They don't just have marketing. They also have products that suck marginally less than their competitors. They also make different bad UI decisions, so once you're used to one usability disaster, moving to the competitor's usability disaster is even harder.
Getting people to take a look is the easy bit. Getting people to post interesting comments is the hard bit. Bruce Perens ran a Slashcode install for a while that typically posted more interesting stories than Slashdot. I don't think I ever saw more than 50 posts in a story, and most were in the 0-25 range. And with only 25 comments in a popular story, there's not enough to encourage you to reply, so even people who knew about it ended up spending more time on Slashdot.
Yes, it says that I'm not American. I learned this usage as a very small child in England, but it may have only been present in English, not in the unofficial fork spoken on the other side of the pond.
I just checked two US English dictionaries and they both listed that use. I don't think we can blame the original poster for your poor command of the language...
Every competitor in this field charges $0 for the OS
Well, kinda. Most Android manufacturers are already paying MS $15/device for a promise not to sue. Paying $15/device to actually get an OS might be worth it. And, sure, you can get the Android software after release for free, but to have access to the under-development versions you need to pay Google. You also need to pay Google if you want to ship their apps. Plus you need to pay your developers to get Android ported to your device, and to keep drivers up to date as kernel interfaces change if you want to allow users to upgrade. Customisation also costs money if you want to differentiate your product at all.
Even then, it's difficult. Precious metals are difficult to sell. For gold, in the quantities found in jewellery, the cost of verifying the quality is often close to the value of the metal.
That amount of energy is huge, but it's an amount that it is conceivable to produce. We can produce antimatter in small quantities, and so it's not implausible to assume that within 100 years we'd have the ability to use a fusion power plant to generate enough that we could power something like this, if we could build it. In contrast, earlier models showed that it would take more than the energy output of a star. This is the difference between 'we don't have any way of building this with current technology' and 'we probably can't ever build this'.
When IPV6 is what we have to work with, we will be swarmed by those bastard botnets with no way to block that many IP addresses that will be used to attack.
Don't block the address, block the prefix. Block a/64 and you're probably blocking a consumer endpoint. With IPv6, addresses are allocated hierarchically, so this becomes even easier. Just shorten the prefix and you'll eventually get the whole ISP. This makes it very easy to block ISPs or even countries that harbour spammers.
Additionally, it becomes much easier for a home user to identify attacks at the router. If you pick a random 32-bit number, odds are that it is a valid IPv4 address. Pick a dozen and you've almost certainly found one that's a home Internet connection. That makes it very easy for malware to spread. Pick a random 64-bit number, and if you're very lucky it's an IPv6 subnet that has some computers on it. Now you have to pick another 64-bit number to find one of the computers on it. For a home Internet connection, most users will be using under 50 of these (and rotating them quite frequently), so you end up with a 50 in 2^64 chance of getting the right one. After a few tries, their router's firewall will notice the suspicious behaviour (lots of connection requests to nonexistent addresses) and block your/64.
For home users, it entails pretty much nothing. If you're running a commodity operating system, it probably already advertises its host name via mDNS. It may also already advertise its link-local IPv6 address. Try sshing to a Mac on your local network by its name and see which address it tries to connect to: you may be surprised...
There are people in the mix. There are huge teams of people analysing the data sent back by the probe. You can't send them all to Mars, so you're still looking at a significant communication latency between the people doing the experiments and the people analysing the results.
Tell me more... While interesting factoids, do you think you might be off by an order of magnitude there, or maybe two? I'm just objecting to arbitrary numbers being thrown around
A shuttle launch costs $500M. That's the public number published by NASA. The total cost of a PhD is around $200K. My PhD was on a grant for a total of £500K, including funding three PhD students, two postdocs, and part of six lecturers for three years. The overheads for a postdoc are about the same as for a PhD student, so if we assume that none of the money went to lecturers' salaries and costs (which is not true) and that PhD students and postdocs cost the same amount (postdocs get a higher salary, but other costs are similar) that gives an upper bound on the cost of a PhD at £100,000. That works out at about $160K, so that's 3125 PhD students per shuttle launch. So, I was being a bit pessimistic in my number, but within an order of magnitude. Costs in the USA are probably a bit higher, as PhDs tend to take longer, so 1500 is probably a reasonable ballpark.
I'm currently on a DARPA-funded project, and my number for that is based on the size of our grant, which is funding two teams, on here in Cambridge and one in SRI, working together on the same project.
Now there's some thinking that will really piss off people in a few billion years should you continue down the path of isolation.
If manned space travel isn't feasible in a few hundred years, then we're doing something badly wrong. The problem is that it, like many other things, has a number of technical prerequisites. A lot of them are in materials science, where research is very expensive and the space program doesn't have enough funding (even if it spent all of its money) to meaningfully influence the speed of development. Apollo needed high-termperature ceramics and it needed computers. Regular, cheap, travel from the Earth to orbit requires a space elevator which requires (among other things) something with the tensile strength on the order of carbon nanotubes (but which can be mass produced) and either 80+% efficient photovoltaic cells or cheap superconductors. These are both likely to appear independent of a space program well within the next 50 years. A space elevator will probably take 10 years to construct and be phenomenally expensive (it will make the Channel Tunnel seem cheap) but has a potentially huge return on investment.
Look at sea and air travel. Current ships and planes are vastly more efficient and safe than early endeavours and a lot of the technology that made this possible was originally created for other uses. To put the cost of manned space travel into perspective, a single shuttle launch cost enough to completely fund about 1,500 PhDs to completion, or to fund about 200 DARPA advanced research projects. And that's just to get the ship into orbit, not counting the costs of the equipment for the mission, the training, the ground personnel, and so on.
A man on the surface of mars could do more in a single day than all the probes have done to date.
At a vastly higher cost. The current rovers mass far less than a man, but on a trip to Mars, the cost of radiation shielding, water recycling and food would dwarf the mass of the man. Plus, of course, all of the propellant required to move all of this into orbit and then to Mars. And the larger landing craft required. The cost of sending a man on a one-way trip to Mars with a year of supplies would be well over a thousand times the cost of sending a rover.
The trace cache, as I said in my original post, is mainly used for loops. It's very small, so it only kicks in if you're running the same sequence of instructions repeatedly. Make it any bigger, and it starts to cut into die area that could be used for instruction cache, so it stops being a win. The trace cache contains the result of translating x86 ops to micro ops. The micro ops are about as complex in terms of encoding as a typical RISC encoding like ARM and so the power cost of decoding them is similar to that of just decoding ARM instructions, except that you need to store them in SRAM twice: once in their encoded and once in their decoded state.
Look at the picture, it's only in ARMv5/6. In ARMv7 it's kind-of there, but the enter Jazelle mode instruction returns failure (check the docs). There's no ARMv8 silicon, but given the lack of market, I'd very much doubt it.
Every modern CISC chip is basically a dynamic translator on top of a RISC core.
And that's the problem for power consumption. You can cut power to execution units that are not being used. You can't ever turn off the decoder ever (except in Xeons, where you do in loops, but you leave on the micro-op decoder, which uses as much power as an ARM decoder) because every instruction needs decoding.
But even high-end ARM chips can do some of this with Jazelle.
Jazelle has been gone for years. None of the Cortex series include it. It gave worse performance to a modern JIT, but in a lower memory footprint. It's only useful when you want to run Java apps in 4MB of RAM.
The code density is better on x86 (yes, even with Thumb), which does mean they tend to use instruction cache more effecitvely
That's not what my tests show, in either compiled core or hand-written assembly.
If that was the logic, the advertising-supported web would have failed as a business model over a decade ago.
It almost did. Then a company came along with plain text adverts that were based on the content of the page and so relevant to your interests (if you were interested in the page). They were not intrusive, but sometimes they were actually relevant. The company made a huge pile of money based on this idea.
They're also fighting Moore's Law. The CPUs and GPUs that are in smartphones today are going to be in featurephones in 18 months. You don't make a difference by aiming for the low end, because the low end keeps moving. There's a lower bound for every generation on the cost of an IC, and putting something that's currently a low-end CPU in a featurephone just won't make sense in a year's time, when the cost difference between it and a significantly faster one is a few cents (and possibly in the wrong direction, because the faster ones are being produced in greater quantities).
Slashdot Mirror
Those who would exchange delicious bacon for a little temporary safety deserve neither bacon nor safety.
They don't just have marketing. They also have products that suck marginally less than their competitors. They also make different bad UI decisions, so once you're used to one usability disaster, moving to the competitor's usability disaster is even harder.
Getting people to take a look is the easy bit. Getting people to post interesting comments is the hard bit. Bruce Perens ran a Slashcode install for a while that typically posted more interesting stories than Slashdot. I don't think I ever saw more than 50 posts in a story, and most were in the 0-25 range. And with only 25 comments in a popular story, there's not enough to encourage you to reply, so even people who knew about it ended up spending more time on Slashdot.
What an excellent post! Have you considered a career in data mining? We have a list of related jobs if you just click here...
I've not seen that before. Is Dan Riccio a cyborg?
I knew there was a reason they wrote Slashcode in obfuscated Perl (but I repeat myself).
Yes, it says that I'm not American. I learned this usage as a very small child in England, but it may have only been present in English, not in the unofficial fork spoken on the other side of the pond.
I just checked two US English dictionaries and they both listed that use. I don't think we can blame the original poster for your poor command of the language...
Every competitor in this field charges $0 for the OS
Well, kinda. Most Android manufacturers are already paying MS $15/device for a promise not to sue. Paying $15/device to actually get an OS might be worth it. And, sure, you can get the Android software after release for free, but to have access to the under-development versions you need to pay Google. You also need to pay Google if you want to ship their apps. Plus you need to pay your developers to get Android ported to your device, and to keep drivers up to date as kernel interfaces change if you want to allow users to upgrade. Customisation also costs money if you want to differentiate your product at all.
Even then, it's difficult. Precious metals are difficult to sell. For gold, in the quantities found in jewellery, the cost of verifying the quality is often close to the value of the metal.
He didn't, his wife did after his death based on some very rough notes about a future series in the Star Trek universe.
That amount of energy is huge, but it's an amount that it is conceivable to produce. We can produce antimatter in small quantities, and so it's not implausible to assume that within 100 years we'd have the ability to use a fusion power plant to generate enough that we could power something like this, if we could build it. In contrast, earlier models showed that it would take more than the energy output of a star. This is the difference between 'we don't have any way of building this with current technology' and 'we probably can't ever build this'.
I may have missed something, when did the UK move to North America?
When IPV6 is what we have to work with, we will be swarmed by those bastard botnets with no way to block that many IP addresses that will be used to attack.
Don't block the address, block the prefix. Block a /64 and you're probably blocking a consumer endpoint. With IPv6, addresses are allocated hierarchically, so this becomes even easier. Just shorten the prefix and you'll eventually get the whole ISP. This makes it very easy to block ISPs or even countries that harbour spammers.
Additionally, it becomes much easier for a home user to identify attacks at the router. If you pick a random 32-bit number, odds are that it is a valid IPv4 address. Pick a dozen and you've almost certainly found one that's a home Internet connection. That makes it very easy for malware to spread. Pick a random 64-bit number, and if you're very lucky it's an IPv6 subnet that has some computers on it. Now you have to pick another 64-bit number to find one of the computers on it. For a home Internet connection, most users will be using under 50 of these (and rotating them quite frequently), so you end up with a 50 in 2^64 chance of getting the right one. After a few tries, their router's firewall will notice the suspicious behaviour (lots of connection requests to nonexistent addresses) and block your /64.
For home users, it entails pretty much nothing. If you're running a commodity operating system, it probably already advertises its host name via mDNS. It may also already advertise its link-local IPv6 address. Try sshing to a Mac on your local network by its name and see which address it tries to connect to: you may be surprised...
There are people in the mix. There are huge teams of people analysing the data sent back by the probe. You can't send them all to Mars, so you're still looking at a significant communication latency between the people doing the experiments and the people analysing the results.
Tell me more... While interesting factoids, do you think you might be off by an order of magnitude there, or maybe two? I'm just objecting to arbitrary numbers being thrown around
A shuttle launch costs $500M. That's the public number published by NASA. The total cost of a PhD is around $200K. My PhD was on a grant for a total of £500K, including funding three PhD students, two postdocs, and part of six lecturers for three years. The overheads for a postdoc are about the same as for a PhD student, so if we assume that none of the money went to lecturers' salaries and costs (which is not true) and that PhD students and postdocs cost the same amount (postdocs get a higher salary, but other costs are similar) that gives an upper bound on the cost of a PhD at £100,000. That works out at about $160K, so that's 3125 PhD students per shuttle launch. So, I was being a bit pessimistic in my number, but within an order of magnitude. Costs in the USA are probably a bit higher, as PhDs tend to take longer, so 1500 is probably a reasonable ballpark.
I'm currently on a DARPA-funded project, and my number for that is based on the size of our grant, which is funding two teams, on here in Cambridge and one in SRI, working together on the same project.
Now there's some thinking that will really piss off people in a few billion years should you continue down the path of isolation.
If manned space travel isn't feasible in a few hundred years, then we're doing something badly wrong. The problem is that it, like many other things, has a number of technical prerequisites. A lot of them are in materials science, where research is very expensive and the space program doesn't have enough funding (even if it spent all of its money) to meaningfully influence the speed of development. Apollo needed high-termperature ceramics and it needed computers. Regular, cheap, travel from the Earth to orbit requires a space elevator which requires (among other things) something with the tensile strength on the order of carbon nanotubes (but which can be mass produced) and either 80+% efficient photovoltaic cells or cheap superconductors. These are both likely to appear independent of a space program well within the next 50 years. A space elevator will probably take 10 years to construct and be phenomenally expensive (it will make the Channel Tunnel seem cheap) but has a potentially huge return on investment.
Look at sea and air travel. Current ships and planes are vastly more efficient and safe than early endeavours and a lot of the technology that made this possible was originally created for other uses. To put the cost of manned space travel into perspective, a single shuttle launch cost enough to completely fund about 1,500 PhDs to completion, or to fund about 200 DARPA advanced research projects. And that's just to get the ship into orbit, not counting the costs of the equipment for the mission, the training, the ground personnel, and so on.
A man on the surface of mars could do more in a single day than all the probes have done to date.
At a vastly higher cost. The current rovers mass far less than a man, but on a trip to Mars, the cost of radiation shielding, water recycling and food would dwarf the mass of the man. Plus, of course, all of the propellant required to move all of this into orbit and then to Mars. And the larger landing craft required. The cost of sending a man on a one-way trip to Mars with a year of supplies would be well over a thousand times the cost of sending a rover.
Technically, so would finding one that counts as work, but it's probably less fun.
The trace cache, as I said in my original post, is mainly used for loops. It's very small, so it only kicks in if you're running the same sequence of instructions repeatedly. Make it any bigger, and it starts to cut into die area that could be used for instruction cache, so it stops being a win. The trace cache contains the result of translating x86 ops to micro ops. The micro ops are about as complex in terms of encoding as a typical RISC encoding like ARM and so the power cost of decoding them is similar to that of just decoding ARM instructions, except that you need to store them in SRAM twice: once in their encoded and once in their decoded state.
Look at the picture, it's only in ARMv5/6. In ARMv7 it's kind-of there, but the enter Jazelle mode instruction returns failure (check the docs). There's no ARMv8 silicon, but given the lack of market, I'd very much doubt it.
Every modern CISC chip is basically a dynamic translator on top of a RISC core.
And that's the problem for power consumption. You can cut power to execution units that are not being used. You can't ever turn off the decoder ever (except in Xeons, where you do in loops, but you leave on the micro-op decoder, which uses as much power as an ARM decoder) because every instruction needs decoding.
But even high-end ARM chips can do some of this with Jazelle.
Jazelle has been gone for years. None of the Cortex series include it. It gave worse performance to a modern JIT, but in a lower memory footprint. It's only useful when you want to run Java apps in 4MB of RAM.
The code density is better on x86 (yes, even with Thumb), which does mean they tend to use instruction cache more effecitvely
That's not what my tests show, in either compiled core or hand-written assembly.
Alarm clocks that you silenced by hitting them were around in the '80s, probably before. Using an existing UI on a new device is not novel.
If that was the logic, the advertising-supported web would have failed as a business model over a decade ago.
It almost did. Then a company came along with plain text adverts that were based on the content of the page and so relevant to your interests (if you were interested in the page). They were not intrusive, but sometimes they were actually relevant. The company made a huge pile of money based on this idea.
They're also fighting Moore's Law. The CPUs and GPUs that are in smartphones today are going to be in featurephones in 18 months. You don't make a difference by aiming for the low end, because the low end keeps moving. There's a lower bound for every generation on the cost of an IC, and putting something that's currently a low-end CPU in a featurephone just won't make sense in a year's time, when the cost difference between it and a significantly faster one is a few cents (and possibly in the wrong direction, because the faster ones are being produced in greater quantities).