what if you were to use it as a ram-jet sort of thingee, with the plasma superheating (and thus expanding) atmospheric gases in a specially shaped chamber?
Ok, so we've gone and turned our fancy plasma drive into an expensive blowtorch. I guess the question is whether the ion flux is enough to actually heat useful amounts of gas, or whether it is just a piddly stream only useful in space (where, by the way, no-one can hear you scream).
The thing about biometrics is that they rely on secure hardware/software. Ie, it's a great idea for ATMs because the bank has incentives to make it tamperproof.
But for home computers in a hostile setting ("cmon, Johnny, help mom get rid of this annoying password scheme on my Bette Midler collection") it is completely unworkable. It is relatively easy to figure out where the biometric input is collected and collated (ie, after the NN has had a chance to guess on whether the variances in typing speed / retina patterns are pass/fail).
It can't stand up to more than five minutes of reverse enginnering.
The other way to get parallax is of course to display multiple views. Instead of just the two, you can get lenticular lenses with 4 or 7 (at least that's what philips has) views. This would buy you the parallax and greated FOV that seems to be the weakness with this approach -- but at a corresponding loss in horisontal res.
You've seen this if you've played with one of thost 7 odd frame animated postcards. In this scenario, tho, each frame isn't a different time, it's rather a view from a different angle.
I have no idea why I didn't get this before.
Eventually, we'll have enough resolution to go to those kinds of multi-way lenses. Then this will be really sweet.
Hrm. I could have sworn Ireplied to this. anyways, no:-( but I can supply a bit more accurate biblio info:
Pessimal algorithms and the simplexity of computations. A. Broder and J. Stolfi. (Satyrical article.) ACM SIGACT News vol. 16 no. 7, 49--53. Fall 1984. [bro-sto-84-pes]
ok, so above I mentioned MIT's display that tracks your head and recalculates the POV for the 3D display based on that, in the context of parallax. In retrospect, I may have glossed over the complications of how they get a wide field of view from a lenticular display. I seem to recall them having some beam combiner thingiee..
but it struck me that a low tech solution might be the best idea -- mount the display on gimbals, and have some dedicated electronics (read: cheap) to always point it towards a reflective dot on your forehead. This is just a simple extension of the track-the-sun that some solar cells use. The current position of the display is then read by the computer to regenerate the POV.
This would solve the FOV, no?
It might be a bit disconcerting, tho, to have a display that tracks you as you work... but it would be cool, at least until your baseball cap obscures the dot and the display goes into target aquisition mode, swivelling about like a headless chicken
90 steps is plenty, given that our vision only uses this sort of depth perception for things that are very close. It's not like you have to split the 90 steps up linearly from distance 0 to infinity. Serendipty helps us out: it works out that you get more resolution for close objects' depth (large horisontal separation) than you do for distant objects (low, or zero horisontal separation (*)), which is ok because motion parallax takes over beyond a certain depth. This is why MIT's display is soo cool.
MIT has a display that uses this sort of lenticular display, but also has a camera that keeps tabs on your head. So that when you move it, the POV is moved too. So you get the parallax for distant objects and "real 3d" for near ones.
I'd like to see it in action one of these days (they're just across the river from here).
(*) can anyone comment on whether it is possible to use negative separation to illuse (a verb of my own invention) an object moving very far away?
There's a great paper about pessimal algorihtms. These two guys from DEC look at slowsort and reluctant search. The idea is that they provably terminate, but take the longest time they can to do so, without ever wasting a computation.
FFTs have been used for everything. Someone once said -- find a way to turn an O(n^2) alg to O(n lg n) time, and they will beat a path to your door finding uses for it.
Also, it we should mention (alhough I've been skimming previous comments, so maybe redundant) that Gauss used FFT to compute fourier series -- he didn't make a big deal out of it, as he no doubt conscidered it a straight forward application of base principles. Smart man that.
after a movie came out featuring him the directors guild decided to change it -- unclear to what. Apparently, in an uncharacteristically honest move, the director's guild will only allow a director to distance him/her self from a movie if they can show it was changed substantially in post-production. And the only name it could be changed to was Alan Smithee.
The movie that broke the camel's back was about a director whose name *was* Alan Smithee, who wanted to get distanced from a movie he'd made. The board said fine. But only to Alan Smithee...
The ultimate joke is that according to IMDB, the movie is so bad that the real life directors would have wanted to be distanced from it too. Too bad, it would have been funny: the movie "Alan Smithee" by Alan Smithee, about Alan Smithee.
A digital signature encrypts a secure hash of the document. You call it a secure hash because to find another document that hashes to the same value should require brute force search.
This is not the case with CRC32, in which all you need to be able to do is to arbitrarily be able to control 32 consecutive bits -- and those bits can be calculated efficently.
Typically a 160 bit hash is used.
So no, you can't just tack on a signature to another document.
This is partially also the motivation of biometrics. If some of your personal features are incorporated into your passphrase, it should be even harder to fake your digital signature.
When you think about it, this has a nice circularity in that pen-and-ink signatures are pure biometrics.
Of course, serious implementation problems remain for biometrics, in that you want to be able to sign something even though you just had an accident and lost a finger, or had root canal surgery and can't speak normally.
I presume that you were being feseatious (sp?) about the knapsack problem. It can be very sucessfully attacked with dynamic programming -- I think. I know it has been shown to be unsuitable for a one-way function.
However, elliptic fields still resist many of the tricks that apply to normal number fields.
Ah, heck, I'm talkin' outta my ass again. I'm just repeating this like gospel, and I'm probably not remembering it right.
Ok, so they say they noticed this virus because it caused unexpected network traffic.
That means that they actively go out and try to download software that might be infected, trying to find new viruses.
imagine the job specification:
"Candidates with at least 3 years experience in searching the 'net for pr0n are invited to apply to NETSEC as system infectors."
So there's this guy sitting in their office, probably dressed in a suit and tie, tongue lolling in his mouth, downloading one exe after another, hoping that this one won't be a virus so that he gets to watch some more porn.
You gotta grin.
Reminds me of the visual-recognition people a few years ago who were using pr0n as input -- it turns out that skin is such a uniform color, it's easy to discriminate. So they were sending their program -- funded by darpa no doubt -- into their collection to find matches for "woman, bent over".
Iguess it is a bit like procmail. There is some program in the mail delivery system that parses my.procmail script and pipes some of my mail through programs that I specify. Obviously, it suids to me before running them.
With this bug, I could specify that the body of the mail be piped to/bin/sh... because the privs weren't dropped, Iget the power of root.
fun!
Re:Microsoft: Subversive Lawmaking?
on
Copyrant
·
· Score: 2
[Microsoft intentionally goading the court into overreacting]
That is an incredibly astute observation. It's a wonderful gamble; they're basically gambling that in two years they'll be able to apply enough PR muscle to make it impossible, politically, for the appelate court to uphold the ruling, which it would likely do otherwise (the historical record is pretty damning).
Given that 1)if there is one thing microsoft does well, it is PR 2) courts are depressingly politicised we conclude 3) they've won.
well, Dan, thanks a bundle for making this a bittersweet victory for the forces of good (that's us, that is). grumble
And the best part is the little gopher guy running around in the cooking pit asking questions.
No, actually, the dubbing is great. Especially when they dub the giggles of the actress. I think everything should be dubbed. Dubbing is easily the most amusign thing that can be done to a show.
You know all those spaghetti westerns with clint -- they were hillarious because of the dubbing.
I was unaware of the incompatibility. Mu retracted. or more muscically:
"undo the mu do that you do so well"
On a more serious note, I can see how RAID or LVMand journalling all have to play nice with each other, in that they have to keep meta-data synced up, but can anyone shed light on why knfsd is different from any other kernel access to filesystems -- such as a read() system call?
How can they tell unrequested trafic as from traffic that is requested via the protocol?
f.ex, I could well imagine a client/server setup where the client sends the server a port number to connect to -- for example a simple RMI/RPC callback.
How can they tell this from a portscan, or do they just disallow incomming connections to ports 1024?
a smurf is virtually untracable, the source addresses points back to the (
misconfigured) amplifier network, which is totally innocent,
My emphasis. I dunno about that. Perhaps what we should be looking at is a sort of administrator license for networked computers.
While I'm not sayin' that the admin of a compromised machine or network is automatically liable for anything that machine is used for, I'm not quite willing to accept them as blameless either.
I propose that if you want to expose your machines directly to the net, opening them up to attacks, you should be responsible for due diligence in making it secure against compromise.
This is exactly the sort of thing that courts are able to decide (they may not understand technology, but they do understand responsiblity and accountability). If you can show logs that the most recent patches were applied and what not, then you are free from liability, while if you run a completely misconfigured system or notoriously insecurable OS without a firewall, then you just might be liable for damages.
Now, before you start moaning about clueless users and cablemodems; if most people get cable modems just to surf the web, they could ask their ISP to put them behind a firewall that blocks most ports, thus avoiding individual configuration responsiblity.
If however, you want to get greater freedom(*), then you sign a waiver of firewalling, and accept that you had better run a tight ship.
Of course, fine points like trojans not blocked by firewalls and http tunneling remain. But the principle is that an easily compromised host is a liablity to the net at large, so I don't see why it isn't the administrator's responibility to secure it.
Johan
(*) ignoring the social engineering factor here: what are you gonna use those freedoms for? Justify in 100 words or less.
That little bit of personal information duely disclosed, I'm wondering how many computations could be expressed as cellular automata. The wonderful thing about them is of course that each iteration only needs to communicate NSWE.
If we had a (and maybe we do, I don't know anything about distributed net) client that was able to talk to other clients, then we should be able to distribute those computations too.
The thing about a cellular network is that it has a global clock; all nodes count in lockstep. So now we'd be bottlenecking the communication with the server -- this is necessary for reporting results and what not. However this could be designed against by only sending updates every n iterations, with different nodes having different offsets.
The second thing we need to think about is recovery from a node failure. Each node could communicate its state each iteration to two neighbors. Then if it died, the server could just reassign that node and it could ask its neighbors what it's state was.
So I have no idea if this would work. Since I'm posting it here though, I suspect it would. However, people don't do it this way, so I'm prolly wrong. Tell me why.
Slashdot Mirror
what if you were to use it as a ram-jet sort of thingee, with the plasma superheating (and thus expanding) atmospheric gases in a specially shaped chamber?
Ok, so we've gone and turned our fancy plasma drive into an expensive blowtorch. I guess the question is whether the ion flux is enough to actually heat useful amounts of gas, or whether it is just a piddly stream only useful in space (where, by the way, no-one can hear you scream).
Johan
The thing about biometrics is that they rely on secure hardware/software. Ie, it's a great idea for ATMs because the bank has incentives to make it tamperproof.
But for home computers in a hostile setting ("cmon, Johnny, help mom get rid of this annoying password scheme on my Bette Midler collection") it is completely unworkable. It is relatively easy to figure out where the biometric input is collected and collated (ie, after the NN has had a chance to guess on whether the variances in typing speed / retina patterns are pass/fail).
It can't stand up to more than five minutes of reverse enginnering.
Did anyone notice the name of Real Media's representative?
Jenny Sorenson
RealNetworks
Probably just a coincidence.
Johan
So I'm not so bright.
The other way to get parallax is of course to display multiple views. Instead of just the two, you can get lenticular lenses with 4 or 7 (at least that's what philips has) views. This would buy you the parallax and greated FOV that seems to be the weakness with this approach -- but at a corresponding loss in horisontal res.
You've seen this if you've played with one of thost 7 odd frame animated postcards. In this scenario, tho, each frame isn't a different time, it's rather a view from a different angle.
I have no idea why I didn't get this before.
Eventually, we'll have enough resolution to go to those kinds of multi-way lenses. Then this will be really sweet.
Hrm. I could have sworn Ireplied to this. anyways, no :-( but I can supply a bit more accurate biblio info:
Pessimal algorithms and the simplexity of computations. A. Broder and J. Stolfi. (Satyrical article.) ACM SIGACT News vol. 16 no. 7, 49--53. Fall 1984. [bro-sto-84-pes]
ok, so above I mentioned MIT's display that tracks your head and recalculates the POV for the 3D display based on that, in the context of parallax. In retrospect, I may have glossed over the complications of how they get a wide field of view from a lenticular display. I seem to recall them having some beam combiner thingiee..
but it struck me that a low tech solution might be the best idea -- mount the display on gimbals, and have some dedicated electronics (read: cheap) to always point it towards a reflective dot on your forehead. This is just a simple extension of the track-the-sun that some solar cells use. The current position of the display is then read by the computer to regenerate the POV.
This would solve the FOV, no?
It might be a bit disconcerting, tho, to have a display that tracks you as you work... but it would be cool, at least until your baseball cap obscures the dot and the display goes into target aquisition mode, swivelling about like a headless chicken
I'm not sure I agree at all.
90 steps is plenty, given that our vision only uses this sort of depth perception for things that are very close. It's not like you have to split the 90 steps up linearly from distance 0 to infinity. Serendipty helps us out: it works out that you get more resolution for close objects' depth (large horisontal separation) than you do for distant objects (low, or zero horisontal separation (*)), which is ok because motion parallax takes over beyond a certain depth. This is why MIT's display is soo cool.
MIT has a display that uses this sort of lenticular display, but also has a camera that keeps tabs on your head. So that when you move it, the POV is moved too. So you get the parallax for distant objects and "real 3d" for near ones.
I'd like to see it in action one of these days (they're just across the river from here).
(*) can anyone comment on whether it is possible to use negative separation to illuse (a verb of my own invention) an object moving very far away?
There's a great paper about pessimal algorihtms. These two guys from DEC look at slowsort and reluctant search. The idea is that they provably terminate, but take the longest time they can to do so, without ever wasting a computation.
FFTs have been used for everything. Someone once said -- find a way to turn an O(n^2) alg to O(n lg n) time, and they will beat a path to your door finding uses for it.
Also, it we should mention (alhough I've been skimming previous comments, so maybe redundant) that Gauss used FFT to compute fourier series -- he didn't make a big deal out of it, as he no doubt conscidered it a straight forward application of base principles. Smart man that.
after a movie came out featuring him the directors guild decided to change it -- unclear to what. Apparently, in an uncharacteristically honest move, the director's guild will only allow a director to distance him/her self from a movie if they can show it was changed substantially in post-production. And the only name it could be changed to was Alan Smithee.
The movie that broke the camel's back was about a director whose name *was* Alan Smithee, who wanted to get distanced from a movie he'd made. The board said fine. But only to Alan Smithee...
The ultimate joke is that according to IMDB, the movie is so bad that the real life directors would have wanted to be distanced from it too. Too bad, it would have been funny: the movie "Alan Smithee" by Alan Smithee, about Alan Smithee.
Oh well, two out of three's not so bad.
A digital signature encrypts a secure hash of the document. You call it a secure hash because to find another document that hashes to the same value should require brute force search.
This is not the case with CRC32, in which all you need to be able to do is to arbitrarily be able to control 32 consecutive bits -- and those bits can be calculated efficently.
Typically a 160 bit hash is used.
So no, you can't just tack on a signature to another document.
This is partially also the motivation of biometrics. If some of your personal features are incorporated into your passphrase, it should be even harder to fake your digital signature.
When you think about it, this has a nice circularity in that pen-and-ink signatures are pure biometrics.
Of course, serious implementation problems remain for biometrics, in that you want to be able to sign something even though you just had an accident and lost a finger, or had root canal surgery and can't speak normally.
I presume that you were being feseatious (sp?) about the knapsack problem. It can be very sucessfully attacked with dynamic programming -- I think. I know it has been shown to be unsuitable for a one-way function.
However, elliptic fields still resist many of the tricks that apply to normal number fields.
Ah, heck, I'm talkin' outta my ass again. I'm just repeating this like gospel, and I'm probably not remembering it right.
the smc at the bruching shuttlecocks thinks that the movie had great hair. John Travolta in dreads!
yeah!
Ok, so they say they noticed this virus because it caused unexpected network traffic.
That means that they actively go out and try to download software that might be infected, trying to find new viruses.
imagine the job specification:
"Candidates with at least 3 years experience in searching the 'net for pr0n are invited to apply to NETSEC as system infectors."
So there's this guy sitting in their office, probably dressed in a suit and tie, tongue lolling in his mouth, downloading one exe after another, hoping that this one won't be a virus so that he gets to watch some more porn.
You gotta grin.
Reminds me of the visual-recognition people a few years ago who were using pr0n as input -- it turns out that skin is such a uniform color, it's easy to discriminate. So they were sending their program -- funded by darpa no doubt -- into their collection to find matches for "woman, bent over".
It's all honest research. really.
Iguess it is a bit like procmail. There is some program in the mail delivery system that parses my .procmail script and pipes some of my mail through programs that I specify. Obviously, it suids to me before running them.
/bin/sh... because the privs weren't dropped, Iget the power of root.
With this bug, I could specify that the body of the mail be piped to
fun!
but Ican't figure out what GEB is. Drop us a clue?
So what crypto algorithm is supported, if not 3DES? I thought that was pretty well regarded.
Also, could you perhaps define those terms?I must have missed that the first time around: ORCON, SAP, and SCI
johan
And the best part is the little gopher guy running around in the cooking pit asking questions.
No, actually, the dubbing is great. Especially when they dub the giggles of the actress. I think everything should be dubbed. Dubbing is easily the most amusign thing that can be done to a show.
You know all those spaghetti westerns with clint -- they were hillarious because of the dubbing.
Ah so.
I was unaware of the incompatibility. Mu retracted. or more muscically:
"undo the mu do that you do so well"
On a more serious note, I can see how RAID or LVMand journalling all have to play nice with each other, in that they have to keep meta-data synced up, but can anyone shed light on why knfsd is different from any other kernel access to filesystems -- such as a read() system call?
Johan
How can they tell unrequested trafic as from traffic that is requested via the protocol?
f.ex, I could well imagine a client/server setup where the client sends the server a port number to connect to -- for example a simple RMI/RPC callback.
How can they tell this from a portscan, or do they just disallow incomming connections to ports 1024?
moo! (or is that spelled "mu"?)
eh? NFS works with any filesystem that you have mounted on your NFS server. loopback, proc, you name it.
as long as read/write perms are ok, that is.
Johan
While I'm not sayin' that the admin of a compromised machine or network is automatically liable for anything that machine is used for, I'm not quite willing to accept them as blameless either.
I propose that if you want to expose your machines directly to the net, opening them up to attacks, you should be responsible for due diligence in making it secure against compromise.
This is exactly the sort of thing that courts are able to decide (they may not understand technology, but they do understand responsiblity and accountability). If you can show logs that the most recent patches were applied and what not, then you are free from liability, while if you run a completely misconfigured system or notoriously insecurable OS without a firewall, then you just might be liable for damages.
Now, before you start moaning about clueless users and cablemodems; if most people get cable modems just to surf the web, they could ask their ISP to put them behind a firewall that blocks most ports, thus avoiding individual configuration responsiblity.
If however, you want to get greater freedom(*), then you sign a waiver of firewalling, and accept that you had better run a tight ship.
Of course, fine points like trojans not blocked by firewalls and http tunneling remain. But the principle is that an easily compromised host is a liablity to the net at large, so I don't see why it isn't the administrator's responibility to secure it.
Johan
(*) ignoring the social engineering factor here: what are you gonna use those freedoms for? Justify in 100 words or less.
Disclaimer: I'm talking outta ma ass.
That little bit of personal information duely disclosed, I'm wondering how many computations could be expressed as cellular automata. The wonderful thing about them is of course that each iteration only needs to communicate NSWE.
If we had a (and maybe we do, I don't know anything about distributed net) client that was able to talk to other clients, then we should be able to distribute those computations too.
The thing about a cellular network is that it has a global clock; all nodes count in lockstep.
So now we'd be bottlenecking the communication with the server -- this is necessary for reporting results and what not. However this could be designed against by only sending updates every n iterations, with different nodes having different offsets.
The second thing we need to think about is recovery from a node failure. Each node could communicate its state each iteration to two neighbors. Then if it died, the server could just reassign that node and it could ask its neighbors what it's state was.
So I have no idea if this would work. Since I'm posting it here though, I suspect it would. However, people don't do it this way, so I'm prolly wrong. Tell me why.