It's a shame that their business model is flawed. Given that they have two products software (X) and hardware (Y) that they always sell together they are charging X+Y to the customer. There is nothing to stop them shifting some of the price from one onto the other. So they could change the split and use the software cost to subsidise the hardware (more than they do). Phystar and the Hackintosh community would be screwed, and any business that they did do would lower the cost of Apple hardware getting them more mainstream sales.
Instead, they've so far resorted to crappy legal tactics and some shoddy technical hacks. It's sad really, they should be playing to win rather than arguing over whose ball it is.
So first you berate people for not reading the article, then when they point out it contradicts you suddenly it's wrong? You have to be a troll....
Ignore my other reply about accumulated error, it turns that I was wrong. The software problem was actually as simple as the loss of precision in the fixed-point multiply. Very simple but complete description is available here.
The problem described is not overflow, it is repeated rounding on the imprecise representation of 0.1. The systems failed after 3.6M ticks. In a 24-bit register overflow is not a problem until at least 16.7M ticks. The lowest bound is because this is not an integer register and the article does not describe the size of the exponent. If you check the figures in the article, when the system failed it was out by about three ticks in 3.6M. Overflow causes the representation to suddenly shift to a completely wrong value. Subtle shifts to nearby values are symptomatic of rounding error.
Rounding is an issue. This particular example is classic textbook stuff and been used in many a software engineering course. Using a 24-bit floating point representation store 0.1. The error will be roughly 1/(2^23). Now repeatedly accumulate this value. The problem occurs because each time the values goes past a power of two boundary in magnitude we need to round the accumulator causing a slight loss of accuracy. Then for each subsequent addition we are adding a less precise representation of 0.1 until they are more likely to round up rather than down.
The system has three sources of accuracy that anybody with experience in floating arithmetic could have pointed out easily: 1. The initial representation of 0.1 is off by a tiny amount (very small impact on the final value) 2. As the accumulator increases in magnitude it's exponent rises, in this case by log_2(3.6M) = 22 places. 3. The subsequents additions of the small value to the much larger value become increasingly imprecise.
The problem is not the clock itself, nor some integer accumulation of the time: it is a designer who chose to use a floating point accumulator. Multiplying the representation of 0.1 by the integer number of ticks at each stage would have eliminated the problem. Accumulating 1/8 ticks in floating point would have worked fine. Doing what they did was stupid.
My memory of how a TPM works is probably off, but if I remember correctly the system passes a hash of the firmware to the TPM, which returns a key to open the system. There is still a physical component that you trust that hashes the firmware before execution. The "dummy" TPM intercepts it, passes on the hash of the real firmware, and then proceeds with the boot.
But the TPM itself has to be trusted - there is no way to verify it. Perhaps the "Evil Maid" isn't that skilled with a soldering iron, but this is a real attack vector. In the UK they decided not to verify the card readers for Chip'n'PIN and as a result attacks using fake readers started to spring up.
Removing the TPM and replacing it with a dummy would allow the introduction of the evil firmware. Whether or not the Maid could accomplish that task while you're in the bathroom is up for debate.
Unfortunately in the security world you can drive a horse and carriage through "almost".
The remaining hole is that the TPM does not verify itself to the user and so you can't really believe that it is doing what it should be doing once it has left your possession. Of course you could make some sort of trusted module to verify the TPM, but it's turtles all the way down...
Sounds cool. I'm seeing some kind of voice interface to get around the lack of writing. To be honest it would be most useful as some sort of mobile platform, rather than desktop software. Then I could easily carry this voice communication around, and hope that it became pervasive enough that all of my friends did too. Although the email paradigm of bouncing messages off of each other works well it would be really good to have a real-time interface for voice between these mobile devices.
Perhaps a speaker, microphone and some buttons to select who I want to talk to. You know this could be really huge.
It's really very simple. Our eyes only capture 2D images. For our brains to create a 3D scene there need to be certain cues present that the brain has evolved to trigger from. One cue is depth - but it is not the only cue, nor is it the only important one. Parallax is also a very important cue without which the brain will not fully (or perhaps) effectively generate a 3D scene.
The reason that I use immersion is because if something is immersive then clearly the cues are correct because nothing is breaking the illusion.
In your second example there is no complex processing required by the brain to convert the 2D image that the retina sees, of the 2D scene, into the brain's internal 2D representation. But for 3D scenes the brain has to compensate for the fact that there is an extra dimension on the outside that it needs to recreate on the inside, using only a 2D channel between the two.
Take a deep breadth, have another think about what fyngyrz said and consider for a moment who the idiot is. The whole point of these technologies is immersion. Clearly depth perception + parallax is more immersive than depth perception alone. A "real" 3D experience - i.e one which provides visual immersion in a 3D world will be one that does not provide jarring cues to break you out of that immersion. The failure to provide parallax stops it being full 3D.
As you've provided two posts without any actual content answer a simple question: have you used one of these systems, and/or a system that offers parallax? Do you understand the difference that you are arguing is unimportant or are you just guessing?
There is a fairground ride near here that offers this type of stero vision. It is entertaining, but it is certainly not 3D and the brain compensates after just a few seconds to realise that. I've also played with systems that offer parallax as we built one as a research prototype a few years ago. They are far superior, yes it does make a real difference if you use your real hard-wired into your brain neck muscles, and no it really doesn't matter if it wouldn't work for Hawking or how unfortunate his condition is as that is irrelevant to the debate.
The idea that any improvement in infrastructure is only for the money is not even a bad thing. I would want infrastructure improvements to save money for the simple reason that we use money as a proxy for resources, and if somebody is going to expend huge amounts of resources on upgrading something then I want it to be more efficient as a result. Of course I would treat safety as a given in any design (that is everything has to meet safety constraints to even be considered by cost effectiveness).
One disadvantage of seeing markets in everything is that you become aware of the holes. (Sssh don't tell the rabid libertarians on slashdot but markets are not perfect allocators of resources). So while you are correct in that knocking out this section doesn't *currently* take all three grids down I would expect that to remain the case. The idea of connecting the three grids is to increase the size of the market-place - in particular resources that are only available part-time can be sold to a wider market to ensure 100% usage during those times.
The problem is that redundancy will be reduced by normal market operations. It's not in my local interest to leave this power station here in Texas idling so that the grid has redundancy. If I'm not getting paid I'll shut it down. In the three smaller markets this wouldn't have happened because I wasn't competing with distant supplies. But once it has happened, if you shut down the interconnect then demand will exceed supply in my area where before it would not. Hence knocking out that section will take the three grids down once market conditions have adapted to expect it being there.
ONE OF the problems with space travel, that has been true for the past 60 years since the first rockets reached the edge of space, has been it costs a HUGE amount of finite resources to get anything into orbit. At least $10,000 a kilogram for a man rated launcher. Better engines that only work out in space do utterly nothing to solve this problem.
Spot the logical fallacy after I've corrected your basic error. With the technology to perform (unmanned) interplanetary missions and retrieve resources from around the solar system the amount of raw stuff that we need to hoist up the gravity well diminishes considerably. Currently if want to attempt a manned interplanetary mission we need to lift every last gram that it needs from the surface. Orbital manufacturing and resource retrieval are orders of magnitude more important than improving our capability to lift things into orbit - because they reduce the amount that we need to lift by orders of magnitude.
Perhaps you are unusual in that you seem to be picking dictionary words uniformly at random, leading to an average complexity of half the search space. Sadly most people are not very good at picking random numbers and if you told them to use this method the probability is exactly one that they would choose fuckdonkey69...
Joking aside, using John is very good advice. It actually sorts the search space to pick common layouts of dictionary words plus fillers first. If it can't find your password then it is pretty secure.
Indeed, just because this early prototype has 75% efficiency we must assume that the maximum that can ever be achieved. Best to just stop investigating it instead of working on improving the range and efficiency. After all scientific progress has advanced quite far enough hasn't it?
God forbid that we improve this technology and use it to replace other sources of loss to reduce energy consumption! After all we are rapidly moving towards an electric infrastructure for vehicles, and they are always this close to the road. Imagine just how bad it would be for global warming if we replaced batteries (and their associated losses) with this technology. Evil scientists.
One is about a man chasing love and escape in a down-trodden totalitarian state. The other is about a man chasing a totalitarian state in a blissed out utopia. They pursue similar themes, but are clearly different stories.
Anyway, that aside, now I can see what you were saying in your first post. Yes, there are very fuzzy boundaries between different ideas due in part to the fact that they are all derived from one another.
Are you really linking to an article that debunks the 7 basic plots myth as part of an argument that we have only seven basic plots? Or are the missing words in your third sentence critical?
What is the benefit to society of allowing patent protection on software?
The benefits of allowing a monopoly on designs for physical processes is to overcome the high barriers of entry that such designs must overcome. Yet software has quite low barriers to entry and designs enjoy more network effects.
Sure. I wouldn't dispute that for a second. It has been hyped as the second coming of online collaboration. But what they didn't claim was that n-way conversations would magically become as simple as 2-way conversations. I think that where DerekLyons was trolling was in claiming that they claimed they would take many types of interaction and collaboration and make them as simple as 2-way conversations.
Your other reply interested me as well. That sounds like an interface issue. It's hard to separate out the potential of the underlying framework, with their v0.9 app stuck on top of it for this beta trial. It might take quite a few versions to iron out those sorts of kinks.
The examples in the review seemed to be treating it like a forum, or possible a usenet thread. I wonder how much of their current interface is tailored to small group IM/email replacement instead of large-group forum/usenet replacement?
Doesn't really qualify as an answer. Point out where Google claimed that they would magically take n-party interactions and make them as simple as two-party. No answer, eh?
Translation: "I don't agree with this review, and thus the reviewer is at fault and ignorant for not agreeing with me. Even though he has seen the software and I... haven't".
Translation: I've decided the GP is wrong without having any evidence, so to rationalise my decision I'll flame him like a bitch.
It seems that a lot of the early reviews are complaining that when they use like a real-time forum, it gets too busy. When a reviewer claims that he's chatting to 12 people at once and it's too much of a time sink - what is he comparing it to? Chatting to 12 people in a normal IM client is a huge time sink because there is always somebody talking.
I'd like to read a review by somebody that knows what that they're talking about. Sure, it's a tool that tries to integrate blogs / forums / chat / email into a single product. But that doesn't magically mean that it can turn forum style interaction between hundreds of people into a linear two-person conversion like email.
If anything, the combination is going to create different conventions for hybrid forms of communication.
Slashdot Mirror
It's a shame that their business model is flawed. Given that they have two products software (X) and hardware (Y) that they always sell together they are charging X+Y to the customer. There is nothing to stop them shifting some of the price from one onto the other. So they could change the split and use the software cost to subsidise the hardware (more than they do). Phystar and the Hackintosh community would be screwed, and any business that they did do would lower the cost of Apple hardware getting them more mainstream sales.
Instead, they've so far resorted to crappy legal tactics and some shoddy technical hacks. It's sad really, they should be playing to win rather than arguing over whose ball it is.
Of course that is the solution they used that caused the problem.
So first you berate people for not reading the article, then when they point out it contradicts you suddenly it's wrong? You have to be a troll....
Ignore my other reply about accumulated error, it turns that I was wrong. The software problem was actually as simple as the loss of precision in the fixed-point multiply. Very simple but complete description is available here.
The problem described is not overflow, it is repeated rounding on the imprecise representation of 0.1. The systems failed after 3.6M ticks. In a 24-bit register overflow is not a problem until at least 16.7M ticks. The lowest bound is because this is not an integer register and the article does not describe the size of the exponent. If you check the figures in the article, when the system failed it was out by about three ticks in 3.6M. Overflow causes the representation to suddenly shift to a completely wrong value. Subtle shifts to nearby values are symptomatic of rounding error.
Rounding is an issue. This particular example is classic textbook stuff and been used in many a software engineering course. Using a 24-bit floating point representation store 0.1. The error will be roughly 1/(2^23). Now repeatedly accumulate this value. The problem occurs because each time the values goes past a power of two boundary in magnitude we need to round the accumulator causing a slight loss of accuracy. Then for each subsequent addition we are adding a less precise representation of 0.1 until they are more likely to round up rather than down.
The system has three sources of accuracy that anybody with experience in floating arithmetic could have pointed out easily:
1. The initial representation of 0.1 is off by a tiny amount (very small impact on the final value)
2. As the accumulator increases in magnitude it's exponent rises, in this case by log_2(3.6M) = 22 places.
3. The subsequents additions of the small value to the much larger value become increasingly imprecise.
The problem is not the clock itself, nor some integer accumulation of the time: it is a designer who chose to use a floating point accumulator. Multiplying the representation of 0.1 by the integer number of ticks at each stage would have eliminated the problem. Accumulating 1/8 ticks in floating point would have worked fine. Doing what they did was stupid.
Is that the same William Stein who writes Sage commenting in an article about numerical impression? I'm impressed, slashdot still has style... :)
Personally I would have used 1/8 second ticks instead of a Python symbolic algebra package, but I'm old school like that.
And it's pretty well behaved as long as you remember the first commandment:
My memory of how a TPM works is probably off, but if I remember correctly the system passes a hash of the firmware to the TPM, which returns a key to open the system. There is still a physical component that you trust that hashes the firmware before execution. The "dummy" TPM intercepts it, passes on the hash of the real firmware, and then proceeds with the boot.
OK. Good correction, time is an important factor.
But the TPM itself has to be trusted - there is no way to verify it. Perhaps the "Evil Maid" isn't that skilled with a soldering iron, but this is a real attack vector. In the UK they decided not to verify the card readers for Chip'n'PIN and as a result attacks using fake readers started to spring up.
Removing the TPM and replacing it with a dummy would allow the introduction of the evil firmware. Whether or not the Maid could accomplish that task while you're in the bathroom is up for debate.
Unfortunately in the security world you can drive a horse and carriage through "almost".
The remaining hole is that the TPM does not verify itself to the user and so you can't really believe that it is doing what it should be doing once it has left your possession. Of course you could make some sort of trusted module to verify the TPM, but it's turtles all the way down...
No. If you don't have physical security then you have nothing, and you can't ever get around that fact.
In your example temporary access to the laptop would allow removal of the TPM, replacement with a dummy, and then execution of the Evil Maid attack.
Sounds cool. I'm seeing some kind of voice interface to get around the lack of writing. To be honest it would be most useful as some sort of mobile platform, rather than desktop software. Then I could easily carry this voice communication around, and hope that it became pervasive enough that all of my friends did too. Although the email paradigm of bouncing messages off of each other works well it would be really good to have a real-time interface for voice between these mobile devices.
Perhaps a speaker, microphone and some buttons to select who I want to talk to. You know this could be really huge.
It's really very simple. Our eyes only capture 2D images. For our brains to create a 3D scene there need to be certain cues present that the brain has evolved to trigger from. One cue is depth - but it is not the only cue, nor is it the only important one. Parallax is also a very important cue without which the brain will not fully (or perhaps) effectively generate a 3D scene.
The reason that I use immersion is because if something is immersive then clearly the cues are correct because nothing is breaking the illusion.
In your second example there is no complex processing required by the brain to convert the 2D image that the retina sees, of the 2D scene, into the brain's internal 2D representation. But for 3D scenes the brain has to compensate for the fact that there is an extra dimension on the outside that it needs to recreate on the inside, using only a 2D channel between the two.
Take a deep breadth, have another think about what fyngyrz said and consider for a moment who the idiot is. The whole point of these technologies is immersion. Clearly depth perception + parallax is more immersive than depth perception alone. A "real" 3D experience - i.e one which provides visual immersion in a 3D world will be one that does not provide jarring cues to break you out of that immersion. The failure to provide parallax stops it being full 3D.
As you've provided two posts without any actual content answer a simple question: have you used one of these systems, and/or a system that offers parallax? Do you understand the difference that you are arguing is unimportant or are you just guessing?
There is a fairground ride near here that offers this type of stero vision. It is entertaining, but it is certainly not 3D and the brain compensates after just a few seconds to realise that. I've also played with systems that offer parallax as we built one as a research prototype a few years ago. They are far superior, yes it does make a real difference if you use your real hard-wired into your brain neck muscles, and no it really doesn't matter if it wouldn't work for Hawking or how unfortunate his condition is as that is irrelevant to the debate.
The idea that any improvement in infrastructure is only for the money is not even a bad thing. I would want infrastructure improvements to save money for the simple reason that we use money as a proxy for resources, and if somebody is going to expend huge amounts of resources on upgrading something then I want it to be more efficient as a result. Of course I would treat safety as a given in any design (that is everything has to meet safety constraints to even be considered by cost effectiveness).
One disadvantage of seeing markets in everything is that you become aware of the holes. (Sssh don't tell the rabid libertarians on slashdot but markets are not perfect allocators of resources). So while you are correct in that knocking out this section doesn't *currently* take all three grids down I would expect that to remain the case. The idea of connecting the three grids is to increase the size of the market-place - in particular resources that are only available part-time can be sold to a wider market to ensure 100% usage during those times.
The problem is that redundancy will be reduced by normal market operations. It's not in my local interest to leave this power station here in Texas idling so that the grid has redundancy. If I'm not getting paid I'll shut it down. In the three smaller markets this wouldn't have happened because I wasn't competing with distant supplies. But once it has happened, if you shut down the interconnect then demand will exceed supply in my area where before it would not. Hence knocking out that section will take the three grids down once market conditions have adapted to expect it being there.
I need a secret underground lair built. Do you think you could design it for me?
Spot the logical fallacy after I've corrected your basic error. With the technology to perform (unmanned) interplanetary missions and retrieve resources from around the solar system the amount of raw stuff that we need to hoist up the gravity well diminishes considerably. Currently if want to attempt a manned interplanetary mission we need to lift every last gram that it needs from the surface. Orbital manufacturing and resource retrieval are orders of magnitude more important than improving our capability to lift things into orbit - because they reduce the amount that we need to lift by orders of magnitude.
Perhaps you are unusual in that you seem to be picking dictionary words uniformly at random, leading to an average complexity of half the search space. Sadly most people are not very good at picking random numbers and if you told them to use this method the probability is exactly one that they would choose fuckdonkey69...
Joking aside, using John is very good advice. It actually sorts the search space to pick common layouts of dictionary words plus fillers first. If it can't find your password then it is pretty secure.
That's completely wrong on so many counts that it doesn't even deserve the criticism. Have you got a copy of the book?
Indeed, just because this early prototype has 75% efficiency we must assume that the maximum that can ever be achieved. Best to just stop investigating it instead of working on improving the range and efficiency. After all scientific progress has advanced quite far enough hasn't it?
God forbid that we improve this technology and use it to replace other sources of loss to reduce energy consumption! After all we are rapidly moving towards an electric infrastructure for vehicles, and they are always this close to the road. Imagine just how bad it would be for global warming if we replaced batteries (and their associated losses) with this technology. Evil scientists.
One is about a man chasing love and escape in a down-trodden totalitarian state. The other is about a man chasing a totalitarian state in a blissed out utopia. They pursue similar themes, but are clearly different stories.
Anyway, that aside, now I can see what you were saying in your first post. Yes, there are very fuzzy boundaries between different ideas due in part to the fact that they are all derived from one another.
Are you really linking to an article that debunks the 7 basic plots myth as part of an argument that we have only seven basic plots? Or are the missing words in your third sentence critical?
What is the benefit to society of allowing patent protection on software?
The benefits of allowing a monopoly on designs for physical processes is to overcome the high barriers of entry that such designs must overcome. Yet software has quite low barriers to entry and designs enjoy more network effects.
Sure. I wouldn't dispute that for a second. It has been hyped as the second coming of online collaboration. But what they didn't claim was that n-way conversations would magically become as simple as 2-way conversations. I think that where DerekLyons was trolling was in claiming that they claimed they would take many types of interaction and collaboration and make them as simple as 2-way conversations.
Your other reply interested me as well. That sounds like an interface issue. It's hard to separate out the potential of the underlying framework, with their v0.9 app stuck on top of it for this beta trial. It might take quite a few versions to iron out those sorts of kinks.
The examples in the review seemed to be treating it like a forum, or possible a usenet thread. I wonder how much of their current interface is tailored to small group IM/email replacement instead of large-group forum/usenet replacement?
Doesn't really qualify as an answer. Point out where Google claimed that they would magically take n-party interactions and make them as simple as two-party. No answer, eh?
Translation: I've decided the GP is wrong without having any evidence, so to rationalise my decision I'll flame him like a bitch.
It seems that a lot of the early reviews are complaining that when they use like a real-time forum, it gets too busy. When a reviewer claims that he's chatting to 12 people at once and it's too much of a time sink - what is he comparing it to? Chatting to 12 people in a normal IM client is a huge time sink because there is always somebody talking.
I'd like to read a review by somebody that knows what that they're talking about. Sure, it's a tool that tries to integrate blogs / forums / chat / email into a single product. But that doesn't magically mean that it can turn forum style interaction between hundreds of people into a linear two-person conversion like email.
If anything, the combination is going to create different conventions for hybrid forms of communication.