It is similar to the Transputer architecture. But which millions of videogame consoles was Sony selling with a Transputer running it? That's the bizmodel. Which makes all the difference, validating the Transputer architecture that is now suitable to the media processing demands of the mass market.
MP3 encoding at a server isn't trivial load for a thousand simul streams on a P4. Your 3 minutes per 45min album is only 15x for about $1000, while a 1TFLOPS GPU card might encode 16 thousand times for $300.
There are many more people coding multistream MP3 servers, but still no port to GPGPU.
Video servers follow the same logic. But video decoders at the client will get better economics from many thousands/millions of ASICs in the mass market, rather than the few thousand servers a year that the market will carry. Unless we're talking about GPUs in 2008-9 which decode HD to "entertainment PCs".
PCI-Express offers 64 "lanes" pumping up to 500MBps each (since January, 250MBps in actual shipping HW). In a switched hub, for 256Gbps total. The Cell's EIB is probably its most interesting feature: 200Gbps token ring that transparently connects offchip. So the new IBM Cells, with 4 cores (Power970 + 8-SPEs each) on one die (or SoC) has 32x 25.6GFLOPS + 4x 970 all moving at 200Gbps. Or just a single Cell at 204GFLOPS feeds 200Gbps to a PCIe stuffed with 20x 10Gig ethernet cards (10 double-10GigE PCIe cards).
The single Cell therefore offers 32 (bits) * 8 (SPEs) = 256 FLOPs per (5 picosecond) loop on a full pipeline. The four-way offers 1KFLOPs. There are 1024-SPE Cells in the product line, which are 32KFLOPs; a 4-way would offer 128KFLOPs per loop. Even complex video codecs and mixing need at most 100MIPS, which such a beast would run at 5ns, or 200 million times realtime. 200Gbps is 5 thousand simul Blu-Ray video streams, so we're talking about the beast working at 40 thousand times its max video throughput, while a single Cell works at 8 times video throughput. Audio throughput is much lower FLOPS:bit.
So the Cell data transfer is certainly ample to its high processing speed.
The Playstation 3 is reported to harness 2 TFLOPS. But "only" 204GFLOPS run on the Cell CPU, 10%. The other 1.8TFLOPS runs on the nVidia G70 GPU. But the G70 runs shaders, very limited application to anything but actually rendering graphics.
The Cell itself is notoriously hard to code for. If just some extra effort can target the nVidia, that's TWO TeraFLOPS in a $500 box. A huge leap past both AMD and Intel.
it's hard to program such GPUs for anything other than graphics applications.
"Anything other" is "general purpose", which they cover at GPGPU.org. But the general community of global developers hasn't gotten hooked on the cheap performance yet. Maybe if someone got an MP3 encoder working on one of these hot new chips, the more general purpose programmers would be delivering supercomputing to the desktop on these chips.
Well, sure, those companies claim their product works. I haven't ever seen any evidence that they do work. Their testimonials don't support claims of surviving transcoding, and are entirely consistent with a watermark that works only on identical copies of the watermarked content.
Since I've actually seen transcoding kill watermarks, and the theory of the exploit is so simple, I will believe they survive transcoding only when I see some real evidence. Or maybe at least a credible endorser of that specific feature - the MPAA and VPs of movie studios don't count.
I'm not really "scared" of watermarks, because I don't violate copyright, and would like to see real copyrights cheaply and accurately enforced. I'm scared of BS pretending to protect copyright that just incriminates not guilty (by fair use right or just false evidence) people. Or props up totally unfair and unsustainable copyright grabs by pretending "surgical strikes" on important violations will actually work. This tech seems to be just more of the collapsable DRM apparatus the copyright industry is building instead of a real infrastructure for managing the minimum essential rights into a healthy market.
Lossless media files are still too big for most people to swap them across the Internet, so exact watermarks protect only a tiny fraction of the files that copyright owners care about.
MP3 is lossy, and there's lots of different MP3 data that sounds close enough to the original MP3 that a song can be transcoded to new data that sounds "the same" within the tolerance for noise that all MP3 listening demands.
Won't this watermark scheme just fall victim to the first revision of an MP3 (or MP4 video, etc) encoder that includes a "scramble" option?
Do you mean AT&T and Verizon? It took decades to split AT&T up, and it's taken AT&T about as long to merge back together within the new laws. Laws that were passed in a much less corporate/monopoly friendly environment.
Much more practical to say "if the Internet were transmitted entirely over distributed phased-array software radios, Net Neutrality will become moot". In the meantime, AT&T and Verizon will dominate all Internet and dependent economic/media/political/etc activity.
The bulk of your post is theory contradicted by actual measurements like the one to which I linked, showing increased bandwidth is a better solution for the backbones themselves than is QoS.
The last bit saying that new legislation isn't required is also theory, and also wrong. AT&T's Whitacre clearly disagrees with you, as he repeats often in public that he wants to charge Google extra. And of course Vonage, and you too if you start a VoIP competitor on his wires. AT&T is a "legal monopoly", working within the maze of regulations they've created since the 1986 divestiture. And if they advertise "QoS" requirements, that's not false advertising.
Take AT&T at its word that it will doublecharge. But don't take its word that it must. On that, look to the computer scientists who have proven he hasn't.
And then realize that the only thing standing between Whitacre and your wallet is your government. But only if you make your government stand there, which it does not today. Otherwise, it's AT&T's government by default, standing between you and your Internet.
Studies of actual traffic congestion mitigation techniques have consistently demonstrated that increasing capacity is a much cheaper and more reliable remedy than QoS on backbones. With extra benefits in the raw capacity. The "quality traffic to quality networks" would require a whole extra architectural layer to route through several different Internet links on realtime route quality decisions, rather than leverage the full capacity of the Net to route anywhere at any time on local congestion conditions or other overall strategies.
These whines are in fact "special consideration" pressure for the telcos to get "Net Doublecharge". They don't need service tiers, but they can use them to demand distant endpoints pay protection money. If they can get the protection money from the government, their favorite source of subsidy and protection for over a century, they certainly will. Especially if they've already used up the capacity for private accounts (people) to pay them directly, which makes them look less competitive.
People learn to act creative. Some of us do manage to learn how to do it, usually alone. It's hard to learn to be creative in a competitive environment. When I was in public school in the 1970s-80s, the curriculum included special training for the "smart kids" in brainstorming. Practicing builds those skills. Not just being creative oneself, but much more of the activity in the group is helping others create ideas without stepping on them. Offering an idea is risky, because it's often wrong, which can discredit you, costing you social power, if even one or a few others in the group are competing.
So while this study might show that "meetings make you dumber", or that we act dumber in meetings, that doesn't have to be true. If we don't just accept it, we can learn to do a lot better. And since "meetings" just means "collaboration with other people", we have a lot better we can do. Better than we can alone, if we get it right.
A funny letter snarking a bank for bouncing a check against a pensioner now insisting banks deal with her own new defensive bureaucracy circulates the Net in an email claiming to be from an old pensioner. It reportedly was written by an Australian columnist as humor. But practically everyone can relate. And now, with our PCs, I hope to see everyone actually apply the policies and procedures the letter mentions.
I dunno, Massachusetts seems to be doing pretty well. Despite people like you calling a car collision without saving the passenger "murder". But then, you voted to send the serial/mass murderer cokehead corporate theocrat to the White House twice, and look at how many problems that's created for Ted Kennedy to make go away.
"Free market libertarian" wants to let the AT&T/Verizon(/Qwest) cartel sell themselves access at puny wholesale rates, but competitors should pay prohibitively inflated retail? Similarly huge-scale competitors like the other 2 of the 3 can equalize roaming charges, but small competitors will never afford to get access?
Libertarians are people who believe in the minimum government possible, but no less. Free marketeers are people who believe that markets work best with the fewest barriers to competition. Free market libertarians recognize that people create governments as our only remedy to monopoly market abuse.
The term you're searching for to describe yourself is "monopolist corporate greedhead".
You can minimize the stand you took apparently without being able to back it up. I'm not surprised you got cornered by some "mob", though I really don't care. Meanwhile, I just asked you to crawl out from behind your keyboard and back up your pose. But now you're just typing back at me, without backing it up with anything real. Just like your performance in this whole thread.
So you like to project your fear at coming out from behind your keyboard on me. And you like to think that my easily showing you up to be a windbag coward is somehow me dancing for you. You're a sad, sad typer. When you want to come to NYC and back up your violent words with a real backbone, try it. Until then, you can stay in the little box you've made for yourself.
So you don't want back up your words, even when you're just "agreeing with the other guy"?
Pussy. Especially with your weird little dance fantasies. Who cares about your bragging? You think you get some kind of credit for getting arrested? Pussy.
This distribution deal is a move that is good for the Internet media revolution. For one, it establishes a real competitor to YouTube with content people will actually want to watch. Without bundling that content with the network: Joost is just the distributor, owning no rights.
But more importantly, it puts copyrighted content into a YouTube competitor that can challenge YouTube if YouTube has the content. That means that YouTube's copyright enforcement doesn't happen in an vacuum of arbitrary claims and baseless decisions. When Joost complains, it will have a copy of the content and a copy of the contract with the content owner. The process to enforce copyright between the two corporations can take place in the well understood realm of corporate negotiations and lawsuits.
Of course, it would be better for everyone (including Viacom, and especially YouTube and Joost) if copyrights didn't slow down every media transaction. But until copyrights actually are peeled back to a legitimate scope, duration and enforcement regime, getting competitors with paper trails to manage it is the best we can do, and better than nothing.
The CDDB was coded as a free repository of CD metadata. Collected by thousands of people around the Net on a worldwide, ongoing basis, by giving away the client SW which many programmers embedded into PC/Mac music players. So millions of people were prompted every time they put in an unknown CD to spend a few seconds typing in artist and song names. In exchange (though no input was required), they got most of their CDs labeled without any effort, after the CDDB was filled.
This kind of read/write database population collaboration is now well known, both in blogs and in more sophisticated databases like Wikipedia. But in the late 1990s it was revolutionary.
Then the CDDB server owners sold out to Gracenote. Gracenote required a login to access the data, which login they supplied only to licensed users. Gracenote first tried to sell CD players integrated with the CDDB, but then found more success in licensing access to iTunes and other online music distributors.
But neither Gracenote nor the CDDB programmers had produced the profitable data. The people who had were locked out. So some new programmers made a new version with the identical API and DB structure, the FreeDB, then datamined the CDDB to populate it. The FreeDB and its contents are GPL, so they cannot be "taken proprietary" (stolen) again. The data is free again, as is the life of this pioneering colalborative project.
If you are generating music metadata, consider submitting it to the FreeDB. And try to use the FreeDB, rather than the privateer CDDB, to support you applications. And send money to the FreeDB operators whenever you can, especially if you use it.
Slashdot Mirror
It is similar to the Transputer architecture. But which millions of videogame consoles was Sony selling with a Transputer running it? That's the bizmodel. Which makes all the difference, validating the Transputer architecture that is now suitable to the media processing demands of the mass market.
MP3 encoding at a server isn't trivial load for a thousand simul streams on a P4. Your 3 minutes per 45min album is only 15x for about $1000, while a 1TFLOPS GPU card might encode 16 thousand times for $300.
There are many more people coding multistream MP3 servers, but still no port to GPGPU.
Video servers follow the same logic. But video decoders at the client will get better economics from many thousands/millions of ASICs in the mass market, rather than the few thousand servers a year that the market will carry. Unless we're talking about GPUs in 2008-9 which decode HD to "entertainment PCs".
PCI-Express offers 64 "lanes" pumping up to 500MBps each (since January, 250MBps in actual shipping HW). In a switched hub, for 256Gbps total. The Cell's EIB is probably its most interesting feature: 200Gbps token ring that transparently connects offchip. So the new IBM Cells, with 4 cores (Power970 + 8-SPEs each) on one die (or SoC) has 32x 25.6GFLOPS + 4x 970 all moving at 200Gbps. Or just a single Cell at 204GFLOPS feeds 200Gbps to a PCIe stuffed with 20x 10Gig ethernet cards (10 double-10GigE PCIe cards).
The single Cell therefore offers 32 (bits) * 8 (SPEs) = 256 FLOPs per (5 picosecond) loop on a full pipeline. The four-way offers 1KFLOPs. There are 1024-SPE Cells in the product line, which are 32KFLOPs; a 4-way would offer 128KFLOPs per loop. Even complex video codecs and mixing need at most 100MIPS, which such a beast would run at 5ns, or 200 million times realtime. 200Gbps is 5 thousand simul Blu-Ray video streams, so we're talking about the beast working at 40 thousand times its max video throughput, while a single Cell works at 8 times video throughput. Audio throughput is much lower FLOPS:bit.
So the Cell data transfer is certainly ample to its high processing speed.
The Playstation 3 is reported to harness 2 TFLOPS. But "only" 204GFLOPS run on the Cell CPU, 10%. The other 1.8TFLOPS runs on the nVidia G70 GPU. But the G70 runs shaders, very limited application to anything but actually rendering graphics.
The Cell itself is notoriously hard to code for. If just some extra effort can target the nVidia, that's TWO TeraFLOPS in a $500 box. A huge leap past both AMD and Intel.
"Anything other" is "general purpose", which they cover at GPGPU.org. But the general community of global developers hasn't gotten hooked on the cheap performance yet. Maybe if someone got an MP3 encoder working on one of these hot new chips, the more general purpose programmers would be delivering supercomputing to the desktop on these chips.
Well, sure, those companies claim their product works. I haven't ever seen any evidence that they do work. Their testimonials don't support claims of surviving transcoding, and are entirely consistent with a watermark that works only on identical copies of the watermarked content.
Since I've actually seen transcoding kill watermarks, and the theory of the exploit is so simple, I will believe they survive transcoding only when I see some real evidence. Or maybe at least a credible endorser of that specific feature - the MPAA and VPs of movie studios don't count.
I'm not really "scared" of watermarks, because I don't violate copyright, and would like to see real copyrights cheaply and accurately enforced. I'm scared of BS pretending to protect copyright that just incriminates not guilty (by fair use right or just false evidence) people. Or props up totally unfair and unsustainable copyright grabs by pretending "surgical strikes" on important violations will actually work. This tech seems to be just more of the collapsable DRM apparatus the copyright industry is building instead of a real infrastructure for managing the minimum essential rights into a healthy market.
I ain't buyin' it.
Lossless media files are still too big for most people to swap them across the Internet, so exact watermarks protect only a tiny fraction of the files that copyright owners care about.
MP3 is lossy, and there's lots of different MP3 data that sounds close enough to the original MP3 that a song can be transcoded to new data that sounds "the same" within the tolerance for noise that all MP3 listening demands.
Won't this watermark scheme just fall victim to the first revision of an MP3 (or MP4 video, etc) encoder that includes a "scramble" option?
"The Internet companies"? Like Google?
Do you mean AT&T and Verizon? It took decades to split AT&T up, and it's taken AT&T about as long to merge back together within the new laws. Laws that were passed in a much less corporate/monopoly friendly environment.
Much more practical to say "if the Internet were transmitted entirely over distributed phased-array software radios, Net Neutrality will become moot". In the meantime, AT&T and Verizon will dominate all Internet and dependent economic/media/political/etc activity.
I wiped you out, dirty hippie monopoly lover.
Without Net Neutrality, the duopoly will vanquish us.
Whoops - the link to a QoS/capacity study that I included in my post didn't make it to the Slashdot page. Should have previewed.
There are more.
Whoops - the link to a QoS/capacity study that I included in my post didn't make it to the Slashdot page. Should have previewed.
The bulk of your post is theory contradicted by actual measurements like the one to which I linked, showing increased bandwidth is a better solution for the backbones themselves than is QoS.
The last bit saying that new legislation isn't required is also theory, and also wrong. AT&T's Whitacre clearly disagrees with you, as he repeats often in public that he wants to charge Google extra. And of course Vonage, and you too if you start a VoIP competitor on his wires. AT&T is a "legal monopoly", working within the maze of regulations they've created since the 1986 divestiture. And if they advertise "QoS" requirements, that's not false advertising.
Take AT&T at its word that it will doublecharge. But don't take its word that it must. On that, look to the computer scientists who have proven he hasn't.
And then realize that the only thing standing between Whitacre and your wallet is your government. But only if you make your government stand there, which it does not today. Otherwise, it's AT&T's government by default, standing between you and your Internet.
Studies of actual traffic congestion mitigation techniques have consistently demonstrated that increasing capacity is a much cheaper and more reliable remedy than QoS on backbones. With extra benefits in the raw capacity. The "quality traffic to quality networks" would require a whole extra architectural layer to route through several different Internet links on realtime route quality decisions, rather than leverage the full capacity of the Net to route anywhere at any time on local congestion conditions or other overall strategies.
These whines are in fact "special consideration" pressure for the telcos to get "Net Doublecharge". They don't need service tiers, but they can use them to demand distant endpoints pay protection money. If they can get the protection money from the government, their favorite source of subsidy and protection for over a century, they certainly will. Especially if they've already used up the capacity for private accounts (people) to pay them directly, which makes them look less competitive.
People learn to act creative. Some of us do manage to learn how to do it, usually alone. It's hard to learn to be creative in a competitive environment. When I was in public school in the 1970s-80s, the curriculum included special training for the "smart kids" in brainstorming. Practicing builds those skills. Not just being creative oneself, but much more of the activity in the group is helping others create ideas without stepping on them. Offering an idea is risky, because it's often wrong, which can discredit you, costing you social power, if even one or a few others in the group are competing.
So while this study might show that "meetings make you dumber", or that we act dumber in meetings, that doesn't have to be true. If we don't just accept it, we can learn to do a lot better. And since "meetings" just means "collaboration with other people", we have a lot better we can do. Better than we can alone, if we get it right.
A funny letter snarking a bank for bouncing a check against a pensioner now insisting banks deal with her own new defensive bureaucracy circulates the Net in an email claiming to be from an old pensioner. It reportedly was written by an Australian columnist as humor. But practically everyone can relate. And now, with our PCs, I hope to see everyone actually apply the policies and procedures the letter mentions.
Who did you vote to be president of presidents in 2000?
I dunno, Massachusetts seems to be doing pretty well. Despite people like you calling a car collision without saving the passenger "murder". But then, you voted to send the serial/mass murderer cokehead corporate theocrat to the White House twice, and look at how many problems that's created for Ted Kennedy to make go away.
"Free market libertarian" wants to let the AT&T/Verizon(/Qwest) cartel sell themselves access at puny wholesale rates, but competitors should pay prohibitively inflated retail? Similarly huge-scale competitors like the other 2 of the 3 can equalize roaming charges, but small competitors will never afford to get access?
Libertarians are people who believe in the minimum government possible, but no less. Free marketeers are people who believe that markets work best with the fewest barriers to competition. Free market libertarians recognize that people create governments as our only remedy to monopoly market abuse.
The term you're searching for to describe yourself is "monopolist corporate greedhead".
You can minimize the stand you took apparently without being able to back it up. I'm not surprised you got cornered by some "mob", though I really don't care. Meanwhile, I just asked you to crawl out from behind your keyboard and back up your pose. But now you're just typing back at me, without backing it up with anything real. Just like your performance in this whole thread.
So you like to project your fear at coming out from behind your keyboard on me. And you like to think that my easily showing you up to be a windbag coward is somehow me dancing for you. You're a sad, sad typer. When you want to come to NYC and back up your violent words with a real backbone, try it. Until then, you can stay in the little box you've made for yourself.
So you don't want back up your words, even when you're just "agreeing with the other guy"?
Pussy. Especially with your weird little dance fantasies. Who cares about your bragging? You think you get some kind of credit for getting arrested? Pussy.
Fascist sockpuppet, how about you deliver some of that social justice here in NYC sometime? Or aren't you just another pisspants rightwinger coward?
This distribution deal is a move that is good for the Internet media revolution. For one, it establishes a real competitor to YouTube with content people will actually want to watch. Without bundling that content with the network: Joost is just the distributor, owning no rights.
But more importantly, it puts copyrighted content into a YouTube competitor that can challenge YouTube if YouTube has the content. That means that YouTube's copyright enforcement doesn't happen in an vacuum of arbitrary claims and baseless decisions. When Joost complains, it will have a copy of the content and a copy of the contract with the content owner. The process to enforce copyright between the two corporations can take place in the well understood realm of corporate negotiations and lawsuits.
Of course, it would be better for everyone (including Viacom, and especially YouTube and Joost) if copyrights didn't slow down every media transaction. But until copyrights actually are peeled back to a legitimate scope, duration and enforcement regime, getting competitors with paper trails to manage it is the best we can do, and better than nothing.
A good client will take multiple entries and merge them, best by "voting".
The CDDB was coded as a free repository of CD metadata. Collected by thousands of people around the Net on a worldwide, ongoing basis, by giving away the client SW which many programmers embedded into PC/Mac music players. So millions of people were prompted every time they put in an unknown CD to spend a few seconds typing in artist and song names. In exchange (though no input was required), they got most of their CDs labeled without any effort, after the CDDB was filled.
This kind of read/write database population collaboration is now well known, both in blogs and in more sophisticated databases like Wikipedia. But in the late 1990s it was revolutionary.
Then the CDDB server owners sold out to Gracenote. Gracenote required a login to access the data, which login they supplied only to licensed users. Gracenote first tried to sell CD players integrated with the CDDB, but then found more success in licensing access to iTunes and other online music distributors.
But neither Gracenote nor the CDDB programmers had produced the profitable data. The people who had were locked out. So some new programmers made a new version with the identical API and DB structure, the FreeDB, then datamined the CDDB to populate it. The FreeDB and its contents are GPL, so they cannot be "taken proprietary" (stolen) again. The data is free again, as is the life of this pioneering colalborative project.
If you are generating music metadata, consider submitting it to the FreeDB. And try to use the FreeDB, rather than the privateer CDDB, to support you applications. And send money to the FreeDB operators whenever you can, especially if you use it.