You could do the equivalent of storing PAR2 style blocks in the image, try to decode the image, and assuming the data had sufficient redundancy, the error correction algorithm would be able to recover the original file.
I'm not talking about cropping. Resizing is more damage than that. Consider an image N pixels wide. Resize to N-1 pixels - each of which is produced by a bicubic interpolation of the original data. That's enough to change each value. You could also mess with the colour palette a bit, e.g. reorder it to further reduce image size or exchange a few entries. Since png is compressed even a one bit change at the start of the image should result in the rest of the file being totally changed. My scheme should produce a stream of bits that was completely different, even though the image would look identical.
A well chosen way of hiding the data with enough redundancy could even survive being recompressed as JPEG, assuming your not using a ridiculously high compression level.
You can certainly put data into an image that can survive JPEG compression or my scheme. Doing that while keeping the data invisible seems harder.
Yeah like the banks! Instead of adapting their business model they hired men with guns and put up bars on the windows. Get with the 20th Century, dudes. The advent of the tommy gun and the Model T ford mean that a business model based on hogging money was clearly obsolete as obsolete as the buggy whip manufacturers'. They should have given their money to entrepreneurs like Mr Capone and instead made money out selling services like tommy gun hire and washing the blood stains off spats.
It seems like you could easily get rid of data hidden in images - just extract the image content and dump any metadata, resize it to one pixel less in each dimension and reencode with a PNG optimizer.
The odd thing is that I never even noticed the stop and refresh buttons - I just used Esc and F5. Still I find when I move from Opera to IE8 or FF the UI just seems completely wrong and the browser is much slower so I move back as soon as I can.
I had a Windows 2000 machine at work until very recently. You can still install the latest Opera or Firefox on Windows 2000, even now. Actually I suspect most Windows code started off supporting Win2k and up and hasn't changed enough to break on Win2k even now..Net 3.5 framework doesn't support Win2k, so that would be an issue for application compatibility.
E.g., someone who's played a lot of WoW or is into classic mythology is a lot more likely to see half-human/half-animal figures, like, say, centaurs, without it involving any kind of alienation.
If they weren't alienated from society, why were they playing WoW?
Published in 1961 and set in 1955, this psychodrama of an ambitiously named development in Connecticut (the source of Yatesâ(TM)s superbly misleading title) recalls us to the period that saw the publication of David Riesman's The Lonely Crowd (1950), Sloan Wilson's novel The Man in the Gray Flannel Suit (1955), the pop sociology of men like William H. Whyte and Vance Packard, whose critiques The Organization Man (1956) and The Hidden Persuaders (1957) made American business seem impersonal and cynical, and-if this isn't too fanciful-Edward Albee's Who's Afraid of Virginia Woolf? and Malvina Reynolds's song "Little Boxes," both of which made their debut in 1962. Pete Seeger had a huge success of his own with the song, which ridiculed the harmless citizens of Daly City, California, and gave us the word ticky-tacky. No less a man than Tom Lehrer was to say that it was "the most sanctimonious song ever written,"
That makes it ok to download NIN and Fleet Foxes, in the unlikely event that Reznor and Pecknold are the sole copyright holders for the music they encourage people to download. It doesn't make it ok to download other artists.
Actually it's pretty unlikely that Reznor and Pecknold are the sole copyright holders - bands often sell copyright to a label and even if they didn't copyright would be shared among all past and present band members. Reznor can't give out a free copyright license anymore than Linus Torvalds can unilaterally change the license of the Linux kernel.
Yeah, I know that even something that only involves two countries could be called "international", just like you could create a "internet" containing only two networks.
Or a World Series where teams from two countries compete.
factor in how much of your tax dollars when into that and then get back to us with a valid point....
That's pretty easy to do. The USPS publishes annual reports just like a company. According to the 2008 annual report, the USPS took in about $75B in revenue, had about $78B in operating costs, and had a contribution of about $3B from the US government. That's about 3% of its costs covered by your taxes.
What's your point again?
And they still had enough cash free to run a website that says they don't depend on tax money
Actually I've read Fortran was faster than C back in the 80's because it did not suffer from pointer aliasing and the like. Basically it was a simpler language and thus easier for a simple compiler to turn into efficient assembler. I've also see a surreal page where Basic code was translated into SSE assembler. Obviously the mapping from Basic arithmetic to SSE is pretty trivial to do. That's probably true of Fortran or Cobol but it defintely isn't true of C or C++ once the compiler has to worry about pointer aliasing. A lot of C programmers tend to concentrate on writing easy to read code and let the compiler turn it into something efficient too, it's a very different mindset from people who told the compiler exactly what sequence of operations they wanted because optimizers weren't practical.
That being said if you look at the output of a modern C compiler it is very very good. Still back in the days when more primitive languages were popular, I think that was not the case.
Actually I don't know why people complain so much about Cobol. There are much fouler propietary languages around in any company that start off as quick and dirty hacks and have gradually been extended and changed to work around the initial flaws and limitations. Most of those are seriously horrible to go in and change because the only documentation is the source code for the buggy tool that parses them.
In A.D. 2101 War was beginning. Captain: What happen ? Mechanic: Somebody set up us the bomb. Operator: We get signal. Captain: What ! Operator: Main screen turn on. Captain: It's You !! CAT5: How are you gentlemen !! CAT5: All your 1000BASE-T are belong to us. CAT5: You are on the way to destruction.
Actually on a lot of phones the ring tone is an MP3. So when someone calls you you need to be able to decode MP3s and handle the network stack. More to the point when the phone is tested by operators they will have a base station simulator attached, so even if the signalling code can manage to keep the connection despite being sabotaged by the application that plays MP3s if it is outside spec it will fail the test.
> It's not irrelevant. The point of the EKA2 kernel is that it guarantees (as far as is possible, etc. but more than a standard Linux or Window Mobile kernel) that the baseband OS will not be affected by the applications the app OS is running. It's designed to avoid the latency problems that would lead to the bugs you list.
Real time OSs are not a panacea for this sort of thing. The application (or more likely a driver) can disable interrupts and wait in a loop. That will kill any real time stuff on the same processor. Even if you have two CPUs if you get the priorities wrong in the external bus interface the application processor could starve out the signalling processor by hogging the bus. I've actually seen signalling code by both of these. I've also seen someone at a customer type in code that turns off interrupts and waits for a second - I pointed out this will cause chaos and he solved the problem another way.
> And not to nit-pick, but in your example, if you switch off the core thats running the GUI, what happens to the display? Does the phone just go blank? How does it wake up when a key is pressed, etc.
The display on a mobile phone has local memory - you DMA stuff from the CPU to the display and then it refreshes itself. On a clamshell phone this is important because the secondary display can be on even when the CPU isn't.
Well it's sort of irrelevant now. Most phone chipsets have two ARM cores - one for the signalling and one for the application.
And as someone put it if you're developing a phone do you really want to deal with bugs like "when I play this Britney Spears MP3 my phone drops calls" or worse "phone fails radio test at the testhouse, seems to depend which application is running but we can't figure out how".
If you put both the applications and the signalling stack on the same ARM you're pretty much asking for this. I'd much rather have say a small ARM9 core and a beefier ARM11 for the application.
The ARM9 doesn't take up much space and you can give it priority access to external flash/sdram and try to run parts of the radio stack as possible from tightly coupled memory, i.e. on die SRAM. That makes it more or less a different machine and minimizes the chance of application code sabotaging the radio. Plus you can run a tiny OS kernel designed for network stacks on the ARM9 and Linux/WinMo/Android or whatever on the ARM11.
And when the phone is in not running applications but needs to stay connected to the network you can shutdown the power hungry application core and the flash, run the ARM9 at a low clock frequency, put the SDRAM in self refresh and have the network stack do its thing mostly low power TCM and cache.
Slashdot Mirror
You could do the equivalent of storing PAR2 style blocks in the image, try to decode the image, and assuming the data had sufficient redundancy, the error correction algorithm would be able to recover the original file.
I'm not talking about cropping. Resizing is more damage than that. Consider an image N pixels wide. Resize to N-1 pixels - each of which is produced by a bicubic interpolation of the original data. That's enough to change each value. You could also mess with the colour palette a bit, e.g. reorder it to further reduce image size or exchange a few entries. Since png is compressed even a one bit change at the start of the image should result in the rest of the file being totally changed. My scheme should produce a stream of bits that was completely different, even though the image would look identical.
A well chosen way of hiding the data with enough redundancy could even survive being recompressed as JPEG, assuming your not using a ridiculously high compression level.
You can certainly put data into an image that can survive JPEG compression or my scheme. Doing that while keeping the data invisible seems harder.
I agree. He should have said "it took us a long time to secure Windows, who knows what security flaws Chrome OS 1.0 will have"
Paypal my fee to the usual account Mr Ballmer.
Yeah like the banks! Instead of adapting their business model they hired men with guns and put up bars on the windows. Get with the 20th Century, dudes. The advent of the tommy gun and the Model T ford mean that a business model based on hogging money was clearly obsolete as obsolete as the buggy whip manufacturers'. They should have given their money to entrepreneurs like Mr Capone and instead made money out selling services like tommy gun hire and washing the blood stains off spats.
It seems like you could easily get rid of data hidden in images - just extract the image content and dump any metadata, resize it to one pixel less in each dimension and reencode with a PNG optimizer.
What cool-aid are you smoking?
I think you mean "What cool aid are you drinking", a reference to the Jonestown cult mass suicide =
http://en.wikipedia.org/wiki/Jim_Jones#Deaths_in_Jonestown
Or maybe "What are you smoking?", a reference to irrational behaviour brought on by smoking crack cocaine.
Still I'm going to have to fine you $150 for using a mixed metaphor. Type carefully in future.
The odd thing is that I never even noticed the stop and refresh buttons - I just used Esc and F5. Still I find when I move from Opera to IE8 or FF the UI just seems completely wrong and the browser is much slower so I move back as soon as I can.
Oh snap.
I had a Windows 2000 machine at work until very recently. You can still install the latest Opera or Firefox on Windows 2000, even now. Actually I suspect most Windows code started off supporting Win2k and up and hasn't changed enough to break on Win2k even now. .Net 3.5 framework doesn't support Win2k, so that would be an issue for application compatibility.
E.g., someone who's played a lot of WoW or is into classic mythology is a lot more likely to see half-human/half-animal figures, like, say, centaurs, without it involving any kind of alienation.
If they weren't alienated from society, why were they playing WoW?
That's the only use I can think of for a laptop.
Oh and number #1 is a laughing killer in a wolf mask.
WHO ARE YOU CALLING A PSYCHO?
http://www.theatlantic.com/doc/200812/hitchens-suburbs
Published in 1961 and set in 1955, this psychodrama of an ambitiously named development in Connecticut (the source of Yatesâ(TM)s superbly misleading title) recalls us to the period that saw the publication of David Riesman's The Lonely Crowd (1950), Sloan Wilson's novel The Man in the Gray Flannel Suit (1955), the pop sociology of men like William H. Whyte and Vance Packard, whose critiques The Organization Man (1956) and The Hidden Persuaders (1957) made American business seem impersonal and cynical, and-if this isn't too fanciful-Edward Albee's Who's Afraid of Virginia Woolf? and Malvina Reynolds's song "Little Boxes," both of which made their debut in 1962. Pete Seeger had a huge success of his own with the song, which ridiculed the harmless citizens of Daly City, California, and gave us the word ticky-tacky. No less a man than Tom Lehrer was to say that it was "the most sanctimonious song ever written,"
Oh snap!
That makes it ok to download NIN and Fleet Foxes, in the unlikely event that Reznor and Pecknold are the sole copyright holders for the music they encourage people to download. It doesn't make it ok to download other artists.
Actually it's pretty unlikely that Reznor and Pecknold are the sole copyright holders - bands often sell copyright to a label and even if they didn't copyright would be shared among all past and present band members. Reznor can't give out a free copyright license anymore than Linus Torvalds can unilaterally change the license of the Linux kernel.
Yeah, I know that even something that only involves two countries could be called "international", just like you could create a "internet" containing only two networks.
Or a World Series where teams from two countries compete.
What do you really want to accomplish by behaving like this?
factor in how much of your tax dollars when into that and then get back to us with a valid point....
That's pretty easy to do. The USPS publishes annual reports just like a company. According to the 2008 annual report, the USPS took in about $75B in revenue, had about $78B in operating costs, and had a contribution of about $3B from the US government. That's about 3% of its costs covered by your taxes.
What's your point again?
And they still had enough cash free to run a website that says they don't depend on tax money
http://www.usps.com/communications/newsroom/postalfacts.htm
0 Tax dollars received for operating the Postal Service
I bet you really enjoyed this movie
http://www.imdb.com/title/tt0119925/
Actually I've read Fortran was faster than C back in the 80's because it did not suffer from pointer aliasing and the like. Basically it was a simpler language and thus easier for a simple compiler to turn into efficient assembler. I've also see a surreal page where Basic code was translated into SSE assembler. Obviously the mapping from Basic arithmetic to SSE is pretty trivial to do. That's probably true of Fortran or Cobol but it defintely isn't true of C or C++ once the compiler has to worry about pointer aliasing. A lot of C programmers tend to concentrate on writing easy to read code and let the compiler turn it into something efficient too, it's a very different mindset from people who told the compiler exactly what sequence of operations they wanted because optimizers weren't practical.
That being said if you look at the output of a modern C compiler it is very very good. Still back in the days when more primitive languages were popular, I think that was not the case.
Actually I don't know why people complain so much about Cobol. There are much fouler propietary languages around in any company that start off as quick and dirty hacks and have gradually been extended and changed to work around the initial flaws and limitations. Most of those are seriously horrible to go in and change because the only documentation is the source code for the buggy tool that parses them.
Slick and creative eh? I bet your one of those types who doesn't wear a suit conforming to DOCNR XQ3451/82.
Or to troll the crap out of people. Trolling leads to page views and more comments.
In A.D. 2101
War was beginning.
Captain: What happen ?
Mechanic: Somebody set up us the bomb.
Operator: We get signal.
Captain: What !
Operator: Main screen turn on.
Captain: It's You !!
CAT5: How are you gentlemen !!
CAT5: All your 1000BASE-T are belong to us.
CAT5: You are on the way to destruction.
So very sorry.
Actually on a lot of phones the ring tone is an MP3. So when someone calls you you need to be able to decode MP3s and handle the network stack. More to the point when the phone is tested by operators they will have a base station simulator attached, so even if the signalling code can manage to keep the connection despite being sabotaged by the application that plays MP3s if it is outside spec it will fail the test.
> It's not irrelevant. The point of the EKA2 kernel is that it guarantees (as far as is possible, etc. but more than a standard Linux or Window Mobile kernel) that the baseband OS will not be affected by the applications the app OS is running. It's designed to avoid the latency problems that would lead to the bugs you list.
Real time OSs are not a panacea for this sort of thing. The application (or more likely a driver) can disable interrupts and wait in a loop. That will kill any real time stuff on the same processor. Even if you have two CPUs if you get the priorities wrong in the external bus interface the application processor could starve out the signalling processor by hogging the bus. I've actually seen signalling code by both of these. I've also seen someone at a customer type in code that turns off interrupts and waits for a second - I pointed out this will cause chaos and he solved the problem another way.
> And not to nit-pick, but in your example, if you switch off the core thats running the GUI, what happens to the display? Does the phone just go blank? How does it wake up when a key is pressed, etc.
The display on a mobile phone has local memory - you DMA stuff from the CPU to the display and then it refreshes itself. On a clamshell phone this is important because the secondary display can be on even when the CPU isn't.
Well it's sort of irrelevant now. Most phone chipsets have two ARM cores - one for the signalling and one for the application.
And as someone put it if you're developing a phone do you really want to deal with bugs like "when I play this Britney Spears MP3 my phone drops calls" or worse "phone fails radio test at the testhouse, seems to depend which application is running but we can't figure out how".
If you put both the applications and the signalling stack on the same ARM you're pretty much asking for this. I'd much rather have say a small ARM9 core and a beefier ARM11 for the application.
The ARM9 doesn't take up much space and you can give it priority access to external flash/sdram and try to run parts of the radio stack as possible from tightly coupled memory, i.e. on die SRAM. That makes it more or less a different machine and minimizes the chance of application code sabotaging the radio. Plus you can run a tiny OS kernel designed for network stacks on the ARM9 and Linux/WinMo/Android or whatever on the ARM11.
And when the phone is in not running applications but needs to stay connected to the network you can shutdown the power hungry application core and the flash, run the ARM9 at a low clock frequency, put the SDRAM in self refresh and have the network stack do its thing mostly low power TCM and cache.