised.
So the person that gets more popular votes and electoral votes wins the election
No that's not what happens. For example, suppose that in state A each of 20 districts is 45% Republican and 55% Democrat so all 20 electoral votes go to Democrats. Then in state B each of 15 districts is 65% Republican and 35% Democrat, so all 15 votes go to the Republicans. Then the Democrat wins the electoral vote by 20 to 15, even though the Republican had 54% of the popular vote. This is despite the fact that the electoral votes in this example were even distributed purely based on population (each district is the exact same size).
Splitting large elections into a series of smaller winner takes all elections does not give the same result as popular vote, because it suppresses the minority vote in each district. And that's before you get into bigger practical problems like gerrymandering. If you want popular vote then go with direct popular vote. If you want to maintain the current handicap for smaller states that the electoral congress provides (I think it's good), then require all states to allocate their electoral votes in proportion to the total vote within the entire state. Don't split things up into even smaller winner-takes-all elections.
In case you were tempted to RTFA, don't. You have to click through two ad-laden pages, and there really isn't any more information than in the summary.
I won't speak for Tesla, but SpaceX does not depend on government money. The Falcon 1 was created entirely with private funding, which includes capital investments to build their entire vertically integrated production facilities (they don't contract anything), some launch facilities, and design, construction, and multiple test flights of an entirely new design of rocket. The Falcon 9 was mostly NASA funding, but it built heavily on the Falcon 1 design, and was thus less expensive to design and test than the Falcon 1 (even without including the huge facilities investments mentioned before). Furthermore, SpaceX already had financing to develop Falcon 9 when they won the NASA contract. The contract allowed them to divert that money into the Dragon Capsule instead, the majority of which is thus privately funded.
So without government funding, they would be about where they are with Falcon 1/9, but just getting started with Dragon. Government money sped them up a bit, but they aren't even close to being dependent on that funding.