The center of weight is in the middle, the middle of the robot is extended a couple of inches beyond the 'stair'. So the robot should be unbalanced, right?
But since the two 'head' segments are nearly horizontal, and the two 'tail' segments are diagonal, the 'head' segments extend their mass further away from the tilt point, and thus shift the center of gravity to the 'head' of the robot.
Except that it forward-fills the gaps, and that is easy.
It stores a snapshot every now and then, and when you want to go back it actually goes forward from the first snapshot before the time index you want.
If you make snapshots 10 times a second and can forward-fill in a tenth of a second, the programmer will not notice this. (And if he does, make more snapshots)
In fact, collision detection between two arbitrary polygons is pretty simple, use the Z-buffer from the graphics card. If you draw a polygon, and the Z-value equals the Z-value in the buffer, it is touching another polygon. Pretty simple, but the problem is that graphics cards don't tell you this...
-XCom: Ufo defense (one of the best games I know) -Silent Storm (turn based 3d!)
Also the Knights of the Old Republic games (and Neverwinter Nights) with all the autopause features could work well. (Don't know about walking around though...)
Only if the protocol is not transparent... like your protocol. Come on, who sends a message first, then maybe a week later the key? Why send the message so soon then, nobody can do anything with it. I remember 1 protocol like that and it's not even serious (very secure though...).
And why is the authentication a red herring? What if the point to point link happens to be a fiber of 150km long (The longest experimental link up to now). Are you going to post sentries along the fiber to make sure nobody is tampering with it? Might as well put the message in a bottle and shoot it through the duct with air-pressure.
Anyways, this is going nowhere... You are a believer in QC and I'm not. QED
The key might not be send for weeks... what protocol are we talking about here?
And how is Bob going to be convinced the message arrived? Is Alice supposed to tell him that? All Alice can tell Bob is that a message arrived, from somebody, and she cannot tell Bob more, unless he sends the key. And then Bob knows that somebody claims to have received a message from somebody (she calls herself Alice, but it could be anyone). So both Bob and Alice don't know anything anyways. Eve can always make up a message from Alice saying "I recieved your message #MSGNUM#, the hash is ###". But it is meaningless without authentication.
This is because crypto is meaningless without authentication. And until there is some form of quantum authentication, this means quantum crypto will always have to stoop to a lower level of security by using normal computationally protected authentication schemes like RSA. And then that will be the weak spot.
>Doh! Eve would have to impersonate both Alice and Bob for every data transaction on every channel.
That is a given for active MITM... And usually cryptographers analyze just the one channel, after that the rest is just more of the same.
>Any channel that Eve cannot sit in the middle of will reveal that eve is in the middle of the quantum channel.
Since there is no quantum authentication yet, how will anybody know who anybody else is? And if you use non-quantum authentication, you will weaken the protocol to non-quantum level, might as well drop the quantum crypto then!
>Bob encrypts a message and sends it to Alice. What does Eve do? She can't read the message because Bob hasn't sent the key yet. Either she forwards it to Alice or she 'loses' it.
Or she keeps it until she gets the key...
>Later, Bob sends the key for the message, now she reads the message "here are my quantum settings, Alice, do they check out?". Does she forward the key and let Alice figure out the quantum channel is compromised?
No, she makes a shiny new message for Alice, with her own 'quantum settings' and her own key, and Alice will be none the wiser because she cannot see the difference between Bob and Eve (no authentication, remember...)
>Or 'lose' this too? Lost messages mean there is someone tampering, so Eve must pretend to BE Alice rather than just tapping into Alice's comms. This is much easier to detect.
No messages lost, Bob and Alice are both fooled... And pretending to BE Alice is quite easy when there is no authentication, there is nothing to detect!
Except that when Bob talks to Alice, Alice happens to be Eve. Oops! And since there isn't any quantum authentication yet, the quantum crypto adds precisely nothing! (since security is only as strong as it's weakest link).
I've said it a million times, and I guess I have to say a million times more: Quantum crypto doesn't protect against an active Monkey-in-the-middle attack! And thus it is not the perfect uncrackable holy grail everybody is so hyped up about.
Linear sweep isn't necessary, just compensate for the unlinearness in software. (or don't use the edges where compression is worst) The second dimension sweep is _very_ slow (85 Hz) so if you wanted to hack it, put the emitter and the mirror on silicon and then sweep the whole assemblage at 85 Hz.
But since they use a laser that is etched onto a silicon chip (think DLP projectors), the mirror can easily vibrate at a couple of hundred Khz.
Let's say I wan't to display 1280x1024 at 85 Hz then the mirror needs to vibrate at ~85 Khz.
From a DLP projector page:
DLP is a way to project a digital video signal to yield a bright, high quality image. The heart of the DLP system is a rectangular array of microscopic mirrors called DMDs, or Digital Micromirror Devices(TM). Each DMD(TM) is a square mirror about 16 um (16 microns, or millionths of a meter) on a side, and can flip ten degrees one way (on) or the other (off). The switching is controlled electrostatically, and takes about 2 uS (two microseconds).
2 uS switch time leads to 250 Khz sweeps, or 2560x2048 at 125 Hz!
Quantum Cryptography doesn't protect against an active Man-in-the-Middle attack, eg: I cut the fiber and put two quantum transcievers on it. This is because Quantum Cryptography cannot authenticate the endpoints. You can of course prevent this by using Public Key Crypto but then why use Quantum Crypto at all?
The cell phone system is designed with small cells (the towers you can see on buildings), this is why the phone is called a cellphone. The cells need to be small otherwise the cellphone needs to output too much power, which leads to bigger batteries and such. Adjacent cells use different frequencies so the cellphone can't interfere with more than one cell. Since a cellphone can usually only see the neighboring cells this works. Normal cells can handle between 40-80 calls (frequencies) at the same time, this is why in cities, where the cellphone density is higher, the cells themselves are smaller and closer together.
Now when you are using the phone on a plane, especially when close to landing or just after take-off, the cellphone 'sees' more cells because the cells are not hidden by buildings or the curvature of the earth. The cellphone is using 1 of the 40-80 frequencies on each of those cells! A cellphone in a plane at an altitude of 1 mile can easily reach 50 cells (especially above cities) and will use up a frequency slot on 1/4 of them (the other 3/4 are on other frequencies). The phone companies don't like this, and what is worse, you are interfering with calls made by other people (on the other cells).
The phonecompanies can use the logfiles from the cells to prove which phones were used in flight (flightpath and multi-cell coverage). They could fine you for it! (which they aren't doing right now, too expensive)
Are you telling me that BitTorrent dosn't verify the whole file using SHA-1 after verifying all the chunks?
If that is the case, and SHA-1 is broken, then BitTorent is sunk.
But I expect that BitTorrent does check the whole file using SHA-1. It would then detect there was an error, and maybe try to get the chunks again.
To break this scheme, you need to find a chunk that collides with a legit chunk, and at the same time makes a file that collides with the legit file. This is a lot harder;)
1. Somebody invents model for market that works (tm) 2. That somebody gets incredibly rich 3. Other somebodys notice 4. Other somebodys license/buy model 5. Everybody uses model 6. Market turns into model, so it doesn't work anymore 7. GOTO 1.
You actually have 2 drives on each channel? So when one drive fails, the other drive on that channel will be unreachable if you are lucky, or have data corruption if unlucky.
So one drive fails, the other fails as well and since RAID5 can only compensate for one failure the entire stack is down.
I'd advise you to buy two more ATA controllers and give each drive it's own channel.
Internet Explorer is a 'smart' browser. Therefore it totally ignores the.gif extension, instead it looks at the first couple of hundred bytes of the file to figure out what it is.
We've had tons of problems with IE rendering an.xhtml file as an.xml file because of this. (And yes, the HTML standard explicitly forbids this behaviour, but this is Microsoft we're talking about)
Well actually, if you use a compacting garbage collector, allocating stuff on the heap is exactly as expensive as allocating on the stack since the heap looks like a stack:)
This is because the compacting GC moves all the live object to the beginning of the heap, therefore allocating heap is as easy as incrementing a pointer.
And that false link up/down attack can be done by spoofing a RST to the TCP/IP session, thereby terminating the link.
With Cisco this is pretty easy to do because: -BGP window size is big -BGP source port for Cisco starts with 11000 (so try 11000, 11001, 11002)
But then again it is trivially simple to defend against:
-use MD5 auth on the BGP link -use loopback addresses for the BGP link -ask your peer to block all traffic to the BGP ports on your router except for their own, and do the same for your peer (ingress filters)
first two workarounds make the attack much more difficult, last one makes the attack impossible. All of the above workaround require that you coordinate the change with your BGP peers and can lead to downtime.
Little price analysis for Google (price per Gbyte)
server: $0 (they will use the servers they have) 160 GB SATA HDD: $99 new harddisk after three years: $12.50 (assuming storage doubling every 12 months) new harddisk after six years: $1.50 new harddisk after nine years: negible redundancy (RAID): $10 (assuming 11 disk RAID5 stripe or 500 disk two dimensional RAID4 matrix) SATA controller: $12.50 per port (cheap ass SATA controller for $50, or just use the controller on the motherboard)
grand total: $140.50 (for 160 GByte or 150 GiByte)
conclusion:
If you already have a big serverpark, redundant (and yes two dimensional RAID4 is very redundant) hardisk storage is very cheap ($1 per GiByte or less)
The $2/Gbyte quote probably relates to them keeping the data in two datacenters for redundancy. Doing RAID between datacenters (Red. Array of Inexpensive Datacenters) is possible, but difficult and the traffic overhead between the datacenters will probably cost you the savings you got by doing RAID between the datacenters.
But since harddisks have to spin faster all the time, they are making the platters smaller. The platters in a 10.000 RPM HDD are almost as small as the platters in a 2.5" HDD. This is why some people are predicting that there won't be any 3.5" HDDs anymore (like in 3 years)
This is true... if you use a good NUMA aware OS. Otherwise the application/data you are trying to run/crunch on four processors resides in the memory of one processor, and the other three need to bug it all the time to get at it (or worse invalidate each other's caches when writing to it).
So if you don't use a NUMA aware OS, the memory bandwidth could sink below that of a single CPU system.
Maybe you need to read the article before posting, or the Opteron technical docs for that matter. The x445 is a Xeon based server with some nasty hacks to get the Xeons to play nice with the multiple memory controllers.
The Opteron has a built in memory controller, and when you build multiprocessor systems, each CPU gets its own memory. The HyperTransport chip interconnects make sure all the CPU's can reach each other's memory, and has been designed from the ground up to be as low latency as possible.
Read this presentation before you try to convince us Xeons in a x445 are a better NUMA solution:
Slashdot Mirror
The center of weight is in the middle, the middle of the robot is extended a couple of inches beyond the 'stair'. So the robot should be unbalanced, right?
But since the two 'head' segments are nearly horizontal, and the two 'tail' segments are diagonal, the 'head' segments extend their mass further away from the tilt point, and thus shift the center of gravity to the 'head' of the robot.
HTH
--Blerik
Except that it forward-fills the gaps, and that is easy.
It stores a snapshot every now and then, and when you want to go back it actually goes forward from the first snapshot before the time index you want.
If you make snapshots 10 times a second and can forward-fill in a tenth of a second, the programmer will not notice this. (And if he does, make more snapshots)
HTH
--Blerik
In fact, collision detection between two arbitrary polygons is pretty simple, use the Z-buffer from the graphics card. If you draw a polygon, and the Z-value equals the Z-value in the buffer, it is touching another polygon. Pretty simple, but the problem is that graphics cards don't tell you this...
--Blerik
Try some turn based strategy games like:
-XCom: Ufo defense (one of the best games I know)
-Silent Storm (turn based 3d!)
Also the Knights of the Old Republic games (and Neverwinter Nights) with all the autopause features could work well. (Don't know about walking around though...)
HTH
--Blerik
Only if the protocol is not transparent... like your protocol. Come on, who sends a message first, then maybe a week later the key? Why send the message so soon then, nobody can do anything with it. I remember 1 protocol like that and it's not even serious (very secure though...).
And why is the authentication a red herring?
What if the point to point link happens to be a fiber of 150km long (The longest experimental link up to now). Are you going to post sentries along the fiber to make sure nobody is tampering with it? Might as well put the message in a bottle and shoot it through the duct with air-pressure.
Anyways, this is going nowhere... You are a believer in QC and I'm not. QED
--Blerik
The key might not be send for weeks... what protocol are we talking about here?
And how is Bob going to be convinced the message arrived? Is Alice supposed to tell him that? All Alice can tell Bob is that a message arrived, from somebody, and she cannot tell Bob more, unless he sends the key. And then Bob knows that somebody claims to have received a message from somebody (she calls herself Alice, but it could be anyone). So both Bob and Alice don't know anything anyways. Eve can always make up a message from Alice saying "I recieved your message #MSGNUM#, the hash is ###". But it is meaningless without authentication.
This is because crypto is meaningless without authentication. And until there is some form of quantum authentication, this means quantum crypto will always have to stoop to a lower level of security by using normal computationally protected authentication schemes like RSA. And then that will be the weak spot.
--Blerik
>Doh! Eve would have to impersonate both Alice and Bob for every data transaction on every channel.
That is a given for active MITM... And usually cryptographers analyze just the one channel, after that the rest is just more of the same.
>Any channel that Eve cannot sit in the middle of will reveal that eve is in the middle of the quantum channel.
Since there is no quantum authentication yet, how will anybody know who anybody else is? And if you use non-quantum authentication, you will weaken the protocol to non-quantum level, might as well drop the quantum crypto then!
>Bob encrypts a message and sends it to Alice. What does Eve do? She can't read the message because Bob hasn't sent the key yet. Either she forwards it to Alice or she 'loses' it.
Or she keeps it until she gets the key...
>Later, Bob sends the key for the message, now she reads the message "here are my quantum settings, Alice, do they check out?". Does she forward the key and let Alice figure out the quantum channel is compromised?
No, she makes a shiny new message for Alice, with her own 'quantum settings' and her own key, and Alice will be none the wiser because she cannot see the difference between Bob and Eve (no authentication, remember...)
>Or 'lose' this too? Lost messages mean there is someone tampering, so Eve must pretend to BE Alice rather than just tapping into Alice's comms. This is much easier to detect.
No messages lost, Bob and Alice are both fooled...
And pretending to BE Alice is quite easy when there is no authentication, there is nothing to detect!
--Blerik
Except that when Bob talks to Alice, Alice happens to be Eve. Oops! And since there isn't any quantum authentication yet, the quantum crypto adds precisely nothing! (since security is only as strong as it's weakest link).
I've said it a million times, and I guess I have to say a million times more: Quantum crypto doesn't protect against an active Monkey-in-the-middle attack! And thus it is not the perfect uncrackable holy grail everybody is so hyped up about.
Nothing to see here, move along...
--Blerik
Linear sweep isn't necessary, just compensate for the unlinearness in software. (or don't use the edges where compression is worst) The second dimension sweep is _very_ slow (85 Hz) so if you wanted to hack it, put the emitter and the mirror on silicon and then sweep the whole assemblage at 85 Hz.
--Blerik
Typo!
They use a _mirror_ that is etched
--Blerik
But since they use a laser that is etched onto a silicon chip (think DLP projectors), the mirror can easily vibrate at a couple of hundred Khz.
Let's say I wan't to display 1280x1024 at 85 Hz then the mirror needs to vibrate at ~85 Khz.
From a DLP projector page:
DLP is a way to project a digital video signal to yield a bright, high quality image. The heart of the DLP system is a rectangular array of microscopic mirrors called DMDs, or Digital Micromirror Devices(TM). Each DMD(TM) is a square mirror about 16 um (16 microns, or millionths of a meter) on a side, and can flip ten degrees one way (on) or the other (off). The switching is controlled electrostatically, and takes about 2 uS (two microseconds).
2 uS switch time leads to 250 Khz sweeps, or 2560x2048 at 125 Hz!
--Blerik
First rule of commenting code:
Comment why not what.
HTH
--Blerik
This is not true!
Quantum Cryptography doesn't protect against an active Man-in-the-Middle attack, eg: I cut the fiber and put two quantum transcievers on it. This is because Quantum Cryptography cannot authenticate the endpoints. You can of course prevent this by using Public Key Crypto but then why use Quantum Crypto at all?
--Blerik
The cell phone system is designed with small cells (the towers you can see on buildings), this is why the phone is called a cellphone. The cells need to be small otherwise the cellphone needs to output too much power, which leads to bigger batteries and such. Adjacent cells use different frequencies so the cellphone can't interfere with more than one cell. Since a cellphone can usually only see the neighboring cells this works. Normal cells can handle between 40-80 calls (frequencies) at the same time, this is why in cities, where the cellphone density is higher, the cells themselves are smaller and closer together.
Now when you are using the phone on a plane, especially when close to landing or just after take-off, the cellphone 'sees' more cells because the cells are not hidden by buildings or the curvature of the earth. The cellphone is using 1 of the 40-80 frequencies on each of those cells! A cellphone in a plane at an altitude of 1 mile can easily reach 50 cells (especially above cities) and will use up a frequency slot on 1/4 of them (the other 3/4 are on other frequencies). The phone companies don't like this, and what is worse, you are interfering with calls made by other people (on the other cells).
The phonecompanies can use the logfiles from the cells to prove which phones were used in flight (flightpath and multi-cell coverage). They could fine you for it! (which they aren't doing right now, too expensive)
HTH
--Blerik
Are you telling me that BitTorrent dosn't verify the whole file using SHA-1 after verifying all the chunks?
;)
If that is the case, and SHA-1 is broken, then BitTorent is sunk.
But I expect that BitTorrent does check the whole file using SHA-1. It would then detect there was an error, and maybe try to get the chunks again.
To break this scheme, you need to find a chunk that collides with a legit chunk, and at the same time makes a file that collides with the legit file. This is a lot harder
--Blerik
It actually works like this:
1. Somebody invents model for market that works (tm)
2. That somebody gets incredibly rich
3. Other somebodys notice
4. Other somebodys license/buy model
5. Everybody uses model
6. Market turns into model, so it doesn't work anymore
7. GOTO 1.
--Blerik
You actually have 2 drives on each channel? So when one drive fails, the other drive on that channel will be unreachable if you are lucky, or have data corruption if unlucky.
So one drive fails, the other fails as well and since RAID5 can only compensate for one failure the entire stack is down.
I'd advise you to buy two more ATA controllers and give each drive it's own channel.
--Blerik
Nope,
.gif extension, instead it looks at the first couple of hundred bytes of the file to figure out what it is.
.xhtml file as an .xml file because of this. (And yes, the HTML standard explicitly forbids this behaviour, but this is Microsoft we're talking about)
Internet Explorer is a 'smart' browser. Therefore it totally ignores the
We've had tons of problems with IE rendering an
--Blerik
Well actually, if you use a compacting garbage collector, allocating stuff on the heap is exactly as expensive as allocating on the stack since the heap looks like a stack :)
a ry /j-jtp10283/
This is because the compacting GC moves all the live object to the beginning of the heap, therefore allocating heap is as easy as incrementing a pointer.
See this page:
http://www-106.ibm.com/developerworks/java/libr
--Blerik
And that false link up/down attack can be done by spoofing a RST to the TCP/IP session, thereby terminating the link.
With Cisco this is pretty easy to do because:
-BGP window size is big
-BGP source port for Cisco starts with 11000 (so try 11000, 11001, 11002)
But then again it is trivially simple to defend against:
-use MD5 auth on the BGP link
-use loopback addresses for the BGP link
-ask your peer to block all traffic to the BGP ports on your router except for their own, and do the same for your peer (ingress filters)
first two workarounds make the attack much more difficult, last one makes the attack impossible. All of the above workaround require that you coordinate the change with your BGP peers and can lead to downtime.
HTH
--Blerik
Little price analysis for Google (price per Gbyte)
server: $0 (they will use the servers they have)
160 GB SATA HDD: $99
new harddisk after three years: $12.50 (assuming storage doubling every 12 months)
new harddisk after six years: $1.50
new harddisk after nine years: negible
redundancy (RAID): $10 (assuming 11 disk RAID5 stripe or 500 disk two dimensional RAID4 matrix)
SATA controller: $12.50 per port (cheap ass SATA controller for $50, or just use the controller on the motherboard)
grand total: $140.50 (for 160 GByte or 150 GiByte)
conclusion:
If you already have a big serverpark, redundant (and yes two dimensional RAID4 is very redundant) hardisk storage is very cheap ($1 per GiByte or less)
The $2/Gbyte quote probably relates to them keeping the data in two datacenters for redundancy. Doing RAID between datacenters (Red. Array of Inexpensive Datacenters) is possible, but difficult and the traffic overhead between the datacenters will probably cost you the savings you got by doing RAID between the datacenters.
--Blerik
Nice and easy solution. Only problem is that multicasting on the internet is hard.
try googling for MBONE for more information.
--Blerik
But since harddisks have to spin faster all the time, they are making the platters smaller. The platters in a 10.000 RPM HDD are almost as small as the platters in a 2.5" HDD. This is why some people are predicting that there won't be any 3.5" HDDs anymore (like in 3 years)
--Blerik
This is true... if you use a good NUMA aware OS. Otherwise the application/data you are trying to run/crunch on four processors resides in the memory of one processor, and the other three need to bug it all the time to get at it (or worse invalidate each other's caches when writing to it).
So if you don't use a NUMA aware OS, the memory bandwidth could sink below that of a single CPU system.
--Blerik
Maybe you need to read the article before posting, or the Opteron technical docs for that matter. The x445 is a Xeon based server with some nasty hacks to get the Xeons to play nice with the multiple memory controllers.
g -c olloquia/talks/Opteron 09.11.03.ppt
The Opteron has a built in memory controller, and when you build multiprocessor systems, each CPU gets its own memory. The HyperTransport chip interconnects make sure all the CPU's can reach each other's memory, and has been designed from the ground up to be as low latency as possible.
Read this presentation before you try to convince us Xeons in a x445 are a better NUMA solution:
http://computing-colloquia.web.cern.ch/computin
check out pages 22 and 23 for a nice comparison between legacy (read Intel) and Opteron multiprocessor memory architecture...
--Blerik