I would rather dive into a swimming pool full of Ebola, triple-edged razor blades, lemon juice, and pig doots than use any software from that "R" named firm.
Every time I've been stupid and downloaded their crap, thinking, "it used to be sooooo bad, any improvement would be a treeemendous improvement", I've had to delete it within the hour. Way too much glop in there.
Ah, no. You can't subtract the latest data and have anything useful revealed, as you don't know the exact waveshape (analog) or x or y location of the latest data bit. Disk heads have about 15% variable and unpredictable write current, another 20% unpredictable head height, plus there is disk speed jitter in the x direction and track positioning error and vibration in the y transverse direction. You cannot reproduce the writing conditions so the subtraction is going to be at least 25% off in amplitude and phase. The residual signal is a bit smaller than that. So you have a very lousy signal to noise ratio. You can't reconstruct data from a signal that is half noise.
A patent means.... nada. the math
on
The Google Navy
·
· Score: 3, Interesting
Something like 99.4% of patents never make a cent.
This one is particularly loopy.
Let's do the math. Let's say Google buys the Queen Mary. 80,000 tons. Let's say they anchor it someplace with an average wave height of 20 feet, wave period of 10 seconds. Raising 80,000 tons at 2 feet per second takes about 160,000 horsepower. Hmmm, that's very close to the original steaming capacity of the QM. In watts, that's about 120 megawatts, about ten times more than you'd need if you packed the ship with servers. Okay, so that looks easily doable.
Problem is, buying the electricity would be much cheaper. 12 megawatts will cost you about $700 an hour. Can you run and maintain and pay on the principal and pay salaries and insurance on $700/hour? No, not a couple of powers of ten.
It's an urban legend. You can't recover erased bits. If you could it would imply that you can store at least two bits in the space of one. Disk companies have a pretty good idea what their heads and surfaces can do. Do you think they'd be passing up big $$$ by under-utilizing their disk's capacity?
There is that one Usenix conference "paper" foating around out there, but if you read it carefully it does not give a single example of one recovered bit.
If you've ever looked at the waveform coming off a disk head, you'd wonder with all the x/y noise and jitter how they can get even ONE bit out of that hairball. The answer is, they can, just barely, by applying all the sync, gating, PLL, and deglitching tricks, just barely reliably recover bits at the maximum recording density possible.
And all those pictures they show of bit patterns lingering under large erased areas are actually counter-examples. They prove that you can detect periodic bit patterns under large erased areas. Duh. In the real world the underlying data is not periodic, and the erasure isn't smooth or periodic either. If you overwrite real typical data with random data, you can't recover the original data. Shannon and company, you know.
But if you think about it a bit, an orbital path can be described by a very few numbers-- the angles to the equator and to Greenwich, and the minor and major radii. All else can be computed on the fly by about 8 lines of code.
Old chinese proverb: "The nail that stands out gets hammered".
I was in a very, very similar situation. In a company with not a shred of software quality control. Everybody listened to my presentations suggesting we get someone with software engineering experience in the loop. Even a "thank you" from the CEO.
Six months later, I got very firmly terminated on wholly made-up charges of poor performance.
There's these things called the basic laws of physics which make this idea a dead end.
Magnetic fields go between two poles and not much further afield. That makes the far field go down as the cube of the distance. Basically insurmountable gotcha.
If you do a little digging, you find there is far less to this story than you might think.
All the lady did is develop a simple way of printing electrical contacts onto the silicon surface.
That's a mighty small part of the overall cell's cost. It's not going to bring cell prices down so the "2 billion" can afford them. heck, the top 2 billion can't afford them.
The devil is in the details. Ray tracing with glossy surfaces is relatively easy. But if you want to simulate real-world textures like orange-peel, bark, hair, or skin, things can really slow down.
Delaying or buffering the analog light signal is just a teensy part of the process. A typical packet needs to be detected, isolated, have its CRC checked, be inspected, have its addresses twiddled, have the CRC recalculated, and then queued for forwarding. It's gonna be really hard to do these things optically.
In addition most optical delay devices are going to have a strong phase shift over frequency characteristic, a very bad thing.
Methinks the materials folks should stick with what they know and not speculate on the uses.
Perhaps one should try looking at a map. Japan is small, habitable areas even smaller. That means wires can be short and cheap. Japan's people are well-trained to pay any amount for whatever biz and govt say they should buy. So you end up with lots of wideband tubes, perhaps not being used to anywhere near capacity.
The USA however, is a BIG place. Expensive to wire up Montana and Texas and the rest. And consumers here while still mildly hypnotized by advertising, occasionally want a choice in speeds and costs.
I did not say that things do not get cheaper. I said that even if it costs only 1/10th the amount to manufacture, that does not affect the retail price much if at all. For example the first round of IC-based computers cost much less to make, but were priced just a teense less than their predecessors.
There have been pretty good rules around for over 60 years regarding what the pilot should do when they can't contact the tower. Similarly the tower has an old red/green light gun for communicating with planes that can't hear.
It's unlikely there was any safety added by the cell phone sms messages. In fact, bypassing the usual no-radio procedures may have compromised safety. There may be some flags dropped on this play.
Nothing in the history of the Universe has ever sold for much less than the competition. It matters not one whit whether by some measure the cells cost 1/10th as much to manufacture. Every time there's been some breakthrough product that costs very little to make, there's a big flash and mobs of managers, QA people, ad agencies, consultants, beaurocrats, distributors, jobbers, salespersons, accountants, bookkeepers, supervisors, warehouses, trucks, and installers all materialize, and wadda you know, the product ends up costing jut a teensy bit less than the competition.
Tracing the x86 back to the 8008 is a mighty tenuous connection.
There are two very weak links.
First, the 8008 to 8080 transition was a major re-do. Like ten times the speed, an external stack, more. The opcodes were upwardly compatible to a point, but that's about the only similarity.
Next, the 8080 to 808x transition was just as abrupt. 16 bit registers, segments, and more. Again there was a certain backward compatibility, if you converted all the mnemonics and register names, but that was about all.
Only problem, a square meter of solar panel for eight hours is going to grab about 800 watt-hours at best. That's about one horsepower-hour. Or thirty HP for the highway for all of two minutes.
Well yes, a Zt of 3 would be nice, but it's not sufficent.
In general devices that take in random energy (heat) and output very structured energy (DC) have very low efficiencies, like single digits. It's hard to corral random molecular motion into coordinated electron motion. Your typical thermocouple puts out millivolts.
Bizness is fine-- there are a lot of good oscilloscopes and such from 1965 to 1980 vintage that nowdays need a few new electrolytics and that's about all. Anything made after that is too miniaturized and digitized to be repaired.
>because the electrolytic chemistry is inherently corrosive.
Yes and no. The whole point is to have a layer of aluminum oxide, so yes that's literally corrosion.
But aluminum oxide is such an inert coating, the corrosion stops after a few microns.
I have 1940 radios with the original electrolytics in them and they work just fine.
Now if you want to talk corrosion, there used to be "wet slug" tantalum capacitors that had sulfuric acid in them! When those leaked, they made a huge mess.
Tantalum is used in small quantities to make high-performance and compact electrolytic capacitors.
Typically a tantalum cap will have lower leakage current and be about 1/4 the size of a aluminum electrolytic, at about twice the cost.
As an electronics repair guy, I just *love* tantalum caps, as they quite often short out given an opportunity. Most repair places won't even try to do component-level repairs anymore, so that leaves lots of nice equipments for me to fix.
Slashdot Mirror
I would rather dive into a swimming pool full of Ebola, triple-edged razor blades, lemon juice, and pig doots than use any software from that "R" named firm.
Every time I've been stupid and downloaded their crap, thinking, "it used to be sooooo bad, any improvement would be a treeemendous improvement", I've had to delete it within the hour. Way too much glop in there.
Ah, no. You can't subtract the latest data and have anything useful revealed, as you don't know the exact waveshape (analog) or x or y location of the latest data bit. Disk heads have about 15% variable and unpredictable write current, another 20% unpredictable head height, plus there is disk speed jitter in the x direction and track positioning error and vibration in the y transverse direction. You cannot reproduce the writing conditions so the subtraction is going to be at least 25% off in amplitude and phase. The residual signal is a bit smaller than that. So you have a very lousy signal to noise ratio. You can't reconstruct data from a signal that is half noise.
Something like 99.4% of patents never make a cent.
This one is particularly loopy.
Let's do the math. Let's say Google buys the Queen Mary. 80,000 tons. Let's say they anchor it someplace with an average wave height of 20 feet, wave period of 10 seconds. Raising 80,000 tons at 2 feet per second takes about 160,000 horsepower. Hmmm, that's very close to the original steaming capacity of the QM. In watts, that's about 120 megawatts, about ten times more than you'd need if you packed the ship with servers. Okay, so that looks easily doable.
Problem is, buying the electricity would be much cheaper. 12 megawatts will cost you about $700 an hour. Can you run and maintain and pay on the principal and pay salaries and insurance on $700/hour? No, not a couple of powers of ten.
It's an urban legend. You can't recover erased bits. If you could it would imply that you can store at least two bits in the space of one. Disk companies have a pretty good idea what their heads and surfaces can do. Do you think they'd be passing up big $$$ by under-utilizing their disk's capacity?
There is that one Usenix conference "paper" foating around out there, but if you read it carefully it does not give a single example of one recovered bit.
If you've ever looked at the waveform coming off a disk head, you'd wonder with all the x/y noise and jitter how they can get even ONE bit out of that hairball. The answer is, they can, just barely, by applying all the sync, gating, PLL, and deglitching tricks, just barely reliably recover bits at the maximum recording density possible.
And all those pictures they show of bit patterns lingering under large erased areas are actually counter-examples. They prove that you can detect periodic bit patterns under large erased areas. Duh. In the real world the underlying data is not periodic, and the erasure isn't smooth or periodic either. If you overwrite real typical data with random data, you can't recover the original data. Shannon and company, you know.
Yep, if the laptops are made in China you can use workers that make $80 to $150 a month.
And as a free bonus, you might get one of the 1,000+ fingers a day that are chopped off in industrial accidents every day.
Way to save a buck, Dell!
I know, it's a wow kind of thing.
But if you think about it a bit, an orbital path can be described by a very few numbers-- the angles to the equator and to Greenwich, and the minor and major radii. All else can be computed on the fly by about 8 lines of code.
Logical conclusion, but.
The last three projects I did got done ahead of schedule, and passed all the tests.
And yes, I was supposed to be working on quality issues.
Apparently middle managers do not really like having their cluelessness pointed out, not even if it is gently and diplomatically presented. Who knew?
Old chinese proverb: "The nail that stands out gets hammered".
I was in a very, very similar situation. In a company with not a shred of software quality control. Everybody listened to my presentations suggesting we get someone with software engineering experience in the loop. Even a "thank you" from the CEO.
Six months later, I got very firmly terminated on wholly made-up charges of poor performance.
Draw your own conclusions.
There's these things called the basic laws of physics which make this idea a dead end.
Magnetic fields go between two poles and not much further afield. That makes the far field go down as the cube of the distance. Basically insurmountable gotcha.
If you do a little digging, you find there is far less to this story than you might think.
All the lady did is develop a simple way of printing electrical contacts onto the silicon surface.
That's a mighty small part of the overall cell's cost. It's not going to bring cell prices down so the "2 billion" can afford them. heck, the top 2 billion can't afford them.
Not quite totally dissimilar to a good comparison.
The allowed amounts of dioxin, TCE, and many other chemicals is down in the parts per billion. So the comparison is off by about five powers of ten.
The devil is in the details. Ray tracing with glossy surfaces is relatively easy. But if you want to simulate real-world textures like orange-peel, bark, hair, or skin, things can really slow down.
Delaying or buffering the analog light signal is just a teensy part of the process. A typical packet needs to be detected, isolated, have its CRC checked, be inspected, have its addresses twiddled, have the CRC recalculated, and then queued for forwarding. It's gonna be really hard to do these things optically.
In addition most optical delay devices are going to have a strong phase shift over frequency characteristic, a very bad thing.
Methinks the materials folks should stick with what they know and not speculate on the uses.
Perhaps one should try looking at a map. Japan is small, habitable areas even smaller. That means wires can be short and cheap. Japan's people are well-trained to pay any amount for whatever biz and govt say they should buy. So you end up with lots of wideband tubes, perhaps not being used to anywhere near capacity.
The USA however, is a BIG place. Expensive to wire up Montana and Texas and the rest. And consumers here while still mildly hypnotized by advertising, occasionally want a choice in speeds and costs.
You decide which regime you want to live under.
I did not say that things do not get cheaper. I said that even if it costs only 1/10th the amount to manufacture, that does not affect the retail price much if at all. For example the first round of IC-based computers cost much less to make, but were priced just a teense less than their predecessors.
There have been pretty good rules around for over 60 years regarding what the pilot should do when they can't contact the tower. Similarly the tower has an old red/green light gun for communicating with planes that can't hear.
It's unlikely there was any safety added by the cell phone sms messages. In fact, bypassing the usual no-radio procedures may have compromised safety. There may be some flags dropped on this play.
Nothing in the history of the Universe has ever sold for much less than the competition. It matters not one whit whether by some measure the cells cost 1/10th as much to manufacture. Every time there's been some breakthrough product that costs very little to make, there's a big flash and mobs of managers, QA people, ad agencies, consultants, beaurocrats, distributors, jobbers, salespersons, accountants, bookkeepers, supervisors, warehouses, trucks, and installers all materialize, and wadda you know, the product ends up costing jut a teensy bit less than the competition.
Tracing the x86 back to the 8008 is a mighty tenuous connection.
There are two very weak links.
First, the 8008 to 8080 transition was a major re-do. Like ten times the speed, an external stack, more. The opcodes were upwardly compatible to a point, but that's about the only similarity.
Next, the 8080 to 808x transition was just as abrupt. 16 bit registers, segments, and more. Again there was a certain backward compatibility, if you converted all the mnemonics and register names, but that was about all.
Only problem, a square meter of solar panel for eight hours is going to grab about 800 watt-hours at best. That's about one horsepower-hour. Or thirty HP for the highway for all of two minutes.
Oh yeah, cause web apps are so great.
I think about 800 million has already gone down that particular rat-hole.
Every try using Citrix or simlar for real work? Would you use it if your company did not force you to do so?
Expect another disaster of Vista-like proportions.
Well yes, a Zt of 3 would be nice, but it's not sufficent.
In general devices that take in random energy (heat) and output very structured energy (DC) have very low efficiencies, like single digits. It's hard to corral random molecular motion into coordinated electron motion. Your typical thermocouple puts out millivolts.
>at zT 3, these materials can replace freon refrigerators.
Uh, how? The Freon cycle can give EERs of 15.
An average Peltier device has an EER of under 0.4.
The numbers don't seem to work out.
Bizness is fine-- there are a lot of good oscilloscopes and such from 1965 to 1980 vintage that nowdays need a few new electrolytics and that's about all. Anything made after that is too miniaturized and digitized to be repaired.
>because the electrolytic chemistry is inherently corrosive.
Yes and no. The whole point is to have a layer of aluminum oxide, so yes that's literally corrosion.
But aluminum oxide is such an inert coating, the corrosion stops after a few microns.
I have 1940 radios with the original electrolytics in them and they work just fine.
Now if you want to talk corrosion, there used to be "wet slug" tantalum capacitors that had sulfuric acid in them! When those leaked, they made a huge mess.
Tantalum is used in small quantities to make high-performance and compact electrolytic capacitors.
Typically a tantalum cap will have lower leakage current and be about 1/4 the size of a aluminum electrolytic, at about twice the cost.
As an electronics repair guy, I just *love* tantalum caps, as they quite often short out given an opportunity. Most repair places won't even try to do component-level repairs anymore, so that leaves lots of nice equipments for me to fix.