Yeah, I've heard that all new sATA drives are shipped with support for that. Also there's a big push among card makers to add the necessary supporting hardware on the cards.
Or not.;-)
Ok I'm going to be honest with you, I can't be bothered to actually calculate how big a capacitor it would take to drive a HDD for a few seconds. My guestimate would be that you're looking at an additional box with capacitors next to your PC however.
Considering that it's not hard to find people who are having the warm fussy feeling of security when they use Raid-0 I bet there's a lot of people that run Raid-[1+] (and no UPS) thinking they are safe from harm.
Yeah but a big write cache on the disk will make all of your fancy RAIDing useless. Because if the system can't reliably tell what parts have been written to disk or not (ie still in the cache) then it can't protect the data from a power outage.
1) Yes, one-way functions are public key cryptography. When I was refering to "public key" in my post I should have written "public key infrastructure" which would have been more correct. Most people don't know the difference in any case.
If you are as clued in regarding cryto as you seem to think you are you'd also know that both symmetric and asymmetric systems are vulnerable and hard to use by themselves. To make a useful system you combine the two. Use an asymetric system to exchange keys, and here you can use a Diffie-Hellman or PKI system, and then use that key in a symmetric system to make it useable. (Since asymetric systems are way too slow to be used for actual data transfer.)
The combination allows you to securely exchange keys over an unsecure channel and to transfer data (reasonably) quickly.
The rest of your points are not really relevant to the situation (stopping someone from snooping your connection) but I'll adress them anyways.
2) While I don't quite agree that it's impossible to prove that no asymmetric system is secure there is no such proof today. If you can prove that a function you have is in fact impossible to break by any other means than brute force then you'd have a provably secure assymertic system. Today you just use a system which is secure within a specific time. (Eg somone sniffing a network need to break the data in pretty much realtime for it to be useful in these applications. Otherwise the data will just pile up and they'll have to start throwing it away.)
Fast number factorisation would indeed break a lot of public key systems currently in use. OTOH number factorization is IIRC a NP-complete problem, so when that day comes it will be such a huge breakthrough in computing that I doubt anyone will cry. (You'll just have to use a one-way function not based on products of primes.) And you make it sound as if figure out a really efficient way to generate lists of prime numbers is an easy task. Since there is an abundant lack of such algorithms you could deduce that it is in fact rather hard.
3) And yes, I'm quite aware that typical loopholes in crypto systems involve attacking the human elements. It can be because the algorithm for generating prime guesses is based on mouse movement (and most poeple just move the mouse in a clockwise circle to do this) or on bugs in the implementation.
This is why I put "implement protocol correctly" in the last sentence of my post.
Re:a more mature evangelion
on
Giant Mecha News
·
· Score: 2, Insightful
I don't know if I'd call it "more mature" but it certainly is good.
The first few episodes I was a bit annoyed with how similar it was to Eva, but I got over it. You could say that if Eva has a religious background then RahXephon has musical influences. Ie music is used in RX in much the same way religion is in Eva. It's very good though. (And I recall watching a TV program and a choir sang one of the arias used in RX.)
RX is a bit easier to get into as well, the characters are more normal then in Eva. But before you call RX deeper you should check out some sites analyzing Eva, there's a lot of things in that show not immideately obvious. Calling it "childish" demonstrates that you haven't done this IMHO.
But I guess I'm influenced by the fact that Eva was one of the first anime series I ever watched. And it had a pretty big impact on me.
I think the point was that you'd encrypt all the protocol data. That way you can't fingerprint it just by looking at the headers. But since encryption is quite slow you might want to send the rest of the data unecrypted which would live it open for attacks. Hence the need for sniffing data.
OTOH if you already have a working encryption system you might as well extend it to the data. (Perhaps using a faster algorithm.) And in that case would be snoopers would be SOL.
What happens if, instead of listening to traffic on a single protocol, they just listen to all traffic, regardless of the headers? Which they, being in control of the routers, are perfectly capable of doing.
Actually they probably can't do that. At least not without some pretty extreme hardware.
Typically you get to a point where you have to use RAM buffers to save data and then have multiple network listeners which swap so they can save the data to disk. If you have a large amount of traffic you soon get to a point where you can't store all data.
But sure, it's their network, so they can do what they want. Just as long as they don't mind me using encrypted channels.;-)
What the AC was probably trying to say (but in a, shall we say, crude way) is that there is no key to extract.
Any encryption protocol worth it's salt (ie generally not those propriatary/secret protocols) is protected against this. Eg SSL or SSH which you can snoop all you want, but there's never a key sent in clear text across the channel. Neither is there any key to find in the source. Instead a key is agreed upon by the clients as they connect, but using "one way functions" which are hard and/or practically impossible to crack. You can also use public key encryption with it to add even more security and authentication to the system.
That's why he suggested that you should read about SSL. A cheap way is to just look at the relevant RFC, although it may be a bit complex if you don't have any experience in number theory. Another hint could be look at crypto sites such as Bruce Schneier's crypto-gram (counterpane.com).
In any case, if you use a well tested protocol and implement it correctly (not always trivial) then the system will be secure.
I read a topic about this on Kuro5hin and one poster came up with the following. I'll just quote directly (I saved a copy of the post, for moments like these.) Unfortunately I don't know who posted it so I can't give credit.
Speciation is more critical to evolution. It has been observed in plants several times, but the only animal occurrance observed as far as I know, is a sample of worms (Nereis Acuminata) collected in Long Beach Harbor, California in 1964 for use as lab specimens at Woods Hole Oceanographic Institute. They were bred in captivity. In the late 80s, they decided to test the population against the wild population and found that the Woods Hole population couldn't breed with the Long Beach population.
In essence the Woods Hole worms are a new species.
Just test if the URL(s) has been in any other article. If it has been add a link to those with a "Possible dupe" warning as the article is up for review.
My point is that it doesn't have to be a automatic system, that is really hard. Just add a quick fix which automates detecting possible dupes. That would kill off most of the dupes in any case.
I never did figure out why the Enterprise's workstations were built of explosive material...
Here at Star Fleet Ship Design we know how important tactice feedback is to our customers. A starship is a very complex piece of machinery to control. If you as a pilot don't get feedback on the ships status it is very easy to misjudge the situation. Furthermore a lot of useability testing showed that it was good for morale if the moron who got you into the mess was the first to blow up.
We also added the Intruder Friendly Computer System to the later models of ships. They include a number of backdoors into the system which will ensure that operators are on the lookout for patches in the future. Also it keeps the crew on their toes and on the lookout for intruders. On the odd chance of a mutiny the captain also has a fair chance to reclaim control of the ship.
We strive to life for star ship crews more interesting.
As long as the child understands the difference between shooting someone in a game and shooting someone in real life I would say it's ok. The kid is going to see a lot of people getting shot or beaten on TV in any case so they need to understand that in any case.
Children are not stupid, and if you meet a child who thinks it would be fun to go and beat some cops then perhaps you should try talking to him. Not to scold him but try to see if he is in fact talking about beating real people and discuss the consequences of that. (And try to do it in a non patronizing way too, so that it might actually have an effect on him.)
When a messed up child goes and kills someone then it's not only the fault of the parents. It's also the fault of those who could have tried to set him straight.
And for the record I don't have kids of my own. (And I didn't intend this post to be as preachy as it might have turned out.;-)
Vector processors are very much in use today. All current processors support in in some way through SSE and similar instructions, and the G4 AltiVec has a lot more to offer in the same area. Furthermore if you have a reasonably current graphics card then it uses vector processing as well.
Nobody said vector processors are dead. They just tend to be overkill for most applications. (And hence they are instead used as a type of co-processors.)
There is research going on in this field. Eg it would allow you to reconfigure part of a pipe-line while data is flowing through the chip.
This is in fact already possible, but the reconfiguration time for large parts of a chip is generally way to slow for it to be usable. But if you have a design which allows you to reconfigure only a very small part of the chip then it's doable during runtime. (Although you may need special boards to do it, I'm not sure how many developer boards actually support reconfiguration while running.)
The idea of having small premade parts is already in use by eg the RAW project at MIT. Doing runtime optimizations is probably never going to happen though because doing routing on a large FPGA can take days to complete.
One interesting project I found a few years back is the RAW project at MIT. It does pretty much the same thing but they are no longer using FPGAs. (They use chips which are similar to FPGAs but specified towards computation.) Their first prototypes used FPGAs though.
Seems like the "programming language" is similar to LabView and such schematic programming languages. (Eg in Matlab you have Simulink.) Apparently there's quite a lot of people who find that easier to work with.
Oh well, it's an interesting field. Let's just hope they don't get a bunch of ludicrous patents that stifle other research in the area.
What are these "bats" you are talking about? Since I haven't played GTA3 I do not know what it is. And how would you go about "beating" someone with them? Interesting concept, apparently there's still a lot of inventing going on in the gaming industry these days. Inventing new ways of killing people and all that.
And sarcasm aside. Yes there is a big fucking difference between using a gun and a bat. If a 6 year old tries to beat me with a bat I'll give him a throuough spanking which he apparently needs. If he comes at me with a rifle that could prove quite a bit harder.
If you think this engine was neat then check out Blueberry3D. It's also a terrain engine but it works on fractals instead of meshes. So you don't need a gigantic mesh database which you then cull or use LOD on. Instead you generate all of the terrain from fractal seeds.
There are a few movies you can download from their site and they are quite impressive. Particularly the trees IMHO. (Which look very real, even close up.)
There are already a bunch of OSS game engines. While more don't hurt it doesn't really help either.
AugstWest's point was that what is really needed is some open minded artists who allow their models to be used in projects where there are ideas but no "artistic gifts".
What is needed is data for skeletal modelling (eg motion capture data), models, textures, levels etc.
And even if general meshes are not really what you want in a project it's better than something that looks butt ugly.
Now people like me (mainly coders) can do a lot of good work by supplying advanced tools and engines. Eg it could be neat to have support for parametric facial animations in a game. (So if you use voice communication it will appear as if the avatar was speaking.) Or continuing on the voice track you could alter voices to better suit the gender/race/species of character talking.
There are also many areas which requires a lot of creativity for programmers. AI is a good example. And coding is quite creative by it's nature.
An HTTP header is on the order of 100B that's not nearly enough to be put into multiple packets. So when the transfer starts you are still in the same position as for FTP.
And seriously, for files of any useable size the phase of the moon is likely to have more impact on transfer speed.
All browsers I've tried (Moz, Netscape, IE) use C:\ as a temporary cache for downloading. At least my C:\ tend to fill up pretty much so this can be a bother if I'm downloading big files.
It's really annoying when you have several GB free on the disk you want to download to and get an "out of space" error. (Or in some cases it just drops the transmission and leave you with either a corrupt file or no file.)
HTTP 1.0 should be slower than FTP, as it requires a new connection with each file. 1.1 has support for multiple files so this issue becomes pretty moot.
There are other rather big differances in implementations as well. But I doubt any are big enough to actually matter to a regular server today.
The big benefit with FTP is that you have user control. If you need permissions FTP is the way to go. If you want users to download multiple files then FTP is also usually more practical. And for big projects there are quite a lot of flexibility in FTP for having server farms and caches. (The computer serving the control connection doesn't have to be the same serving the data connection.)
That said if you need big files to a lot of users get something like BitTorrent or SwarmCast. More efficient than either HTTP or FTP. (But you still need a HTTP server for those.)
Slashdot Mirror
The XBox DVD does not read "backwards". It's a perfectly normal DVD player, though it's apparently not that good at reading CDRs.
You're probably thinking of the Gamecube which is non-standard and read the disc from the outside instead of from the innermost tracks first.
Yeah, I've heard that all new sATA drives are shipped with support for that. Also there's a big push among card makers to add the necessary supporting hardware on the cards.
;-)
Or not.
Ok I'm going to be honest with you, I can't be bothered to actually calculate how big a capacitor it would take to drive a HDD for a few seconds. My guestimate would be that you're looking at an additional box with capacitors next to your PC however.
Sounds like a good idea, until you calculate how much electricity regular household appliances require.
;-)
Just on your typical PC rig the monitor is likely the biggest power consumer. (If you have a fairly large CRT.)
If I were to turn on my 32" TV and stereo system a UPS would be drained before I even hit the DVD menu.
What you want for these types of applications are a diesel generator. They come in varying sizes and are much more realistic for this scenario.
Besides, you could try to hook it up to your heating system so you had warm water for "free" by collecting heat given off by the engine.
"Never underestimate the power of stupidity."
Considering that it's not hard to find people who are having the warm fussy feeling of security when they use Raid-0 I bet there's a lot of people that run Raid-[1+] (and no UPS) thinking they are safe from harm.
Yeah but a big write cache on the disk will make all of your fancy RAIDing useless. Because if the system can't reliably tell what parts have been written to disk or not (ie still in the cache) then it can't protect the data from a power outage.
That's why RAID is no substitute for UPS.
1) Yes, one-way functions are public key cryptography. When I was refering to "public key" in my post I should have written "public key infrastructure" which would have been more correct. Most people don't know the difference in any case.
If you are as clued in regarding cryto as you seem to think you are you'd also know that both symmetric and asymmetric systems are vulnerable and hard to use by themselves. To make a useful system you combine the two. Use an asymetric system to exchange keys, and here you can use a Diffie-Hellman or PKI system, and then use that key in a symmetric system to make it useable. (Since asymetric systems are way too slow to be used for actual data transfer.)
The combination allows you to securely exchange keys over an unsecure channel and to transfer data (reasonably) quickly.
The rest of your points are not really relevant to the situation (stopping someone from snooping your connection) but I'll adress them anyways.
2) While I don't quite agree that it's impossible to prove that no asymmetric system is secure there is no such proof today. If you can prove that a function you have is in fact impossible to break by any other means than brute force then you'd have a provably secure assymertic system. Today you just use a system which is secure within a specific time. (Eg somone sniffing a network need to break the data in pretty much realtime for it to be useful in these applications. Otherwise the data will just pile up and they'll have to start throwing it away.)
Fast number factorisation would indeed break a lot of public key systems currently in use. OTOH number factorization is IIRC a NP-complete problem, so when that day comes it will be such a huge breakthrough in computing that I doubt anyone will cry. (You'll just have to use a one-way function not based on products of primes.) And you make it sound as if figure out a really efficient way to generate lists of prime numbers is an easy task. Since there is an abundant lack of such algorithms you could deduce that it is in fact rather hard.
3) And yes, I'm quite aware that typical loopholes in crypto systems involve attacking the human elements. It can be because the algorithm for generating prime guesses is based on mouse movement (and most poeple just move the mouse in a clockwise circle to do this) or on bugs in the implementation.
This is why I put "implement protocol correctly" in the last sentence of my post.
I don't know if I'd call it "more mature" but it certainly is good.
The first few episodes I was a bit annoyed with how similar it was to Eva, but I got over it. You could say that if Eva has a religious background then RahXephon has musical influences. Ie music is used in RX in much the same way religion is in Eva. It's very good though. (And I recall watching a TV program and a choir sang one of the arias used in RX.)
RX is a bit easier to get into as well, the characters are more normal then in Eva. But before you call RX deeper you should check out some sites analyzing Eva, there's a lot of things in that show not immideately obvious. Calling it "childish" demonstrates that you haven't done this IMHO.
But I guess I'm influenced by the fact that Eva was one of the first anime series I ever watched. And it had a pretty big impact on me.
I think the point was that you'd encrypt all the protocol data. That way you can't fingerprint it just by looking at the headers. But since encryption is quite slow you might want to send the rest of the data unecrypted which would live it open for attacks. Hence the need for sniffing data.
OTOH if you already have a working encryption system you might as well extend it to the data. (Perhaps using a faster algorithm.) And in that case would be snoopers would be SOL.
To protect against MIM use PKI. (Public Key Infrastructure.)
Actually they probably can't do that. At least not without some pretty extreme hardware.
Typically you get to a point where you have to use RAM buffers to save data and then have multiple network listeners which swap so they can save the data to disk. If you have a large amount of traffic you soon get to a point where you can't store all data.
But sure, it's their network, so they can do what they want. Just as long as they don't mind me using encrypted channels.
What the AC was probably trying to say (but in a, shall we say, crude way) is that there is no key to extract.
Any encryption protocol worth it's salt (ie generally not those propriatary/secret protocols) is protected against this. Eg SSL or SSH which you can snoop all you want, but there's never a key sent in clear text across the channel. Neither is there any key to find in the source. Instead a key is agreed upon by the clients as they connect, but using "one way functions" which are hard and/or practically impossible to crack. You can also use public key encryption with it to add even more security and authentication to the system.
That's why he suggested that you should read about SSL. A cheap way is to just look at the relevant RFC, although it may be a bit complex if you don't have any experience in number theory. Another hint could be look at crypto sites such as Bruce Schneier's crypto-gram (counterpane.com).
In any case, if you use a well tested protocol and implement it correctly (not always trivial) then the system will be secure.
I read a topic about this on Kuro5hin and one poster came up with the following. I'll just quote directly (I saved a copy of the post, for moments like these.) Unfortunately I don't know who posted it so I can't give credit.
CIH was around in '98, so the basic concept has been round for quite some time already.
Just test if the URL(s) has been in any other article. If it has been add a link to those with a "Possible dupe" warning as the article is up for review.
My point is that it doesn't have to be a automatic system, that is really hard. Just add a quick fix which automates detecting possible dupes. That would kill off most of the dupes in any case.
Here at Star Fleet Ship Design we know how important tactice feedback is to our customers. A starship is a very complex piece of machinery to control. If you as a pilot don't get feedback on the ships status it is very easy to misjudge the situation. Furthermore a lot of useability testing showed that it was good for morale if the moron who got you into the mess was the first to blow up.
We also added the Intruder Friendly Computer System to the later models of ships. They include a number of backdoors into the system which will ensure that operators are on the lookout for patches in the future. Also it keeps the crew on their toes and on the lookout for intruders. On the odd chance of a mutiny the captain also has a fair chance to reclaim control of the ship.
We strive to life for star ship crews more interesting.
VTY,
--SFSD propaganda manager
As long as the child understands the difference between shooting someone in a game and shooting someone in real life I would say it's ok. The kid is going to see a lot of people getting shot or beaten on TV in any case so they need to understand that in any case.
;-)
Children are not stupid, and if you meet a child who thinks it would be fun to go and beat some cops then perhaps you should try talking to him. Not to scold him but try to see if he is in fact talking about beating real people and discuss the consequences of that. (And try to do it in a non patronizing way too, so that it might actually have an effect on him.)
When a messed up child goes and kills someone then it's not only the fault of the parents. It's also the fault of those who could have tried to set him straight.
And for the record I don't have kids of my own. (And I didn't intend this post to be as preachy as it might have turned out.
Vector processors are very much in use today. All current processors support in in some way through SSE and similar instructions, and the G4 AltiVec has a lot more to offer in the same area. Furthermore if you have a reasonably current graphics card then it uses vector processing as well.
Nobody said vector processors are dead. They just tend to be overkill for most applications. (And hence they are instead used as a type of co-processors.)
There is research going on in this field. Eg it would allow you to reconfigure part of a pipe-line while data is flowing through the chip.
This is in fact already possible, but the reconfiguration time for large parts of a chip is generally way to slow for it to be usable. But if you have a design which allows you to reconfigure only a very small part of the chip then it's doable during runtime. (Although you may need special boards to do it, I'm not sure how many developer boards actually support reconfiguration while running.)
The idea of having small premade parts is already in use by eg the RAW project at MIT. Doing runtime optimizations is probably never going to happen though because doing routing on a large FPGA can take days to complete.
One interesting project I found a few years back is the RAW project at MIT. It does pretty much the same thing but they are no longer using FPGAs. (They use chips which are similar to FPGAs but specified towards computation.) Their first prototypes used FPGAs though.
Seems like the "programming language" is similar to LabView and such schematic programming languages. (Eg in Matlab you have Simulink.) Apparently there's quite a lot of people who find that easier to work with.
Oh well, it's an interesting field. Let's just hope they don't get a bunch of ludicrous patents that stifle other research in the area.
What are these "bats" you are talking about? Since I haven't played GTA3 I do not know what it is. And how would you go about "beating" someone with them? Interesting concept, apparently there's still a lot of inventing going on in the gaming industry these days. Inventing new ways of killing people and all that.
And sarcasm aside. Yes there is a big fucking difference between using a gun and a bat. If a 6 year old tries to beat me with a bat I'll give him a throuough spanking which he apparently needs. If he comes at me with a rifle that could prove quite a bit harder.
If you think this engine was neat then check out Blueberry3D. It's also a terrain engine but it works on fractals instead of meshes. So you don't need a gigantic mesh database which you then cull or use LOD on. Instead you generate all of the terrain from fractal seeds.
There are a few movies you can download from their site and they are quite impressive. Particularly the trees IMHO. (Which look very real, even close up.)
There are already a bunch of OSS game engines. While more don't hurt it doesn't really help either.
AugstWest's point was that what is really needed is some open minded artists who allow their models to be used in projects where there are ideas but no "artistic gifts".
What is needed is data for skeletal modelling (eg motion capture data), models, textures, levels etc.
And even if general meshes are not really what you want in a project it's better than something that looks butt ugly.
Now people like me (mainly coders) can do a lot of good work by supplying advanced tools and engines. Eg it could be neat to have support for parametric facial animations in a game. (So if you use voice communication it will appear as if the avatar was speaking.) Or continuing on the voice track you could alter voices to better suit the gender/race/species of character talking.
There are also many areas which requires a lot of creativity for programmers. AI is a good example. And coding is quite creative by it's nature.
An HTTP header is on the order of 100B that's not nearly enough to be put into multiple packets. So when the transfer starts you are still in the same position as for FTP.
And seriously, for files of any useable size the phase of the moon is likely to have more impact on transfer speed.
Actually he has a big point here.
All browsers I've tried (Moz, Netscape, IE) use C:\ as a temporary cache for downloading. At least my C:\ tend to fill up pretty much so this can be a bother if I'm downloading big files.
It's really annoying when you have several GB free on the disk you want to download to and get an "out of space" error. (Or in some cases it just drops the transmission and leave you with either a corrupt file or no file.)
HTTP 1.0 should be slower than FTP, as it requires a new connection with each file. 1.1 has support for multiple files so this issue becomes pretty moot.
There are other rather big differances in implementations as well. But I doubt any are big enough to actually matter to a regular server today.
The big benefit with FTP is that you have user control. If you need permissions FTP is the way to go. If you want users to download multiple files then FTP is also usually more practical. And for big projects there are quite a lot of flexibility in FTP for having server farms and caches. (The computer serving the control connection doesn't have to be the same serving the data connection.)
That said if you need big files to a lot of users get something like BitTorrent or SwarmCast. More efficient than either HTTP or FTP. (But you still need a HTTP server for those.)