for the most part, the human eye can't tell the difference between one random function and another. Ture, this is very lossy, but it looks the same.
This idea has been used in computer graphics by substituting portions of images with "detail" textures which are essentially the high frequency components of a part of the image overlayed with the lower frequency components. The lower frequency components compress very well with convential techniques and the higher frequency components look identitical at all scales.
So, as far is the human eye is concerned, white noise adds very little information to the picture and it can be thrown away and replaced with software generated white noise.
I was thinking about a real-time version of this. Recall how "live" programs on TV and radio are cencored by using a 7 second delay? What if we did the same thing for commercials. Someone else watches your programs and sends live information on the internet ragarding when a commercial starts and ends. Most of the work could be done by looking for the blank screen.
When a commercial comes on, your TV automatically mutes or switches to another channel.
If a ISP did that you could launch some very nasty attacks against their network by sending them back bad data.
Well, you could have other customers verify each other. The same problem occurs in networked games, people can forge calculation so client-side calculations have to verified by the central server or by another client. If the client base is large enough then it's difficult to do collaborative "cheating".
The other problem occurs is if the ISP's data is confidential AND the customer has to have access to it to process it. I did some work on this last year ago using "crypographic computing" to solve this problem. Basically you use the ideas of permutation and rotation to encrypt both the executing program and the required data. The client machine runs a VM that is capable of executing the encrypted instructions. Actually each instruction becomes a tree of instructions that calculate the same result. Trees from neighboring instructions can be woven together as long as data-dependencies are observed. Because you are using encrypted data, operations like "add/mul" won't work in their native form. They would have to decrypt 2 numbers before they could be multiplied together. This would give away the data, so instead you have to develop a cryptographic add/multiply function than can work with numbers in their encrypted form.
Advantages: - the client doesn't know what he/she is executing - the client doesn't know what data he/she is processing - both the program and data can have MACs to prevent tampering of either.
Disadvantages: - Memory usage can grow exponentially with the number of instructions executed - Encryption / Decryption of program data is much more expensive than processing itself. Decryption time can grow by the number of instructions executed, because some operation cannot operate on encrypted data.
Although it's an interesting academic problem, it has little real world use because of speed and memory issues. But, it has lead me to my current project which is very useful.
Furthermore, why do they feel the need to make a movie/show on something that MIGHT happen? It's only speculation. Yes, I know that's what most movies are made of, but I can't see this having a good effect on the generally uneducated populace.
Reminds me of a little radio show called War of the Worlds.:) Some people did freak out and do weird stuff, but you have to admit that is what made it cool. I see no difference here. Scare a few people who don't know any better and it's fun for the whole family.
Perlin was working on something like this at NYU, do a search for his name and you'll find it. IIR, his idea was to use something like a treadmill that would go in any direction.
I take that back.. Too my knowedge, no one in the black market has broken these things yet. If someone has let me know.;) I was talking to the guys from cryptography.com who break smart cards for a living, and they were saying several of the new sat decoder systems are very secure.
Note that there's really no such thing as "tamper-proof" hardware.
heh. It's impossible, for the near future, to physically extract significant information from the human brain (such as a PIN or password). Some work has been done in biological computing to use the "brains" of smaller insects to store information - because it shows promise to being 100% tamper resistant. Though, nothing working has any practical use yet - as far as I'm aware.
But actually what the newer sat decoder boxes do is have a slow smart card that decrypts a new key every second or so. The smart card is "tamper-resistant" to a large degree and passes the key to the decoder hardware which is not tamper-resistant.
No one has broken the smart cards yet... but that's not to say it can't be done with a lot of money. Getting a key for 1 second doesn't do much for you. However, if you can buffer your transmission data and delay it for a few seconds, you can have someone else continously send you the decoded keys over the internet. Having to send data continously makes the operation risker because there is a way to trace you... though you can send "almost untraceable" data by using ICMP to ping-with-data to a remote host and a forged return address. With all that money flying around, I'm sure someone will give it a try.;)
I really hate telephone spam because I sleep wierd hours and they wake me up. Seems like everyone is using auto-dialers these days. These things are easily detectable, just say hello and if no one responds in 1.5 seconds then it's an autodialer. I was thinking about setting up a modem that would say an automated "hello" and listen for silence. If it wasn't an auto-dialer then it would beep me and tell the person to hang on a second. Is there an established API under linux for doing full-duplex sound on a modem? I don't mind buying a specific brand of modem.
This means ISPs who have been lazy about closing their spam relay holes will have to take the RBL seriously now. If you are running a server and want to make sure you don't have any holes that will put you on the RBL telnet to mail-abuse.org
Internet Explorer, having to survive on its own without the rest of a big corporation to subsidize it will have to charge money and will lose market share to the open source Mozilla project which should be in good shape by the time the ruling comes down.
IE doesn't have to charge money. It could still rake in tons of money by being a hot portal site. Millions of people never change the default homepage. I'm surprised RedHat hasn't started doing this for NS under linux... maybe they have some sort of agreement.
I've been using 3 21" ViewSonic monitors for about 1.5 years. After that I developed an eye stigmatism. The refresh rates on monitors is much more visiable out of the corner of your eye (try it). Somehow my left eyelid would spaz out at random interval even when I wasn't look at at moniters. I started turning off the side monitors when I'm not looking at them, which is most of the time. About 2-3 months later my eye problem stopped.
This probably won't happen with LCD monitors, but be careful! If you are using normal monitors put them at the highest possible refresh rate they will do.
You are correct, except that mp3 was meant to be a fixed-bandwidth streaming format. Thus regardless of the sound playing there is fixed amount of bandwidth allocated to play it (128kb in this case). Playing sin(x) obviously doesn't require that much bandwidth - but it's still allocated 128kb, while more complex suffer when they exceed this rate. Some encoders will dynamically switch encoding rates depending on the complexity of the sound, which can save you a good bit of space.
I think we are going to see a new generation of music encoders that can do several times better than mp3. There is much self-similarity that is not exploited. I have an album called Andy Warhul - Ah yes, Ah no. In this album andy plays back 2 audio clips one saying "ah yes" and one saying "ah no." He uses the same clips with no pitch change over and over thousands of times (it's quite boring - but somehow interesting). Compressing such a stream should result in a file 100K or less, yet it takes 30MB. This is an extreme example where self-similarity (fractals if you want to call it that) could be be used to compress music.
Most music has very repetive patterns that can be exploited. Mp3 is designed for movies not music. There hasn't been much need for better music compression until recently because no one makes money off of it. Now that people are, I think we will see better algorithms replace mp3, and it's not MS's format.
It looks like the took apache and spiffied it up. Some of nicer new features (besides SSL) are:
" Remote Configuration: a browser-based configuration tool to allo[w manipulation of the server configuration via a GUI.
Machine Translation Support: This new function, when used with an available IBM Machine Translation Engine, enables the IBM HTTP Server to translate English Web pages into other languages without human intervention. This permits a Web site visitor to read the page in his native language, effectively broadening the reach of your Web site. IBM Machine Translation Engines are included in the WebSphere Application Server 3.0 and include: German, Simplified Chinese and Traditional Chinese. Additional languages will be available in the future. "
Is it just me or does Corel Draw crash every 5 minutes (windows version)? I'm afraid to use it, because I lose my work every time. Even saving after every operation doesn't work because somehow it manages to corrupt it's files. I've upgraded all the patches they have on their web site, but they seem to have no effect on it's stability.
Corel draw would be a wonderful program if it was more stable, but in it's current state it's very unusable. Anyone have suggestions for a more stable vector based drawing program?
NetCraft counts over four million web servers running Apache.
I assume they count servers by IP address not by domain names right? I wonder how accurate this information is since a popular way to do virtual hosting is to assign multiple IP addresses to a single host. If there are a small number of people assigning whole sub-domains to a single machine these numbers could be an order of magnitude off.
That page might load faster, but it doesn't have a document.f.q.focus(); like google's front page does. you have to hit tab 3 times or click on the field (yuck!)
Wierd how I never noticed that. That is a bit of a pain.
as for a start page.. i use a page on my local computer
I used to do that, but I use so many computers I kept losing the file. I should set one up on my own homepage, but....sigh... I'm so lazy.:)
Most VC firms worth anything should be willing to do that as they regularily deal with proprietary or secret information.
From my experience, most VC firms won't sign a NDA (non disclosure agreement) because they get so many business plans many of them overlap and they would have a conflict of interest. They are more likely to sign a NDA if you have a proven track record, but then you probably don't need money.
If it's a really good idea, people will steal it no matter what you do. The advantage you have is that you have a working proto-type and hopefully not to far off from production, where other people will be playing catch-up. Patents now take 18 months to go through, so they are not worth much in a startup environment. Better just to file a sloppy patent yourself and slap a patent-pending on your product to scare people off. If you go to lawyer you are looking at $20k for a good application.
I suggest leaving out the really technical details from your business plan and wait till you find someone who sounds really interested before disclosing the rest.
Thought I mention I have a page with pictures and brief bios of some of my favorite CS heros. (Turing, Shannon, Huffman, Whitefield, Miyamoto, RMS, and Ken Perlin). Click on my sig then [heros].
In a worst-case scenario the proliferation of such devices will discourage young people from exploring the opportunities of teaching as a career, and may decrease our chances of.....
Computers are better than humans at chess, but that hasn't stopped people from playing or getting into it. If anything, I think it has helped because now people can always find a challenging player/teacher.
CDDB the online CD database claims "still the world's largest CD database with over 390,000 titles and 4,500,000 audio tracks".
Estimate an audio track to be on average 3MB, and you are looking at 12 terabytes of music right there. From my experiences CDDB has pretty good coverage of english music, but it's lacking some foreign titles. So add 10-20% more to the estimate. They are currently working on a version with international character sets, so it might be a lot higher if they don't have any Asain titles. Also you add maybe another few TBs for new bands and old bands that are not available in CD form.
I wonder how many years it will be before 16TB is easily affordable? Less than 10 if moore's law holds for storage. hmm.. it would take you ~64 years to listen to it all though. course there is very little of that which you actually *want* to listen to. that's where group filtering comes in.
Anyhow, my prediction is that within 10 years an ordinary person will have a complete collection of the world's published music in their home. Legal issues aside, I think this is pretty exciting.
All this time I thought people were joking when they put a (tm) next to "Good Thing". I just did a trademark search, and guess what. The mark is owned by Martha Steward! Why am I not surprised?
Word Mark GOOD THING Owner Name (REGISTRANT) MARTHA STEWART LIVING OMNIMEDIA LLC Owner Address 20 West 43rd Street, 25th Floor New York NEW YORK 10036 LIMITED LIABILITY COMPANY DELAWARE Attorney of Record HOWARD J SHIRE Serial Number 75-516347 Registration Number 2272142 Filing Date 07/09/1998 Registration Date 08/24/1999 Mark Drawing Code (1) TYPED DRAWING Register PRINCIPAL Published for Opposition 06/01/1999 Type of Mark SERVICE MARK
Slashdot Mirror
White noise is extremely compressable :
pixel = random();
for the most part, the human eye can't tell the difference between one random function and another. Ture, this is very lossy, but it looks the same.
This idea has been used in computer graphics by substituting portions of images with "detail" textures which are essentially the high frequency components of a part of the image overlayed with the lower frequency components. The lower frequency components compress very well with convential techniques and the higher frequency components look identitical at all scales.
So, as far is the human eye is concerned, white noise adds very little information to the picture and it can be thrown away and replaced with software generated white noise.
I don't think these arguments hold up if you compare with the car analogy. Claiming your product is unsafe does not exempt you from the law.
I was thinking about a real-time version of this. Recall how "live" programs on TV and radio are cencored by using a 7 second delay? What if we did the same thing for commercials. Someone else watches your programs and sends live information on the internet ragarding when a commercial starts and ends. Most of the work could be done by looking for the blank screen.
When a commercial comes on, your TV automatically mutes or switches to another channel.
What is ICrave ?
If a ISP did that you could launch some very nasty attacks against their network by sending them back bad data.
Well, you could have other customers verify each other. The same problem occurs in networked games, people can forge calculation so client-side calculations have to verified by the central server or by another client. If the client base is large enough then it's difficult to do collaborative "cheating".
The other problem occurs is if the ISP's data is confidential AND the customer has to have access to it to process it. I did some work on this last year ago using "crypographic computing" to solve this problem. Basically you use the ideas of permutation and rotation to encrypt both the executing program and the required data. The client machine runs a VM that is capable of executing the encrypted instructions. Actually each instruction becomes a tree of instructions that calculate the same result. Trees from neighboring instructions can be woven together as long as data-dependencies are observed. Because you are using encrypted data, operations like "add/mul" won't work in their native form. They would have to decrypt 2 numbers before they could be multiplied together. This would give away the data, so instead you have to develop a cryptographic add/multiply function than can work with numbers in their encrypted form.
Advantages:
- the client doesn't know what he/she is executing
- the client doesn't know what data he/she is processing
- both the program and data can have MACs to prevent tampering of either.
Disadvantages:
- Memory usage can grow exponentially with the number of instructions executed
- Encryption / Decryption of program data is much more expensive than processing itself. Decryption time can grow by the number of instructions executed, because some operation cannot operate on encrypted data.
Although it's an interesting academic problem, it has little real world use because of speed and memory issues. But, it has lead me to my current project which is very useful.
Furthermore, why do they feel the need to make a movie/show on something that MIGHT happen? It's only speculation. Yes, I know that's what most movies are made of, but I can't see this having a good effect on the generally uneducated populace.
:) Some people did freak out and do weird stuff, but you have to admit that is what made it cool. I see no difference here. Scare a few people who don't know any better and it's fun for the whole family.
Reminds me of a little radio show called War of the Worlds.
Perlin was working on something like this at NYU, do a search for his name and you'll find it. IIR, his idea was to use something like a treadmill that would go in any direction.
No one has broken the smart cards yet.
;) I was talking to the guys from cryptography.com who break smart cards for a living, and they were saying several of the new sat decoder systems are very secure.
I take that back.. Too my knowedge, no one in the black market has broken these things yet. If someone has let me know.
Note that there's really no such thing as "tamper-proof" hardware.
;)
heh. It's impossible, for the near future, to physically extract significant information from the human brain (such as a PIN or password). Some work has been done in biological computing to use the "brains" of smaller insects to store information - because it shows promise to being 100% tamper resistant. Though, nothing working has any practical use yet - as far as I'm aware.
But actually what the newer sat decoder boxes do is have a slow smart card that decrypts a new key every second or so. The smart card is "tamper-resistant" to a large degree and passes the key to the decoder hardware which is not tamper-resistant.
No one has broken the smart cards yet... but that's not to say it can't be done with a lot of money. Getting a key for 1 second doesn't do much for you. However, if you can buffer your transmission data and delay it for a few seconds, you can have someone else continously send you the decoded keys over the internet. Having to send data continously makes the operation risker because there is a way to trace you... though you can send "almost untraceable" data by using ICMP to ping-with-data to a remote host and a forged return address. With all that money flying around, I'm sure someone will give it a try.
I really hate telephone spam because I sleep wierd hours and they wake me up. Seems like everyone is using auto-dialers these days. These things are easily detectable, just say hello and if no one responds in 1.5 seconds then it's an autodialer. I was thinking about setting up a modem that would say an automated "hello" and listen for silence. If it wasn't an auto-dialer then it would beep me and tell the person to hang on a second. Is there an established API under linux for doing full-duplex sound on a modem? I don't mind buying a specific brand of modem.
Thanks!
This means ISPs who have been lazy about closing their spam relay holes will have to take the RBL seriously now. If you are running a server and want to make sure you don't have any holes that will put you on the RBL telnet to mail-abuse.org
Internet Explorer, having to survive on its own without the rest of a big corporation to subsidize it will have to charge money and will lose market share to the open source Mozilla project which should be in good shape by the time the ruling comes down.
IE doesn't have to charge money. It could still rake in tons of money by being a hot portal site. Millions of people never change the default homepage. I'm surprised RedHat hasn't started doing this for NS under linux... maybe they have some sort of agreement.
I've been using 3 21" ViewSonic monitors for about 1.5 years. After that I developed an eye stigmatism. The refresh rates on monitors is much more visiable out of the corner of your eye (try it). Somehow my left eyelid would spaz out at random interval even when I wasn't look at at moniters. I started turning off the side monitors when I'm not looking at them, which is most of the time. About 2-3 months later my eye problem stopped.
This probably won't happen with LCD monitors, but be careful! If you are using normal monitors put them at the highest possible refresh rate they will do.
You are correct, except that mp3 was meant to be a fixed-bandwidth streaming format. Thus regardless of the sound playing there is fixed amount of bandwidth allocated to play it (128kb in this case). Playing sin(x) obviously doesn't require that much bandwidth - but it's still allocated 128kb, while more complex suffer when they exceed this rate. Some encoders will dynamically switch encoding rates depending on the complexity of the sound, which can save you a good bit of space.
I think we are going to see a new generation of music encoders that can do several times better than mp3. There is much self-similarity that is not exploited. I have an album called Andy Warhul - Ah yes, Ah no. In this album andy plays back 2 audio clips one saying "ah yes" and one saying "ah no." He uses the same clips with no pitch change over and over thousands of times (it's quite boring - but somehow interesting). Compressing such a stream should result in a file 100K or less, yet it takes 30MB. This is an extreme example where self-similarity (fractals if you want to call it that) could be be used to compress music.
Most music has very repetive patterns that can be exploited. Mp3 is designed for movies not music. There hasn't been much need for better music compression until recently because no one makes money off of it. Now that people are, I think we will see better algorithms replace mp3, and it's not MS's format.
It looks like the took apache and spiffied it up. Some of nicer new features (besides SSL) are:
"
Remote Configuration: a browser-based configuration tool to allo[w manipulation of the server configuration via a GUI.
Machine Translation Support: This new function, when used with an available IBM Machine Translation Engine, enables the IBM HTTP Server to translate English Web pages into other languages without human intervention. This permits a Web site visitor to read the page in his native language, effectively broadening the reach of your Web site. IBM Machine Translation Engines are included in the WebSphere Application Server 3.0 and include: German, Simplified Chinese and Traditional Chinese. Additional languages will be available in the future.
"
Is it just me or does Corel Draw crash every 5 minutes (windows version)? I'm afraid to use it, because I lose my work every time. Even saving after every operation doesn't work because somehow it manages to corrupt it's files. I've upgraded all the patches they have on their web site, but they seem to have no effect on it's stability.
Corel draw would be a wonderful program if it was more stable, but in it's current state it's very unusable. Anyone have suggestions for a more stable vector based drawing program?
300! is
:)
306057512216440636035370461297268629
388588804173576999416776741259476533
176716867465515291422477573349939147
888701726368864263907759003154226842
927906974559841225476930271954604008
012215776252176854255965356903506788
725264321896264299365204576448830388
909753943489625436053225980776521270
822437639449120128678675368305712293
681943649956460498166450227716500185
176546469340112226034729724066333258
583506870150169794168850353752137554
910289126407157154830282284937952636
580145235233156936482233436799254594
095276820608062232812387383880817049
600000000000000000000000000000000000
000000000000000000000000000000000000
000
I have no idea what 300M! is
NetCraft counts over four million web servers running Apache.
I assume they count servers by IP address not by domain names right? I wonder how accurate this information is since a popular way to do virtual hosting is to assign multiple IP addresses to a single host. If there are a small number of people assigning whole sub-domains to a single machine these numbers could be an order of magnitude off.
That page might load faster, but it doesn't have a document.f.q.focus(); like google's front page does. you have to hit tab 3 times or click on the field (yuck!)
....sigh... I'm so lazy. :)
Wierd how I never noticed that. That is a bit of a pain.
as for a start page.. i use a page on my local computer
I used to do that, but I use so many computers I kept losing the file. I should set one up on my own homepage, but
If you use google as your homepage as I do, you may want to use this link instead. It loads faster and looks cleaner than the page they have up.
http://www.google.com/search?q=
Most VC firms worth anything should be willing to do that as they regularily deal with proprietary or secret information.
From my experience, most VC firms won't sign a NDA (non disclosure agreement) because they get so many business plans many of them overlap and they would have a conflict of interest. They are more likely to sign a NDA if you have a proven track record, but then you probably don't need money.
If it's a really good idea, people will steal it no matter what you do. The advantage you have is that you have a working proto-type and hopefully not to far off from production, where other people will be playing catch-up. Patents now take 18 months to go through, so they are not worth much in a startup environment. Better just to file a sloppy patent yourself and slap a patent-pending on your product to scare people off. If you go to lawyer you are looking at $20k for a good application.
I suggest leaving out the really technical details from your business plan and wait till you find someone who sounds really interested before disclosing the rest.
Thought I mention I have a page with pictures and brief bios of some of my favorite CS heros. (Turing, Shannon, Huffman, Whitefield, Miyamoto, RMS, and Ken Perlin). Click on my sig then [heros].
In a worst-case scenario the proliferation of such devices will discourage young people from exploring the opportunities of teaching as a career, and may decrease our chances of.....
Computers are better than humans at chess, but that hasn't stopped people from playing or getting into it. If anything, I think it has helped because now people can always find a challenging player/teacher.
CDDB the online CD database claims
"still the world's largest CD database with over 390,000 titles and 4,500,000 audio tracks".
Estimate an audio track to be on average 3MB, and you are looking at 12 terabytes
of music right there. From my experiences CDDB has pretty good coverage of english
music, but it's lacking some foreign titles. So add 10-20% more to the estimate. They are
currently working on a version with international character sets, so it might be a lot higher
if they don't have any Asain titles. Also you add maybe another few TBs for new
bands and old bands that are not available in CD form.
I wonder how many years it will be before 16TB is easily affordable? Less than 10 if moore's
law holds for storage. hmm.. it would take you ~64 years to listen to it all though. course
there is very little of that which you actually *want* to listen to. that's where group filtering comes in.
Anyhow, my prediction is that within 10 years an ordinary person will have a complete collection of the world's published music in their home. Legal issues aside, I think this is pretty exciting.
All this time I thought people were joking when they put a (tm) next to "Good Thing". I just did a trademark search, and guess what. The mark is owned by Martha Steward! Why am I not surprised?
Word Mark GOOD THING
Owner Name (REGISTRANT) MARTHA STEWART LIVING OMNIMEDIA LLC
Owner Address 20 West 43rd Street, 25th Floor New York NEW YORK 10036 LIMITED LIABILITY COMPANY DELAWARE
Attorney of Record HOWARD J SHIRE
Serial Number 75-516347
Registration Number 2272142
Filing Date 07/09/1998
Registration Date 08/24/1999
Mark Drawing Code (1) TYPED DRAWING
Register PRINCIPAL
Published for Opposition 06/01/1999
Type of Mark SERVICE MARK