Amen. There's an ideal length for a tax code. That length is under one page. Charge every corporation operating in the US 2% of gross receipts from the US market. Charge every non-incorporated business 2% of gross receipts regardless of source. Charge every household 15% above the median cost of housing, food, clothing, and transportation in their state. There, that gives a break to the poor and still takes less than a page. Quit taxing on profit/loss and making exceptions for everything under the Sun.
It could be a great tax earner. People are willing to pay $200 an ounce for really good stuff, and $30 an ounce for leaf tips from ditch weed. If it's grown widely enough that a good ounce costs $5 to grow, process, package, and ship, then tax it at $25 or even $45 an ounce. People'd still be getting it cheaper than they are now.
Any pitch can be both a strike and not a strike until the home plate umpire calls it, especially at a Cardinals/Cubs game. With 60,000 in attendance, you can have around 30,000 arguments about what the call should be, but only one call.
Running surface water in a channel (stream, creek, river) has certain limits on it in pretty much every US state.
First off, anything deemed "navigable" belongs to the US government even if your land surrounds it on three sides. Nobody owns part of the Mississippi, for example, and if the river washes away 400 square feet of banks from your property you've lost 400 square feet of property. OTOH, if 400 square feet of land washes up next to your land through accretion, that's yours.
In Illinois and some other states, a creek, pond, or lake that is surrounded on all sides by your property you can utilize how you want. That may be different in other states, and I wouldn't know (other than that Missouri appears to be the same) which to tell you are different.
If a creek flows onto your property or onto your neighbor's property from yours, there is joint ownership of the water rights in Illinois. This is true even if the creek starts from a spring completely within one owner's parcel of land. Ponds or lakes that cover land owned by more than one person are likewise joint resources. You can't just take up someone else's water from the land and use it how you want.
All of that's not legal advice, as I'm not a lawyer. It is what they teach in real estate agency classes in Illinois for the agents to be able to advise their clients about water rights.
The specifics of rain water capture vs. running surface water are another matter yet again. In drier areas or areas where the water resources are spread thin by irrigation and large urban populations placed nonsensically far away from adequate water supplies I can imagine the issues are much more contentious.
I see what you did there, and I can't understand why it's not modded funny yet. Man, I wish my mod points came today. It's one of those things that really shouldn't be funny, but definitely is.
You silly, this is where that "social engineering" comes into play.
You convince all these people they want to help you for free, then you sell the fruits of their labor for money.
Some see crowd sourcing as communist, but it's actually quite the opposite. It's capitalism in hyperdrive: you put up a little bit of capital, organize a whole bunch of ultra-cheap willing (so technically non-slave) labor, and you profit more from the higher margin. Those paid laborers can just keep working for those stupid enough to keep paying them.
I think you're analyzing the Streisand effect from the opposite direction of those tagging the story that way.
Apple didn't want a few hobbyists on OdioWorks talking about making the iPod work with software other than iTunes. Now, because they tried to stifle that publicity, there are these suits. Now Apple will have a bunch of people aware that there's a group wanting to make iPods interoperable with other software.
It's Apple getting more publicity because they didn't want it that earned the story the tag. You're right that the EFF wants to raise awareness of issues like this, though.
That's exactly my point. The high-speed trains will do nothing for most of America, because they won't stop in most of America.
Trains actually are much more efficient per ton than cars or semi trucks. Each ton that CSX or BNSF transports goes over 400 miles on a gallon of diesel. Each ton on a semi 40-ton semi getting a typical 8 miles per gallon is only going to move 320 miles per ton per gallon. Part of this is the long stretches of motion with fewer stops, and part of it is the lower rolling friction. Personal cars have more vehicle weight to cargo and passenger weight yet again.
Van pools actually do rate better than rail for typical passenger loads. In fact, they rate even better than motorcycles if we can trust the US DOE's statistics.
One problem with rail is getting to capacity of passengers. You could get more riders if it was easier to get around at your destination. Reliably clean and well-maintained rentals near the stations would go a long way toward that.
The trick is to load your vehicles to capacity. Use a lighter vehicle for fewer people and less stuff or put more people and stuff on vehicles that have economies of scale.
The die for the Phenom II is about 16mm * 16mm and has 758 million transistors (or about 379 million idealized logic gates, not accounting for flip-flops, latches, and such) at 45 nm per transistor and then a bunch of routing wires and such that are longer.
With over 8MB of cache (6 MB L3, 2 MB L2, and then the L1 as well), most of which will never be touched by an individual instruction, there's at least 64 million gates an instruction won't use right there.
After the cache, there's only 315 million gates left. There are over 300 instructions in the architecture. So figure 3.15 million gates per instruction (discounting the registers, pad IO, etc).
At 64 bits, you're looking at fewer than 50 thousand logic gates per bit per instruction. Again, this is discounting the transistors used in registers, used to stabilize the results of gates, in buffers, etc.
If you consider that a gate is two 45nm features, and consider the gates themselves only 10% of the distance the signal is required to travel, then 50,000 * 900 nm = 45,000,000 nm. That's 4.5 cm.
As others have said, thanks to pipelines an instruction can be dispatched while another is already in the process of completing. In fact, within the last decade at least some instructions on some CISC chips have taken as many as 20 or 25 clock cycles to complete. Two cycles per instruction on a pipelined system with very complex instructions (an adder is only 3 gate delays per bit, and we accounted for 50k gates per instruction) would be fabulous.
Of course, things don't really work that way. You do have to account for latches, flip-flops, the serial nature of inputs and outputs among the parts of the CPU, and the fact that silicon doesn't propagate electrons at the speed of light in a vacuum even near absolute zero. It certainly doesn't at room temperature.
The propagation of an EM wave through copper or aluminum at 1.7 or so volts is much lower than c. There's electrical resistance to worry about, which is one reason these ultra low-temperature benchmarks are able to crank so high. It's not just dissipating the heat from the resistance. It's actually lowering the resistance itself at those kinds of temperatures in the first place.
Within roughly the same family of chip using roughly the same logic the speed at which you run that logic is going to scale pretty reliably. What you can't do is take two different chip designs and compare performance by comparing clock speeds.
The Athlon isn't the same chip as the Athlon XP, which isn't the same as the Athlon X2. The Athlon x2 isn't the same as the Phenom, which isn't the same as the Phenom II. The Pentium III isn;t a Pentium 4 which isn't the same as the Pentium M. That Pentium M isn't the same as the Core, which isn't the same as the Core 2. The Core 2 isn't the same chip design as the Core i7.
Even bigger differences come in when you look at entirely different types of chips running other instruction sets. The x86, x86-64, Itanium, MIPS 32bit, MIPS 64 bit, SPARC, UltraSPARC, Arm 9, Arm 11, Alpha, Power, and Dragonball/ColdFire clock-for-clock are going to have very different performance characteristics. If you clocked each one at exactly 2 Ghz, you'd still have a wide variety of different types of instructions doing different amounts of work per instruction.
This simple discussion avoids RISC vs. CISC, pipeline length, microprogramming, cache sizes, cache types, instruction dispatch time vs. instruction completion time, fetch alignments, fetch stalls, instruction reordering, FPU interconnect speeds, memory bandwidth, number of instructions required to do perform common tasks (for example, one generation of the Alpha does some very common things in a single instruction what took the previous generation 10 to 12 instructions), and probably many other performance characteristics I'm failing to name at the moment.
That's true, but the i7-related Xeons and the Opterons are in fact NUMA and not SMP. The chips including their own memory controllers and talking to each other through HyperTransport and Intel's knock-off of it pretty much require that.
Vector processors like NVidia GPUs are really good at doing the same transformation on lots of similar data values all at once. For serial decision trees like most applications require, they don't really offer much benefit.
What would be really nice would be a blazing fast two core or so CPU mated to three or four GPUs, a couple of FPGAs, and a couple of DSPs. Then, though, developers would have to write code that gets compiled to four different instruction sets.
If a build system could automatically determine which parts of an app went to which processors and handle optimizing for the communications among the different chips, that's where you'd see really huge performance wins. Right now the hardest part of application design is getting multiple identical cores to coordinate their work properly. This is often due to the serial nature of the tasks.
One place you can really notice a difference is in your RAM. A CAS latency of 2 on a system that defaulted to 4 or 5 in the BIOS will make the system run noticeably faster.
Another place you'll find exceptional, noticeable speed increases is when removing dependencies on rotating drives. Switch to an SSD for files you read often. Add a bunch more RAM and set your temporary directory up on a RAM drive. Use a caching drive controller (preferably one with battery-backed writes) if you have the cash and expansion slot for it.
A 5% increase in processor speeds can be noticeable in very few circumstances. A 5% bus speed increase is more noticeable because all your components are communicating at that increased 5%, like your RAM. The CPU usually isn't the biggest bottleneck, but speeding up the bus, memory controller, or both mitigates a bottleneck that isn't just in your CPU.
What part of "once the crackers get to exploiting it against Windows boxen" is so difficult to understand?
The exploit is a motherboard logic chip issue for one core logic chip set. That's why it's not going to be a market-crashing problem for Intel.
This is not a Linux v. Microsoft issue at all. It's an Intel motherboard issue. I think you're both wrong to some extent, but you're far more wrong than Ungrounded Lightning. Linux just makes it easier for someone who is already the administrative user on the box to access the box. It doesn't make it easier for someone to become the administrative user who wasn't already.
Once Windows exploit code is available, the news will be worse for Intel. That's because it will affect many more Intel customers who have the board. Duh.
It still won't be catastrophic. The Pentium division bug didn't kill them, the Pentium IV didn't kill them, and this certainly isn't a bigger problem than those.
Gold doesn't get more valuable. The price of gold goes up when the value of the currency tanks. The price of a commodity and its and value do not equate across a change in value of the currency used to measure them.
Oh, sorry to reply yet again. Yes, I can name a number of project forks in which the original fell by the wayside as a fork took off. Not all the originals are dead, mind you, but the forks are much more popular. A few of these are situations in which the original is still viable (Debian, for example), but in which the fork has a huge number advantage or a lot of momentum.
The company that writes some application that talks to a GPLed MySQL will have to follow the GPL in distributing MySQL. They will not have to license their own code as GPL. What's the problem with giving someone the source to something they could have gone to the website and downloaded themselves anyway?
So replace the motherboard as well, since that's the vulnerable and compromised component. You might even ask for a refund or replacement, as this is a design defect.
Considering this requires you to already be root ("Administrator" in Windows talk), I'm not sure how low-hanging I'd call it. If someone's your system's administrator, they can probably do all sorts of easier and nastier things.
iLife '09 already tries (and does a decent job, if the demos are to be believed) of categorizing your photos by setting and subject. It uses face recognition and any embedded GPS data in the image file from your camera to do so.
BTW, I'm not an Apple fanboy, and I'm pissed that's what was covered in their presentation Sunday that was supposed to be about how environmentally friendly their systems and manufacturing processes are.
Slashdot Mirror
Amen. There's an ideal length for a tax code. That length is under one page. Charge every corporation operating in the US 2% of gross receipts from the US market. Charge every non-incorporated business 2% of gross receipts regardless of source. Charge every household 15% above the median cost of housing, food, clothing, and transportation in their state. There, that gives a break to the poor and still takes less than a page. Quit taxing on profit/loss and making exceptions for everything under the Sun.
It could be a great tax earner. People are willing to pay $200 an ounce for really good stuff, and $30 an ounce for leaf tips from ditch weed. If it's grown widely enough that a good ounce costs $5 to grow, process, package, and ship, then tax it at $25 or even $45 an ounce. People'd still be getting it cheaper than they are now.
Any pitch can be both a strike and not a strike until the home plate umpire calls it, especially at a Cardinals/Cubs game. With 60,000 in attendance, you can have around 30,000 arguments about what the call should be, but only one call.
Running surface water in a channel (stream, creek, river) has certain limits on it in pretty much every US state.
First off, anything deemed "navigable" belongs to the US government even if your land surrounds it on three sides. Nobody owns part of the Mississippi, for example, and if the river washes away 400 square feet of banks from your property you've lost 400 square feet of property. OTOH, if 400 square feet of land washes up next to your land through accretion, that's yours.
In Illinois and some other states, a creek, pond, or lake that is surrounded on all sides by your property you can utilize how you want. That may be different in other states, and I wouldn't know (other than that Missouri appears to be the same) which to tell you are different.
If a creek flows onto your property or onto your neighbor's property from yours, there is joint ownership of the water rights in Illinois. This is true even if the creek starts from a spring completely within one owner's parcel of land. Ponds or lakes that cover land owned by more than one person are likewise joint resources. You can't just take up someone else's water from the land and use it how you want.
All of that's not legal advice, as I'm not a lawyer. It is what they teach in real estate agency classes in Illinois for the agents to be able to advise their clients about water rights.
The specifics of rain water capture vs. running surface water are another matter yet again. In drier areas or areas where the water resources are spread thin by irrigation and large urban populations placed nonsensically far away from adequate water supplies I can imagine the issues are much more contentious.
I see what you did there, and I can't understand why it's not modded funny yet. Man, I wish my mod points came today. It's one of those things that really shouldn't be funny, but definitely is.
You silly, this is where that "social engineering" comes into play.
You convince all these people they want to help you for free, then you sell the fruits of their labor for money.
Some see crowd sourcing as communist, but it's actually quite the opposite. It's capitalism in hyperdrive: you put up a little bit of capital, organize a whole bunch of ultra-cheap willing (so technically non-slave) labor, and you profit more from the higher margin. Those paid laborers can just keep working for those stupid enough to keep paying them.
I think you're analyzing the Streisand effect from the opposite direction of those tagging the story that way.
Apple didn't want a few hobbyists on OdioWorks talking about making the iPod work with software other than iTunes. Now, because they tried to stifle that publicity, there are these suits. Now Apple will have a bunch of people aware that there's a group wanting to make iPods interoperable with other software.
It's Apple getting more publicity because they didn't want it that earned the story the tag. You're right that the EFF wants to raise awareness of issues like this, though.
That's exactly my point. The high-speed trains will do nothing for most of America, because they won't stop in most of America.
Trains actually are much more efficient per ton than cars or semi trucks. Each ton that CSX or BNSF transports goes over 400 miles on a gallon of diesel. Each ton on a semi 40-ton semi getting a typical 8 miles per gallon is only going to move 320 miles per ton per gallon. Part of this is the long stretches of motion with fewer stops, and part of it is the lower rolling friction. Personal cars have more vehicle weight to cargo and passenger weight yet again.
Van pools actually do rate better than rail for typical passenger loads. In fact, they rate even better than motorcycles if we can trust the US DOE's statistics.
One problem with rail is getting to capacity of passengers. You could get more riders if it was easier to get around at your destination. Reliably clean and well-maintained rentals near the stations would go a long way toward that.
The trick is to load your vehicles to capacity. Use a lighter vehicle for fewer people and less stuff or put more people and stuff on vehicles that have economies of scale.
The die for the Phenom II is about 16mm * 16mm and has 758 million transistors (or about 379 million idealized logic gates, not accounting for flip-flops, latches, and such) at 45 nm per transistor and then a bunch of routing wires and such that are longer.
With over 8MB of cache (6 MB L3, 2 MB L2, and then the L1 as well), most of which will never be touched by an individual instruction, there's at least 64 million gates an instruction won't use right there.
After the cache, there's only 315 million gates left. There are over 300 instructions in the architecture. So figure 3.15 million gates per instruction (discounting the registers, pad IO, etc).
At 64 bits, you're looking at fewer than 50 thousand logic gates per bit per instruction. Again, this is discounting the transistors used in registers, used to stabilize the results of gates, in buffers, etc.
If you consider that a gate is two 45nm features, and consider the gates themselves only 10% of the distance the signal is required to travel, then 50,000 * 900 nm = 45,000,000 nm. That's 4.5 cm.
As others have said, thanks to pipelines an instruction can be dispatched while another is already in the process of completing. In fact, within the last decade at least some instructions on some CISC chips have taken as many as 20 or 25 clock cycles to complete. Two cycles per instruction on a pipelined system with very complex instructions (an adder is only 3 gate delays per bit, and we accounted for 50k gates per instruction) would be fabulous.
Of course, things don't really work that way. You do have to account for latches, flip-flops, the serial nature of inputs and outputs among the parts of the CPU, and the fact that silicon doesn't propagate electrons at the speed of light in a vacuum even near absolute zero. It certainly doesn't at room temperature.
The propagation of an EM wave through copper or aluminum at 1.7 or so volts is much lower than c. There's electrical resistance to worry about, which is one reason these ultra low-temperature benchmarks are able to crank so high. It's not just dissipating the heat from the resistance. It's actually lowering the resistance itself at those kinds of temperatures in the first place.
Within roughly the same family of chip using roughly the same logic the speed at which you run that logic is going to scale pretty reliably. What you can't do is take two different chip designs and compare performance by comparing clock speeds.
The Athlon isn't the same chip as the Athlon XP, which isn't the same as the Athlon X2. The Athlon x2 isn't the same as the Phenom, which isn't the same as the Phenom II. The Pentium III isn;t a Pentium 4 which isn't the same as the Pentium M. That Pentium M isn't the same as the Core, which isn't the same as the Core 2. The Core 2 isn't the same chip design as the Core i7.
Even bigger differences come in when you look at entirely different types of chips running other instruction sets. The x86, x86-64, Itanium, MIPS 32bit, MIPS 64 bit, SPARC, UltraSPARC, Arm 9, Arm 11, Alpha, Power, and Dragonball/ColdFire clock-for-clock are going to have very different performance characteristics. If you clocked each one at exactly 2 Ghz, you'd still have a wide variety of different types of instructions doing different amounts of work per instruction.
This simple discussion avoids RISC vs. CISC, pipeline length, microprogramming, cache sizes, cache types, instruction dispatch time vs. instruction completion time, fetch alignments, fetch stalls, instruction reordering, FPU interconnect speeds, memory bandwidth, number of instructions required to do perform common tasks (for example, one generation of the Alpha does some very common things in a single instruction what took the previous generation 10 to 12 instructions), and probably many other performance characteristics I'm failing to name at the moment.
That's true, but the i7-related Xeons and the Opterons are in fact NUMA and not SMP. The chips including their own memory controllers and talking to each other through HyperTransport and Intel's knock-off of it pretty much require that.
Vector processors like NVidia GPUs are really good at doing the same transformation on lots of similar data values all at once. For serial decision trees like most applications require, they don't really offer much benefit.
What would be really nice would be a blazing fast two core or so CPU mated to three or four GPUs, a couple of FPGAs, and a couple of DSPs. Then, though, developers would have to write code that gets compiled to four different instruction sets.
If a build system could automatically determine which parts of an app went to which processors and handle optimizing for the communications among the different chips, that's where you'd see really huge performance wins. Right now the hardest part of application design is getting multiple identical cores to coordinate their work properly. This is often due to the serial nature of the tasks.
One place you can really notice a difference is in your RAM. A CAS latency of 2 on a system that defaulted to 4 or 5 in the BIOS will make the system run noticeably faster.
Another place you'll find exceptional, noticeable speed increases is when removing dependencies on rotating drives. Switch to an SSD for files you read often. Add a bunch more RAM and set your temporary directory up on a RAM drive. Use a caching drive controller (preferably one with battery-backed writes) if you have the cash and expansion slot for it.
A 5% increase in processor speeds can be noticeable in very few circumstances. A 5% bus speed increase is more noticeable because all your components are communicating at that increased 5%, like your RAM. The CPU usually isn't the biggest bottleneck, but speeding up the bus, memory controller, or both mitigates a bottleneck that isn't just in your CPU.
What part of "once the crackers get to exploiting it against Windows boxen" is so difficult to understand?
The exploit is a motherboard logic chip issue for one core logic chip set. That's why it's not going to be a market-crashing problem for Intel.
This is not a Linux v. Microsoft issue at all. It's an Intel motherboard issue. I think you're both wrong to some extent, but you're far more wrong than Ungrounded Lightning. Linux just makes it easier for someone who is already the administrative user on the box to access the box. It doesn't make it easier for someone to become the administrative user who wasn't already.
Once Windows exploit code is available, the news will be worse for Intel. That's because it will affect many more Intel customers who have the board. Duh.
It still won't be catastrophic. The Pentium division bug didn't kill them, the Pentium IV didn't kill them, and this certainly isn't a bigger problem than those.
Gold doesn't get more valuable. The price of gold goes up when the value of the currency tanks. The price of a commodity and its and value do not equate across a change in value of the currency used to measure them.
Oh, sorry to reply yet again. Yes, I can name a number of project forks in which the original fell by the wayside as a fork took off. Not all the originals are dead, mind you, but the forks are much more popular. A few of these are situations in which the original is still viable (Debian, for example), but in which the fork has a huge number advantage or a lot of momentum.
GCC -> EGCS -> GCC
Mosaic -> Netscape
Netscape -> Mozilla
Mozilla (Seamonkey) -> Mozilla Firefox
KHTML -> WebKit
Debian -> Ubuntu
XFree -> X.org
StarOffice -> OpenOffice
SSH -> OpenSSH
Hack -> NetHack
osCommerce -> ZenCart
AT&T Unix -> BSD
Crap, they didn't buy SAP. They bought PeopleSoft. Sorry.
Actually, there's nothing about InnoDB or SAP that are third-party to Oracle, since it already acquired both of those.
The company that writes some application that talks to a GPLed MySQL will have to follow the GPL in distributing MySQL. They will not have to license their own code as GPL. What's the problem with giving someone the source to something they could have gone to the website and downloaded themselves anyway?
Repeat this sentence three times: "Bundling is not binding." There, feel better?
The GNU General Public License v2 - An Overview (subsection on bundling and aggregation).
You can also read the GPL, v2 and the GPL, v3 for more information rather than spreading disinformation.
If you run a virtual server farm, please don't run it on desktop motherboards with cache overflow vulnerabilities. ;-)
So replace the motherboard as well, since that's the vulnerable and compromised component. You might even ask for a refund or replacement, as this is a design defect.
I think you missed the pun. To "queue" a group means to have them form a line so they can each have their turn at something.
Considering this requires you to already be root ("Administrator" in Windows talk), I'm not sure how low-hanging I'd call it. If someone's your system's administrator, they can probably do all sorts of easier and nastier things.
iLife '09 already tries (and does a decent job, if the demos are to be believed) of categorizing your photos by setting and subject. It uses face recognition and any embedded GPS data in the image file from your camera to do so.
BTW, I'm not an Apple fanboy, and I'm pissed that's what was covered in their presentation Sunday that was supposed to be about how environmentally friendly their systems and manufacturing processes are.