Note the slash between vacuum and sealed.
Also, once frozen, the presure should be able to be lowered to arbitrarily low values, no?
Re:Implications to Cryptography
on
Does P = NP?
·
· Score: 1
I was under the impression that "were" was the 2nd person singular subjunctive. Is is also the 3rd?
Re:Huge crytography implications!
on
Does P = NP?
·
· Score: 1
What I meant by my second point was that by having a P solutoin, we inherently have some structure to the problem. If we are able to exploint that structure, we might be able to apply it to the brute force approach and cut down the search space.
The implicity assumption there is that we get diminishing returns in our solution. Take sorting as an example. If we have a quick sort running over an almost sorted list, we get diminishing returns -- after just a few iterations, the list may be completely sorted except for one element, but we'll be running O(n^2) iterations just to move that element to where it was supposed to be.
If we have an O(n^4096) solution to 3sat, then we might be able to run it for a few hundred iterations and analyse the partial result. Say that 75% of the time, we learn something useful about the structure of the answer (enough to practically apply a brute force approach, say). Since this a probablistic result, we couldn't use it to improve the O() time of the algorithm, but it would be a very effective attack against cryptography.
Re:Huge crytography implications!
on
Does P = NP?
·
· Score: 1
I dunno. I would be hard pressed to think of any case where the worst-case polynomial doesn't simplify down to one of order \leq to the lenght of the input. Still not so good, but a limit. Sheer speculation.
Another avenue is that perhaps a partial solution would be useful from a cryptographic point of view. Perhaps we do O(n^10) units of work and use the partial result to construct a 2^60 bit brute force attack against a key. also speculation, but I'd appreciate your thoughts.
Re:Implications to Cryptography
on
Does P = NP?
·
· Score: 1
The blow comes from transforming an exponential complexity
k^n
problem (where k is a constant -- say 2 or 4096) into a polynomial problem
n^k.
We can always grow n to suit our security needs, but if the problem is polynomial, the problem complexity just won't grow fast enough to keep up with the exponential growth of our computing resources.
Johan
Re: Is this problem NP complete?
on
Does P = NP?
·
· Score: 1
Expanding on the under-modded previous reply, Cook showed that 3sat is able to solve any problem that was in NP. He did this by constructing a 3sat problem from the turing machine(*) that would solve the NP problem. This construction can be performed in P time (important!).
Thus, Cook showed that if we are able to solve 3sat, then any problem in NP can be solved by constructing a 3sat problem in P time and then solving that.
The completeness part comes from the fact that 3sat is itself in NP -- this is fairly easy to show. All you need to show is that you can in P time verify that a proposed 3sat answer is correct. this is easy.
(*) I should probably admit that I am a bit hazy on this detail; obviously we can't use this construction to solve the halting problem, so the construction can only work for problems in NP, but turing machines can also solve Non-NP problems. Can anyone clarigy
It would be a trivial excersise to add the option to use p/gp/g for outgoing mail. I'm sort of suprised they haven't implemented that already. Incoming pgp mail could also be dealt with in a similar way (though you would have to trust the service enough to send it your private pgp key -- which is more than I would like to do. As soon as I enter the pass phrase, a compromised front end could capture my private key. Not good).
Hushmail's biggest problem is exactly to ensure that no one can compromise the java byte codes in transit. This requires signatures and authentication infrastructure that is hard to assume accross architecures, no?
far fetched if you look at the image as a whole, but very believable if you look at small enough blocks.
But perhaps I mislead you by using the term periodic rather than self-similar. All was saying was that many blocks would be very similar to blocks from previous frames -- and was trying to make the point that the best frame to match against might be 4 or even 15 frames ago. Take a car driving past a background; the blocks making up the bg will be the same before and after the car passes, so the best source of image composites would be maybe 15 frames back.
the motivating idea behind 3d compression being that motion in movies tends to be periodic (people talking, walking, f*cking..., wheels turning) the time relationship should be quite rewarding, as long as the window is large (long?) enough.
But why do you say that wavelets could not be plugged in? The basic steps of transform, quantize, untransform remain the same, no? Wavelets just have more appealing behavior under the rather extreme quantification (quanitization? same word?) needed for low bit rate.
'course there is a way to combine wavelets and motion compensation.
My understanding of how video codecs work is that there is basically a two step process: 1) you try to simplify each frame by extracting information out of it that refers to historical data, 2) you compress the rest as graphics. The intent being that compressing the difference will be more efficient than compressing each frame anew. Makes sense, no?
so 1) is where motion compensation comes in
and 2) is where wavelets come in.
A very simple codec might be to simply subtract each frame from the preceeding one (which means that unchanging areas are black) and compress the results with jpg. More advanced systems might split the frame up into segments and do a transformation (== motion compensation / rotation) on each segment before subtraction and compression.
Now for my question: the segmentation can be very hard to get right - exhaustive search being time consuming and all that. Is anyone doing it in the time domain as well, so that not only is the previous frame searched for a good match, but perhaps the last 30 or so. If the video is very flashy (think mtv), this might work quite well.
That makes more sense. Now I can calm down; I was afraid that all the fertilizer and diesel that I've been hoarding was going to get the boys in blue to come aknocking.:-)
Isn't iodine used to clean out cuts? Or is that something-or-other of iodine, in the same way that the lithium taken as an anti-depressant isn't pure lithium?
Not to question your integrity, but this I cannot believe. The use of shopping records by lawenforcement before a crime was commited? I could see how they would be able to subpoena them afterwards for use in the trial, but only if they had a suspect and wanted circumstantial evidence or prehaps to prove premeditation. Your example smells of -- oh I forget the term -- something infringing on privacy, presumed innocence.
... and that was the best job of design they could do? The thing sucks at something like three different levels. It looks bad. It has no good encraption. It is distributed in a way that is a licencing nightmare.
You know, everytime a company does something like this, I have to smile, 'cause if the barrier is this low, I'll make a killing.
As soon as I loose my scruples.
You know, I realise that I'm not 100% on what single user mode is. Only one UID can have any processes in the run queue? Only one process runs -- task switching disabled?
Time to whip out the Tannenbaum. On that note, is he still the best general purpose OS reference text out there?
Not a quote, but an aside:Grumble. I've already moderated, now that point is wasted. Will someone go and mod up the question about whether anyone has actually died in space, as opposed to the lauch pad or mid-air?
now onto the meat. More meat!
how pucnture resistant are those tanks (or can they be made to be?). What about structural integrity in general; I presume they're mostly empty when they reach the low-presure part of the atmosphere, thus they don't really need to withstand a 1 atmosphere presure diff.
Back to the deaths in space. A question for y'all: how much is x% extra saftey worth? Back in the good old days, life was cheap. This is what enabled explorers to risk life and limb to discover new trade routes (and flaunt certain death at the edge of the world). These days, life is much more expensive. Is it too expensive for us to afford exploration? discuss
Blue sky warning here. How hard would it be to checkpoint a kernel?
What I'm after is a way of bringing a machine down in such a way that application processes can be frozen, a new kernel swapped in, and the applications unthawed.
It all comes down to is how much kernel state a process has; by definition very little. It has at most handles for internal kernel datastructures. So as long as the two kernel versions know enough about each other to translate those, you shouldn't need to reboot a machine to upgrade the kernel.
The big thing I've overlooked is hardware state; things like BIOS/network/whatever. These services will likely need to be restarted.
Since you're sending mail to the recipient, and you need it encrypted, you must trust them somewhat. At least their intentions. But perhaps not their technical skill; they don't understand swap files, that sort of thing.
So you send them email in a format that makes it maliciously hard to let compromising information leak into the insecure enviroment (after all, the person could just blab, but you presumably trust them enough not to do that). So this doesn't decryt to file, it decrypts to screen. Likewise, the timeout features can be circumvented, but only be a malicious recipient. I actually can't think of a scenario where that would be necessary. Perhaps the person you are corresponding with will be exchanged with a new person, and you don't want them reading your past exchanges?
Anyways, if what you are concerned with is deniablity, you need to not sign any messages. That is the only protection availible in the ever-forgable digital world. Perhaps that is what the company means by auto-shredding. Their server will no longer verify the sender of a message after it has timed out. That could be useful.
erm. this seems like a problem that is solvable in any number of ways. Replication seems to be easiest. Cache popular content onto fast pipes (provisions for bandwidth limiting are assumed). Encode a forwarding requirement into the protocol -- every file you download, you have to allow someone to grab that file from you. Use multicast and PPV style scheduling (requesters register for a file, letting the server determine when (within a short timeperiod) to multicast it).
I suprised by this being an issue at all. I haven't looked at the gnutella infrastructure, but these are issues that I would have thought tackled during the initial design.
No, you type it into a java applet running locally on your browser, which communicates with hushmail's servers over some public-key-exchange encrypted channel (likely RSA, but that's a guess).
So there are three points of attack:
1) compromise your browser/vm.
2) compromise the hushmail server.
3) compromise the bytecodes intransit.
Obviously number 3 is the easiest way to go. Interestingly, microsoft's ideas with signed binaries would be a [partial] solution to that. You would then have to
4) compromise signer's certificate
and as soon as that happens, basically the attacker needs to compromise the whole infrastructure, which we assume is impractical.
so presumably the boil off / evaporation of black holes comes from the negative particle being sucked in, and thus causing a net loss in mass.
However, this leaves a few questions
1) how come more negative (un-, if you wish) than positive particles are attracted?
1.1) how come gravity works the same on unparticles? Shouldn't they be repelled?
2) why does this happen at a faster rate for small black holes (I understand that rate of evaporation is inversely proportional to mass)
That was my first distro! 1.2.13; took fscking hours to install, 'cause after each (common) reboot in the install process, it would laboriously chug through all my the io offsets looking for the cdrom. Once I got the hint that adding cdrom=hdd to the boot line, life was sweeter.
Slashdot Mirror
Note the slash between vacuum and sealed.
Also, once frozen, the presure should be able to be lowered to arbitrarily low values, no?
I was under the impression that "were" was the 2nd person singular subjunctive. Is is also the 3rd?
What I meant by my second point was that by having a P solutoin, we inherently have some structure to the problem. If we are able to exploint that structure, we might be able to apply it to the brute force approach and cut down the search space.
The implicity assumption there is that we get diminishing returns in our solution. Take sorting as an example. If we have a quick sort running over an almost sorted list, we get diminishing returns -- after just a few iterations, the list may be completely sorted except for one element, but we'll be running O(n^2) iterations just to move that element to where it was supposed to be.
If we have an O(n^4096) solution to 3sat, then we might be able to run it for a few hundred iterations and analyse the partial result. Say that 75% of the time, we learn something useful about the structure of the answer (enough to practically apply a brute force approach, say). Since this a probablistic result, we couldn't use it to improve the O() time of the algorithm, but it would be a very effective attack against cryptography.
I dunno. I would be hard pressed to think of any case where the worst-case polynomial doesn't simplify down to one of order \leq to the lenght of the input. Still not so good, but a limit. Sheer speculation.
Another avenue is that perhaps a partial solution would be useful from a cryptographic point of view. Perhaps we do O(n^10) units of work and use the partial result to construct a 2^60 bit brute force attack against a key. also speculation, but I'd appreciate your thoughts.
The blow comes from transforming an exponential complexity
k^n
problem (where k is a constant -- say 2 or 4096) into a polynomial problem
n^k.
We can always grow n to suit our security needs, but if the problem is polynomial, the problem complexity just won't grow fast enough to keep up with the exponential growth of our computing resources.
Johan
Expanding on the under-modded previous reply, Cook showed that 3sat is able to solve any problem that was in NP. He did this by constructing a 3sat problem from the turing machine(*) that would solve the NP problem. This construction can be performed in P time (important!).
Thus, Cook showed that if we are able to solve 3sat, then any problem in NP can be solved by constructing a 3sat problem in P time and then solving that.
The completeness part comes from the fact that 3sat is itself in NP -- this is fairly easy to show. All you need to show is that you can in P time verify that a proposed 3sat answer is correct. this is easy.
(*) I should probably admit that I am a bit hazy on this detail; obviously we can't use this construction to solve the halting problem, so the construction can only work for problems in NP, but turing machines can also solve Non-NP problems. Can anyone clarigy
It would be a trivial excersise to add the option to use p/gp/g for outgoing mail. I'm sort of suprised they haven't implemented that already. Incoming pgp mail could also be dealt with in a similar way (though you would have to trust the service enough to send it your private pgp key -- which is more than I would like to do. As soon as I enter the pass phrase, a compromised front end could capture my private key. Not good).
Hushmail's biggest problem is exactly to ensure that no one can compromise the java byte codes in transit. This requires signatures and authentication infrastructure that is hard to assume accross architecures, no?
far fetched if you look at the image as a whole, but very believable if you look at small enough blocks.
But perhaps I mislead you by using the term periodic rather than self-similar. All was saying was that many blocks would be very similar to blocks from previous frames -- and was trying to make the point that the best frame to match against might be 4 or even 15 frames ago. Take a car driving past a background; the blocks making up the bg will be the same before and after the car passes, so the best source of image composites would be maybe 15 frames back.
Johan
Prost! (pause 1 second) Alain Prost! is my favorite toast when drinking at dinners. Only works with european males, tho.
the motivating idea behind 3d compression being that motion in movies tends to be periodic (people talking, walking, f*cking..., wheels turning) the time relationship should be quite rewarding, as long as the window is large (long?) enough.
But why do you say that wavelets could not be plugged in? The basic steps of transform, quantize, untransform remain the same, no? Wavelets just have more appealing behavior under the rather extreme quantification (quanitization? same word?) needed for low bit rate.
'course there is a way to combine wavelets and motion compensation.
My understanding of how video codecs work is that there is basically a two step process: 1) you try to simplify each frame by extracting information out of it that refers to historical data, 2) you compress the rest as graphics. The intent being that compressing the difference will be more efficient than compressing each frame anew. Makes sense, no?
so 1) is where motion compensation comes in
and 2) is where wavelets come in.
A very simple codec might be to simply subtract each frame from the preceeding one (which means that unchanging areas are black) and compress the results with jpg. More advanced systems might split the frame up into segments and do a transformation (== motion compensation / rotation) on each segment before subtraction and compression.
Now for my question: the segmentation can be very hard to get right - exhaustive search being time consuming and all that. Is anyone doing it in the time domain as well, so that not only is the previous frame searched for a good match, but perhaps the last 30 or so. If the video is very flashy (think mtv), this might work quite well.
discuss.
That makes more sense. Now I can calm down; I was afraid that all the fertilizer and diesel that I've been hoarding was going to get the boys in blue to come aknocking. :-)
Thanks for the clarification.
Isn't iodine used to clean out cuts? Or is that something-or-other of iodine, in the same way that the lithium taken as an anti-depressant isn't pure lithium?
Mercury is toxic and it evaporates
Keep the mirror in a vacuum / sealed chamber.
A cool aspect of the design is that you can change the magnifying power of the telescope by spinning it faster or slower. Neat!
No!
Not to question your integrity, but this I cannot believe. The use of shopping records by lawenforcement before a crime was commited? I could see how they would be able to subpoena them afterwards for use in the trial, but only if they had a suspect and wanted circumstantial evidence or prehaps to prove premeditation. Your example smells of -- oh I forget the term -- something infringing on privacy, presumed innocence.
Could you please provide a link or backup data?
The people behind this are big.
... and that was the best job of design they could do? The thing sucks at something like three different levels. It looks bad. It has no good encraption. It is distributed in a way that is a licencing nightmare.
You know, everytime a company does something like this, I have to smile, 'cause if the barrier is this low, I'll make a killing.
As soon as I loose my scruples.
You know, I realise that I'm not 100% on what single user mode is. Only one UID can have any processes in the run queue? Only one process runs -- task switching disabled?
Time to whip out the Tannenbaum. On that note, is he still the best general purpose OS reference text out there?
Not a quote, but an aside: Grumble. I've already moderated, now that point is wasted. Will someone go and mod up the question about whether anyone has actually died in space, as opposed to the lauch pad or mid-air?
now onto the meat. More meat!
how pucnture resistant are those tanks (or can they be made to be?). What about structural integrity in general; I presume they're mostly empty when they reach the low-presure part of the atmosphere, thus they don't really need to withstand a 1 atmosphere presure diff.
Back to the deaths in space. A question for y'all: how much is x% extra saftey worth? Back in the good old days, life was cheap. This is what enabled explorers to risk life and limb to discover new trade routes (and flaunt certain death at the edge of the world). These days, life is much more expensive. Is it too expensive for us to afford exploration? discuss
Blue sky warning here. How hard would it be to checkpoint a kernel?
What I'm after is a way of bringing a machine down in such a way that application processes can be frozen, a new kernel swapped in, and the applications unthawed.
It all comes down to is how much kernel state a process has; by definition very little. It has at most handles for internal kernel datastructures. So as long as the two kernel versions know enough about each other to translate those, you shouldn't need to reboot a machine to upgrade the kernel.
The big thing I've overlooked is hardware state; things like BIOS/network/whatever. These services will likely need to be restarted.
Thoughts?
Since you're sending mail to the recipient, and you need it encrypted, you must trust them somewhat. At least their intentions. But perhaps not their technical skill; they don't understand swap files, that sort of thing.
So you send them email in a format that makes it maliciously hard to let compromising information leak into the insecure enviroment (after all, the person could just blab, but you presumably trust them enough not to do that). So this doesn't decryt to file, it decrypts to screen. Likewise, the timeout features can be circumvented, but only be a malicious recipient. I actually can't think of a scenario where that would be necessary. Perhaps the person you are corresponding with will be exchanged with a new person, and you don't want them reading your past exchanges?
Anyways, if what you are concerned with is deniablity, you need to not sign any messages. That is the only protection availible in the ever-forgable digital world. Perhaps that is what the company means by auto-shredding. Their server will no longer verify the sender of a message after it has timed out. That could be useful.
Johan
erm. this seems like a problem that is solvable in any number of ways. Replication seems to be easiest. Cache popular content onto fast pipes (provisions for bandwidth limiting are assumed). Encode a forwarding requirement into the protocol -- every file you download, you have to allow someone to grab that file from you. Use multicast and PPV style scheduling (requesters register for a file, letting the server determine when (within a short timeperiod) to multicast it).
I suprised by this being an issue at all. I haven't looked at the gnutella infrastructure, but these are issues that I would have thought tackled during the initial design.
No, you type it into a java applet running locally on your browser, which communicates with hushmail's servers over some public-key-exchange encrypted channel (likely RSA, but that's a guess).
So there are three points of attack:
1) compromise your browser/vm.
2) compromise the hushmail server.
3) compromise the bytecodes intransit.
Obviously number 3 is the easiest way to go. Interestingly, microsoft's ideas with signed binaries would be a [partial] solution to that. You would then have to
4) compromise signer's certificate
and as soon as that happens, basically the attacker needs to compromise the whole infrastructure, which we assume is impractical.
so presumably the boil off / evaporation of black holes comes from the negative particle being sucked in, and thus causing a net loss in mass.
However, this leaves a few questions
1) how come more negative (un-, if you wish) than positive particles are attracted?
1.1) how come gravity works the same on unparticles? Shouldn't they be repelled?
2) why does this happen at a faster rate for small black holes (I understand that rate of evaporation is inversely proportional to mass)
If the women don't find you (+1, Interesting), they might as well find you (+1, Informative).
Red Green kicks Ass!
That was my first distro! 1.2.13; took fscking hours to install, 'cause after each (common) reboot in the install process, it would laboriously chug through all my the io offsets looking for the cdrom. Once I got the hint that adding cdrom=hdd to the boot line, life was sweeter.
fun fun.