Real world trials generally don't use placebos. Instead, the control is the best treatment currently available. After all, you're usually not interested in whether some drug works better than sugar, but whether it's worth using it instead of or in addition to some currently available treatment. So the control group get the best treatment they could have got if they were not in the study, while the experimental group may or may not fare better.
The needs of the many outweigh the needs of the one. Getting clean data for evidence-based-medicine is absolutely necessary for the well-being of possibly thousands or millions of potential patients.
We already have an unpatented, royalty-free, unencumbered, lowest-common-denominator video codec for use on the internet: H.261.
H.323 specifies it as the lowest common denominator for video-over-IP, so all video phones already support it, including hardware implementations. It was published in 1990 - twenty years ago - so it is as patent-free as you can get. And it's published by the ITU, so the specification is freely available.
If the return type of the main function is a type compatible with int, a return from the initial call to the main function is equivalent to calling the exit function with the value returned by the main function as its argument;10) reaching the } that terminates the main function returns a value of 0.
Reaching the end of main without a return is the equivalent of calling exit(0); as the last line.
Why, that's a nice gift horse, mind if I take a look at it's teeth?
Laptops with 3G aren't new, but now that Apple is (maybe) planning on adding a 3G capable radio chipset (to add to the bluetooth and wifi 2.4/5GHz radio chipset), it's just bringing "more problems"?
Well, if you don't want to use it, that's fine, but I would be happy with a laptop which is able to talk to the world outwith the confines of my home or work wifi networks.
Does anyone find it amusing that after all the ridicule the French heaped on Americans for electing Bush that they went and elected somebody even worse?
I know, cutting people off from the internet is so much worse than torture.
producing larger files than most alternatives (MPEG4, for instance) for the same level of quality
What exactly do you mean by MPEG 4? Simple Profile? Advanced Simple Profile? Advanced Video Coding? Any other strange and wonderful aspects of MPEG 4 Visual? Maybe you mean MPEG 4 part 22, which can achieve very high resolution with a very compact representation?
You have failed to point out where Apple are lying. I have demonstrated repeatedly that colour resolution is not necessarily the same as full resolution, in many places where we take it for granted. You have yet to show me where Apple have said their screens are "x by y pixels, where each individual pixel is capable of producing 2^24 colours". They give a pixel resolution. They also give a colour depth. Where do they say there is a one-to-one mapping between colour resolution and pixels? Sure they don't say the displays use a 6-bit DAC, but when have they ever claimed to have an 8-bit DAC? It is possible to produce millions of colours on these 6-bit displays by dithering, so their claim of producing "millions of colours", is absolutely correct.
colors which are not red, green, or blue are necessarily produced by a combination of the primary colors, a tristimulus display can produce any color that is a combination of those primary shades, for any reasonable definition of the word "produce".
You're confusing colour perception with colour reproduction. It is true that to perceive cyan there must be at least two cones stimulated, but there are an infinite number of spectra consisiting of different wavelengths which can add up to produce the same colour of cyan. If this were not so, display colour calibration would be impossible.
Further it is not necessary to build a display from the red, green and blue primaries. One could just as easily build a display out of the primaries cyan, magenta and yellow, since these are just filters on a white spectrum. In this case cyan can be produced by a single subpixel while blue would probably need a whole pixel.
If your method of dithering requires using multiple pixels, then your effective resolution is lower and thus claiming full resolution and the ability to produce that many colors is a LIE. You CANNOT do both; you can do one or the other. You can -either- have full resolution with a lower color count, OR you can display more colors at the expense of image data. Just as it would be a lie to claim the total number of subpixels as your resolution, then claim 16 million colors at that resolution.
Where have Apple stated "x by y pixels, where each individual pixel is capable of producing 2^24 colours"? Show me where and I will agree with you. However, Apple have never made such a claim. They give a specific pixel resolution, and a colour depth. There is not a one-to-one mapping between pixels (rgb-triplets) and colours, and Apple have never claimed there is. Would you dispute that a video encoded in YUV4:2:0 is not HD? You don't actually get individually addressable pixel colours; each colour refers to a 2x2 block of brightness pixels. And yet there is no-one suing Sony for Blue-ray video not being 1920x1080 pixels.
Look up the Bayer pattern, and find out how resolution is measured. It is equally valid to count individual sensor elements (each capable of red, green or blue), or 2x2 blocks of red, two greens and a blue. The reason this is valid is because the human eye is more sensitive to resolution in brightness than in colour. The output pixels are dithered to produced a full brightness-resolution image. This is why Foveon get away with claiming triple the resolution of their individual sensor elements.
If dithering is not allowed, then neither is YUV, Bayer patterns, or any method of picture compression which exploits the limitations of human perception. Let's sue the joint photographics expect group, since the JPEG format doesn't store individual pixels - my camera claims to produce pictures of 3072x2304 pixels, but not only is it captured from a Bayer pattern, but the colour planes are subsampled and individual pixels are not even stored, just an approximation of a 2d discrete cosine transform. It is about as far from 3072x2304 pixels as you can get. Shall I sue Sony for making a 7 megapixel camera which uses a Bayer pattern, or will I happily accept that not having to account for individual colours at the full resolution is what gives me those 7 million pixels in the first place?
My point is that colours which are not pure red, green or blue can only be produced by combinations of subpixels, as a tristimulus display can only produce shades of these three colours. We are in agreement about this.
Further, since the above, it is necessary to dither subpixels to get anything close to "millions" of colours (or even anything higher than my original hyperbolic claim of 766 colours). I believe we are in agreement about this.
Finally, since dithering between subpixels is a necessary artifact of colour reproduction, arguing about whether one method of dithering is "right" and produces "true 16 million colours" and another is "wrong" and produces "only 200 thousand colours" is pointless. An LCD screen could easily use a Bayer pattern and be no less "wrong" about its colour production capabilities.
I did not say Nikon are lying about their resolution, I was pointing out that the rgb-triplet method is not the only valid way of dithering to produce more than the very limited range of colours which n-bit RGB can produce or capture.
In conclusion: dithering between arbitrary subpixels is a valid method of colour reproduction, and it is not false advertising to state full resolution when an arbitrary block of subpixels is required to produce an arbitrary colour.
you're arguing that cyan is not a color and in fact RGB are the ONLY colors
No, I'm arguing that RGB are in fact the only colours produced by a tristimulus-based display. The eye combines the separate wavelengths with a weighted response to produce something that we in our own perceptually encumbered heads call "cyan".
You seem to believe that a single group of rgb subpixels is a distict entity. In fact most monitors are laid out in the pattern RGBRGBRGBRGB A colour cyan would be displayed by turning off some of the subpixels _GB_GB_GB_GB Now, there is absolutely no difference in a cyan pixel produced by the group of pixels _GB and the group of pixels GB_ or even the group of pixels B_G
The distinction of where one pixel ends and another begins is arbitrary, and so to claim that an 8-bit display is capable of displaying 16 million colors at its native resolution is wrong because the native resolution is not 1920x1080 pixels, but 5760x1080 subpixels, and it is irrelevant which source subpixels make up a single visual entity which we call "cyan".
To both claim 16million colors and full resolution is a blatant lie
Then sue Nikon for using Bayer patterns and stating full (interpolated) resolution.
Let's start with, it's multiplicative, not additive. That's 255^3, not 255*3. This is because, as you mentition later, the eye combines all three subpixels into a new color.
You only get to multiply colours together if you take multiple subpixels to form one colour. You can do that on a rgb-triplet basis, or any arbitrary block of red, green and blue subpixels. cf Bayer pattern.
So, even if one were to concede all your points, these aren't really 1920x1280x24 displays are they then. Because that 1920x1280 resolution has to get shortchanged for the dithering. So you can say that Apple lied about the resolution instead of the color if you like, but it's awful pedantic.
Since the human eye is less sensitive to colour resolution, the fact that colour is dithered is irrelevant - you still get full brightness resolution - or you do dispute that 1080p YUV422 is actually 1920x1080 pixels?
A 1920x1280x24 screen was advertised and not delivered.
This is simply not true. Apple have always said "millions of colours" and never gone into detail as to how those colours are arrived at.
Not quite. The human eye is less sensitive to colour resolution than colour depth. It is more sensitive to brightness (both resolution and depth - as detected by all three cones and rods) than colour (detected by individual cones).
For example, 1080p is HDTV. But only the luma plane is 1920x1080 pixels; colour information is (usually) transmitted or stored at 1/4 the resolution of the luma plane. Even broadcast-quality 10-bit-per-channel HD-SDI is YUV 4:2:2, where the chroma is at half the vertical resolution. You don't go suing Sky because they don't have full colour resolution, do you?
It is a very sensible engineering tradeoff to reduce colour resolution, since the human eye is not very good at telling the difference.
I did not say that three 8-bit subpixels produce 766 permutations - clearly there are ~16 million. I said there are 766 colours, and that it is the combination of multiple subpixels to produce the illusion of more than that. Dithering is simply the process of producing more colours using more than three subpixels.
Agreed. It will be good when Apple finally release ultra-high-dpi displays. When you get to around 300ppi, dithering really does become indistinguisable.
Indeed with high enough resolution, each subpixel can be 1-bit.
A pixel on an LCD monitor emits three different colours; one from each constituent subpixel. The pixel filters out a set of wavelengths from a white light source to produce a triplet of subpixels, each with a single, even, colour.
Just an FYI, these kinds of trolls work better if you save the obvious nonsense for later rather than starting right off the bat.
Ad hominem. And I thought everyone was reading Paul Graham nowadays.
What, precisely, is incorrect about the following statement: A single pixel on an LCD screen is made of three subpixels, each of which can be lit at a particular brightness; the number of distinct brightnesses is dependant on the width of the DAC.
What the hell is the complaint about? Even a screen with an 8-bit DAC is only capable of displaying 766 colours - each subpixel can show 255 brightnesses of three distinct wavelengths of light (as each subpixel can show the same black this makes 766, not 768). And if you want to get really picky, you can only display three colours - a flourescent backlit display does not emit light like a blackbody, it has a particular spectrum which is filtered by one of three filters. No matter the brightness, each red subpixel displays the same spectrum.
So why the claims of millions of colours? Because the eye dithers. Light from all three subpixels land on cosited cones on the retina, and the optic nerve processes this weighted tristimulus response so that the brain perceives the equivalent of a particular wavelength.
So a single pixel can appear to produce 16 million colours by being made up of three different coloured subpixels. In some rendering situations, subpixels can be individually lit. This all works because the eye has very poor resolution for colours. This is also why video is invariable encoded in a YUV colour-space.
Whether an individual subpixel can display 256 levels is quite irrelevant since dithering is capable of producing a higher colour depth at the expense of colour resolution. You still get full brightness resolution. And this is ok, because its not really possible to tell the difference.
What next, suing Nikon for daring to include Bayer filters on their CCDs? Yes, it is possible to build CCDs where the R, G and B are cosited, nobody actually uses the Foveon sensor because the difference in the capture picture is not discernable.
This whole thing is stupid. It sounds like people nitpicking advertising, without actually being aware of the technical concepts involved the image display process.
Blurrydisc is still a stupid name. HD DVD is clear - it's a High Definition DVD, and people know what a DVD is. A blurry disc? What's High def about that?
the rather stupid limit of around 3GB of memory in 32 bit Vista.
superfetch actually makes sense when so little memory can be adressed
Actually, no. There is a limit of 3GB address-space for applications (actually it's 2GB unless you boot with a switch to reduce the amount of kernel address-space, which you really shouldn't be doing unless you know what you're doing).
Windows Vista, Windows XP, Linux, Mac OS X et al support PAE, which allows 32-bit Intel CPUs to support a 36-bit address bus, or up to 64GB of physical memory. This physical memory is shared between each processes virtual address space. Since there is more physical than virtual address space, there is less contention for RAM and less swapping. Of course, applications in the know can indirectly address extended memory, and manipulate its virtual address space, but that is so horrible its easier just to spawn a new process.
Superfetch is basically a page-cache to a fast disk. It can't affect the virtual address space, so it does nothing for the 3GB-per-process limit. It allegedly helps with Windows' stupid ideas about the working set, and the fact that Windows will page out applications that you're actively using.
a very poor way of handling what is in resident memory
I agree with you there. The fact that Windows can take several minutes to redraw the desktop, in 2007, is absurd.
Slashdot Mirror
Real world trials generally don't use placebos. Instead, the control is the best treatment currently available. After all, you're usually not interested in whether some drug works better than sugar, but whether it's worth using it instead of or in addition to some currently available treatment. So the control group get the best treatment they could have got if they were not in the study, while the experimental group may or may not fare better.
The needs of the many outweigh the needs of the one. Getting clean data for evidence-based-medicine is absolutely necessary for the well-being of possibly thousands or millions of potential patients.
We already have an unpatented, royalty-free, unencumbered, lowest-common-denominator video codec for use on the internet: H.261.
H.323 specifies it as the lowest common denominator for video-over-IP, so all video phones already support it, including hardware implementations. It was published in 1990 - twenty years ago - so it is as patent-free as you can get. And it's published by the ITU, so the specification is freely available.
Um, no. Main is special.
Reaching the end of main without a return is the equivalent of calling exit(0); as the last line.
Why, that's a nice gift horse, mind if I take a look at it's teeth?
Laptops with 3G aren't new, but now that Apple is (maybe) planning on adding a 3G capable radio chipset (to add to the bluetooth and wifi 2.4/5GHz radio chipset), it's just bringing "more problems"?
Well, if you don't want to use it, that's fine, but I would be happy with a laptop which is able to talk to the world outwith the confines of my home or work wifi networks.
I know, cutting people off from the internet is so much worse than torture.
Nonsense. I only ever download between 64kB and 4MB of any film at a time.
Fixed that for you. Or were you asleep when Steve wrote his Open Letter?
2) RFID doesn't require power.
What exactly do you mean by MPEG 4? Simple Profile? Advanced Simple Profile? Advanced Video Coding? Any other strange and wonderful aspects of MPEG 4 Visual? Maybe you mean MPEG 4 part 22, which can achieve very high resolution with a very compact representation?
iPhone + Project Gutenberg. I can read thousands of classics, wherever I am, on the best looking screen I've found. And with no DRM.
You have failed to point out where Apple are lying. I have demonstrated repeatedly that colour resolution is not necessarily the same as full resolution, in many places where we take it for granted. You have yet to show me where Apple have said their screens are "x by y pixels, where each individual pixel is capable of producing 2^24 colours". They give a pixel resolution. They also give a colour depth. Where do they say there is a one-to-one mapping between colour resolution and pixels? Sure they don't say the displays use a 6-bit DAC, but when have they ever claimed to have an 8-bit DAC? It is possible to produce millions of colours on these 6-bit displays by dithering, so their claim of producing "millions of colours", is absolutely correct.
Show me where Apple are lying.
You're confusing colour perception with colour reproduction. It is true that to perceive cyan there must be at least two cones stimulated, but there are an infinite number of spectra consisiting of different wavelengths which can add up to produce the same colour of cyan. If this were not so, display colour calibration would be impossible.
Further it is not necessary to build a display from the red, green and blue primaries. One could just as easily build a display out of the primaries cyan, magenta and yellow, since these are just filters on a white spectrum. In this case cyan can be produced by a single subpixel while blue would probably need a whole pixel.
Where have Apple stated "x by y pixels, where each individual pixel is capable of producing 2^24 colours"? Show me where and I will agree with you. However, Apple have never made such a claim. They give a specific pixel resolution, and a colour depth. There is not a one-to-one mapping between pixels (rgb-triplets) and colours, and Apple have never claimed there is. Would you dispute that a video encoded in YUV4:2:0 is not HD? You don't actually get individually addressable pixel colours; each colour refers to a 2x2 block of brightness pixels. And yet there is no-one suing Sony for Blue-ray video not being 1920x1080 pixels.
Look up the Bayer pattern, and find out how resolution is measured. It is equally valid to count individual sensor elements (each capable of red, green or blue), or 2x2 blocks of red, two greens and a blue. The reason this is valid is because the human eye is more sensitive to resolution in brightness than in colour. The output pixels are dithered to produced a full brightness-resolution image. This is why Foveon get away with claiming triple the resolution of their individual sensor elements.
If dithering is not allowed, then neither is YUV, Bayer patterns, or any method of picture compression which exploits the limitations of human perception. Let's sue the joint photographics expect group, since the JPEG format doesn't store individual pixels - my camera claims to produce pictures of 3072x2304 pixels, but not only is it captured from a Bayer pattern, but the colour planes are subsampled and individual pixels are not even stored, just an approximation of a 2d discrete cosine transform. It is about as far from 3072x2304 pixels as you can get. Shall I sue Sony for making a 7 megapixel camera which uses a Bayer pattern, or will I happily accept that not having to account for individual colours at the full resolution is what gives me those 7 million pixels in the first place?
My point is that colours which are not pure red, green or blue can only be produced by combinations of subpixels, as a tristimulus display can only produce shades of these three colours. We are in agreement about this.
Further, since the above, it is necessary to dither subpixels to get anything close to "millions" of colours (or even anything higher than my original hyperbolic claim of 766 colours). I believe we are in agreement about this.
Finally, since dithering between subpixels is a necessary artifact of colour reproduction, arguing about whether one method of dithering is "right" and produces "true 16 million colours" and another is "wrong" and produces "only 200 thousand colours" is pointless. An LCD screen could easily use a Bayer pattern and be no less "wrong" about its colour production capabilities.
I did not say Nikon are lying about their resolution, I was pointing out that the rgb-triplet method is not the only valid way of dithering to produce more than the very limited range of colours which n-bit RGB can produce or capture.
In conclusion: dithering between arbitrary subpixels is a valid method of colour reproduction, and it is not false advertising to state full resolution when an arbitrary block of subpixels is required to produce an arbitrary colour.
No, I'm arguing that RGB are in fact the only colours produced by a tristimulus-based display. The eye combines the separate wavelengths with a weighted response to produce something that we in our own perceptually encumbered heads call "cyan".
You seem to believe that a single group of rgb subpixels is a distict entity. In fact most monitors are laid out in the pattern
RGBRGBRGBRGB
A colour cyan would be displayed by turning off some of the subpixels
_GB_GB_GB_GB
Now, there is absolutely no difference in a cyan pixel produced by the group of pixels
_GB
and the group of pixels
GB_
or even the group of pixels
B_G
The distinction of where one pixel ends and another begins is arbitrary, and so to claim that an 8-bit display is capable of displaying 16 million colors at its native resolution is wrong because the native resolution is not 1920x1080 pixels, but 5760x1080 subpixels, and it is irrelevant which source subpixels make up a single visual entity which we call "cyan".
Then sue Nikon for using Bayer patterns and stating full (interpolated) resolution.
You only get to multiply colours together if you take multiple subpixels to form one colour. You can do that on a rgb-triplet basis, or any arbitrary block of red, green and blue subpixels. cf Bayer pattern.
Since the human eye is less sensitive to colour resolution, the fact that colour is dithered is irrelevant - you still get full brightness resolution - or you do dispute that 1080p YUV422 is actually 1920x1080 pixels?
This is simply not true. Apple have always said "millions of colours" and never gone into detail as to how those colours are arrived at.
Not quite. The human eye is less sensitive to colour resolution than colour depth. It is more sensitive to brightness (both resolution and depth - as detected by all three cones and rods) than colour (detected by individual cones).
For example, 1080p is HDTV. But only the luma plane is 1920x1080 pixels; colour information is (usually) transmitted or stored at 1/4 the resolution of the luma plane. Even broadcast-quality 10-bit-per-channel HD-SDI is YUV 4:2:2, where the chroma is at half the vertical resolution. You don't go suing Sky because they don't have full colour resolution, do you?
It is a very sensible engineering tradeoff to reduce colour resolution, since the human eye is not very good at telling the difference.
Ad hominem. Attack the argument, not the person.
Thank you, you seem to be the only person on this entire thread to get my point. Perhaps I should have explained myself better?
Agreed. It will be good when Apple finally release ultra-high-dpi displays. When you get to around 300ppi, dithering really does become indistinguisable.
Indeed with high enough resolution, each subpixel can be 1-bit.
A pixel on an LCD monitor emits three different colours; one from each constituent subpixel. The pixel filters out a set of wavelengths from a white light source to produce a triplet of subpixels, each with a single, even, colour.
Ad hominem. And I thought everyone was reading Paul Graham nowadays.
What, precisely, is incorrect about the following statement:
A single pixel on an LCD screen is made of three subpixels, each of which can be lit at a particular brightness; the number of distinct brightnesses is dependant on the width of the DAC.
What the hell is the complaint about? Even a screen with an 8-bit DAC is only capable of displaying 766 colours - each subpixel can show 255 brightnesses of three distinct wavelengths of light (as each subpixel can show the same black this makes 766, not 768). And if you want to get really picky, you can only display three colours - a flourescent backlit display does not emit light like a blackbody, it has a particular spectrum which is filtered by one of three filters. No matter the brightness, each red subpixel displays the same spectrum.
So why the claims of millions of colours? Because the eye dithers. Light from all three subpixels land on cosited cones on the retina, and the optic nerve processes this weighted tristimulus response so that the brain perceives the equivalent of a particular wavelength.
So a single pixel can appear to produce 16 million colours by being made up of three different coloured subpixels. In some rendering situations, subpixels can be individually lit. This all works because the eye has very poor resolution for colours. This is also why video is invariable encoded in a YUV colour-space.
Whether an individual subpixel can display 256 levels is quite irrelevant since dithering is capable of producing a higher colour depth at the expense of colour resolution. You still get full brightness resolution. And this is ok, because its not really possible to tell the difference.
What next, suing Nikon for daring to include Bayer filters on their CCDs? Yes, it is possible to build CCDs where the R, G and B are cosited, nobody actually uses the Foveon sensor because the difference in the capture picture is not discernable.
This whole thing is stupid. It sounds like people nitpicking advertising, without actually being aware of the technical concepts involved the image display process.
Blurrydisc is still a stupid name. HD DVD is clear - it's a High Definition DVD, and people know what a DVD is. A blurry disc? What's High def about that?
Actually, no. There is a limit of 3GB address-space for applications (actually it's 2GB unless you boot with a switch to reduce the amount of kernel address-space, which you really shouldn't be doing unless you know what you're doing).
Windows Vista, Windows XP, Linux, Mac OS X et al support PAE, which allows 32-bit Intel CPUs to support a 36-bit address bus, or up to 64GB of physical memory. This physical memory is shared between each processes virtual address space. Since there is more physical than virtual address space, there is less contention for RAM and less swapping. Of course, applications in the know can indirectly address extended memory, and manipulate its virtual address space, but that is so horrible its easier just to spawn a new process.
Superfetch is basically a page-cache to a fast disk. It can't affect the virtual address space, so it does nothing for the 3GB-per-process limit. It allegedly helps with Windows' stupid ideas about the working set, and the fact that Windows will page out applications that you're actively using.
I agree with you there. The fact that Windows can take several minutes to redraw the desktop, in 2007, is absurd.