"It depends. The likely result would depend on why you didn't make it. Can verify you had to go to the hospital? Most businesses would give you a refund or credit. Something less drastic? Most businesses would give you a credit."
I really disagree with this idea. Why should someone else be able to arbiter whether my reason for not attending the concert is a "valid excuse"? In a similar way I never tell my boss why I want time off work: if I have leave available then they can either approve or deny my request. I refuse to allow others the power to adjudicate my own life.
So? I think this is precisely the sort of thing that should never be "illegal" and is absolutely not a criminal matter.
If the ticket seller has a "terms of service" that re-selling tickets is against their policy, then they can take appropriate civil action against those that break their policy.
I thought the first sale doctrine was pretty strong in the US and most items could be resold without regard.
What is different about ticket sales? Can someone explain it to me?
These countries still have a healthy pharmaceutical industry and obviously turn a profit for private companies. Remember, that sick people do need medicine and the pharmaceutical industry meets that need nicely and does a good job of it.
Chickens are modern dinosaurs and will gladly eat meat.
We have a few sturdy wooden boxes into which we regularly throw animal carcasses. The bodies are mostly scavenged from the road-side, others from shooting.
After a half week the body is crawling with maggots. We tip the whole lot out into the chicken coop. The little bastards go crazy for the wrigglers, but what may be surprising to many is that after a few days the bones will be stripped clean: the chickens happily eat the lot.
Very good for the chickens, and we feel good about keeping the neighborhood roads clear of road kill.
I too am an experimentalist and I was thinking about how to do it cheaply...
I think the simplest approach is to run the experiment twice, once with the device "on", and once with the device "off", and look for differences in the thrust, rather than trying to take a direct measurement of the thrust since as you correctly point out, there is heaps to take into account.
But I put scare quotes around "on" and "off" since how to do this will have an obvious impact on the device and the coupling to the device. Running power through cables will change their temperature, and thermal expansion may change the way they tug on the device. Same goes for the microwave guide tubes. My only idea worth sharing at this time is to set everything up as standard and then run it, and then replace one end of the cavity with a material transparent to microwaves and run it again. A high density plastic should do the trick, so replace 1mm of copper (or whatever the cavity is made from, brass? Aluminum? I don't think it matters...) with 10mm of plastic and the mass should be about the same. The only difference I would expect is that the cavity may run cooler in the second run. It should definitely be monitored for temperature, and if different by more than a degree (C) then some active cooling should be applied, which would then be the same for each run. Of course blowing a fan on it will probably produce more thrust than the device is generating:)
The thrust measurement can be directly measured against a very light spring scale provided the linear bearing has sufficiently low stiction. I'd recommend porous graphite air bearings to take the load of the device, most particularly because they have low stiction. The absolute position of the device can be read using an optical encoder, or a microscope viewing an appropriately fine vernier scale or grating: such a system should be able to measure a displacement of 10um with confidence: F = kD, so to measure a mN of thrust, the spring constant needs to be around 1000 N/m. I'm not a spring expert, but scouting around online this seems to be easily doable.
If I wasn't working so much at the moment, I'd even consider setting up the rig in the basement, since this is a cool experiment that with appropriate skill and care can actually be reproduced by an amateur, since no specific or crazy equipment should be needed.
"However I would likely not start my PhD thesis with it."
You've never written a thesis then. My own most definitely starts with a very basic and verified observable truth. The idea of a thesis is to build on knowledge and so the best place to start is to lay out what is considered to be accepted knowledge, with appropriate citations and references.
But do not be surprised that some people want to talk about it. And certainly do not be surprised when foreign nations decide not to buy US hardware and software. I'm not saying Chinese hardware is any more innocent, that's not the point I'm making, the decision that the NSA made (was anyone consulted?) to subvert the security of every PC on the planet will have repercussions. And this will have significant impact on the IT economy of the US.
As an individual wanting to use a trusted computing platform, for my own computing needs and those of my small business, I now have to look far and wide for hardware I can trust. Personally I'm keeping a close eye on the Russian effort to shrug off x86 dependence.
If they are, then the poster is ignoring the atrocities being committed over the wall in Palestine. Once the young soldiers have worked off all their energy shooting Palestinians of course they won't feel like shooting each other, or their local civilians.
If gay people in a Christian society had to tolerate being asked "1. I don't like what you're doing so I'm going to try to "fix" you." a few times a year then I think they'd be pissed, but over all no more put-out than being asked for change by the train station.
Do not under-sell the discrimination that the Christian church has extended towards gay people. Discrimination that affects you daily, limits your freedom, impairs you financially, and alters your self esteem.
There is nothing redeeming about the Church's policy against gay people.
So yes, perhaps it isn't as bad as "hang, behead, stone, or kill you in some other gruesome way" but it's still inexcusable.
Sorry, we're well off topic. Small X-Box. Hmmm, didn't Sony do this with the Playstation?
I'm no expert, but my understanding is that the device fails to work on two levels.
1) The device samples a tiny drop of blood from the periphery of the body. Regardless of the sensitivity of the analyzing device, any blood sample taken in this way (big or small) will not represent the composition of the blood in the body. It may work for some factors (everyone has seen a diabetic test their finger, but even that test is only "close enough" for them to self medicate with it) but not for all, including some significant ones. Theranos claimed they could "correct" for this. It was an extraordinary claim that required extraordinary evidence that it would work. No such evidence has been presented.
2) The device works with a tiny drop of blood. Performing the analysis with such a small sample is problematic. Sure, it will work for some factors, since we have extremely sensitive tests for them, or they are abundant in the sample and easily detected. However, there are other factors that are difficult to detect, even with the largest and most sophisticated machines on the planet sampling an entire vial of blood. It was an extraordinary claim that their testing machine could sample a tiny drop of blood and perform reliable tests. No extraordinary evidence for this claim has ever been presented.
So that's my summary, feel free to read online for a while if you want to find some citations, but it's all been reported pretty well here on Slashdot and in the online media.
You're absolutely right about the mesh network being insecure: but SO IS THE INTERNET (see NSA and Snowdon) and the physical and routing layers should always be regarded as untrusted.
The only solution is to use a secure VPN at all times. Of course you'll need an end point at some point, but I'd rather my end point was on a different continent, and ideally a country with minimal treaties with my own.
Of course performance for such a thing will be in the toilet. Ever used Tor? It's slow as well. But if faced with a choice between security and privacy (lumping them together) vs convenience, I know exactly which one I choose every time. Of course I'm not normal.
Guys, I was asking after "practical density". Of course the chip density is insane, just look at a 256 or 512GB micro SD card!!!!
There must be some practical issues getting in the way, whether it's heat, interconnects, or just plain dumb cost. However, what I was pondering, was when will we be able to buy a 10TB 3.5" form factor SSD? Just as we can for spinning rust.
They would probably burn pretty hot, but yes, the possibilities are rather startling. It will be an interesting day when SSDs overtake HDDs on practical metrics such as data density, and even more so on price.
Sounds great, choosing the correct level of atomicity is as much an art as a science. I was looking at an old project from five years ago and kept wondering "what the hell was I thinking" while looking at the multi-threading code. I've either gained some deeper insight over time, or (more likely) I've forgotten all of the domain-specific knowledge relevant to that project:)
"I try to break up the data into chunks that can each be operated on independently."
At first I thought you meant copies of the data, which was confusing since you then go on to talk about thread-safe code with mutex locking. My comment here is that if you truly break the data up into chunks then you don't need any locking or "thread-safety" since each thread has its own data. But I now realise you meant "work chunks" or "tasks": thread1 do rows 1-1000, thread2 do rows 1001-2000, etc... and of course as you outline, locking is the way to go there. As I wrote in my previous post, if you can separate the reads from the writes then you'll improve performance, but not all frameworks/languages/architectures can handle this.
Your approach sounds like you've rolled your own "temporary table" and it's great that its working well for you. My only spot of advice would be to model the failure mode carefully. It sounds like you're post-processing older data, and if there was a power failure then you'd only lose the latest job unit, which can then be re-calculated anew. If this scheme was acting on live data then the benefits of a relational database have now been lost: you may as well just run MongoDB in RAM with no backing store and cross your fingers!
Good luck with it, I have some coding to do myself, but my final comment is that I think it would be fair to say that writing software to utilise multiple cores is an excellent area to practice and develop in, as that is what the future will inevitably bring. I have some CS PhD friends working on "drop-in" parallel libraries, which perform truly magic stuff so that a serial program can be trans-piled into a parallel version. It sounds awesome, but they all tell me it's a long way away, if it will ever work at all....
"and do you really want to get on the same nightmare in 2020 with Win7"
It won't be 2020.
XP support ended in 2014. It's taken two years for a decent chunk of software to show up as unsupported on XP.
Win7 support will start to peter out in 2020, probably running until 2022 before things start to change. However, the landscape is going to change dramatically, and I wouldn't want to try and predict what things will be like 5 years from now. The current Win10 "continuous update" scheme of Microsoft may not work out so well, and software authors will look for other platforms to offer their wares. Virtualisation will continue to make a difference: so a software author may choose to dodge Win10 completely and offer their program in a self-contained Win7 VM. If Microsoft makes that impossible then we'll see a resurgence of interest in Wine and Mono. Getting all of that to work on a case-by-case basis can be tricky, but if a software developer simply provides a self-contained appliance with it all worked out then the user now has complete freedom to run the base OS of their choice.
So my guess is that anything could happen in the near future, so don't go claiming that in 2020 we'll all just miraculously shift over to Win10 regardless. My own software house now sees Win10 as a dead-end and are looking at all possible options to ensure that in the future we are not tied to its inevitable demise. But we're not doom and gloom about it, and personally I think this diversification is going to be a really good thing.
Sure, but if you tell me you've proven that the first million digits of pi don't repeat and you generated them with a computer, I can at least do some basic checks:
1) There are a million digits.
2) They don't repeat.
3) To the limit of the resources available to me I can run my own pi generation routine and ensure that my output matches yours.
That's how I was thinking about this proof, but I do realise it's different. Handling 200TB of boolean logic operations is kind of funky!!:)
You are right, and I'm not surprised in the least that you didn't see the performance gains you were hoping for.
Re-working a process to extract more performance from multiple concurrent operations depends more on the fundamental type of operation being performed than it ever does on the core count or the hardware being used.
Some tasks simply can't be processed in parallel, with some do so trivially. Most are somewhere in between, and deciding the best strategy depends on lots of factors.
You mentioned "spinning up the threads" and "scheduling them" but I can assure you that over a 10 minute time-frame those delays are trivial. Context-switching can take 10s of ms to process, which is significant for real-time high definition video, but not at all compared to your 10 minute time-frame. The most important detail will be how you handle disk access and shared memory resources. The "optimal" approach is to load one copy of the data from the HDD at the beginning, use a software framework that allows you to share access to the same data in memory, and set your ten threads in motion. They all read from a shared memory space, but write to their own output space. At the end the results should be aggregated and written back to the HDD in one go. Of course if the threads are reading and writing back from a common memory space then contention between the threads for write locks (depends a lot on the memory model, for argument sake I'll assume here writes can't be concurrent) will slow everything down. Going forwards, sharing access to the memory may not be possible in all cases, so if you have RAM to burn, make ten copies of the data and let each thread work on its own copy. This will be very fast, but takes ten times as much RAM. If RAM is limited, or there are so many data they must be constantly read and written from the HDD then you *must* be extremely careful about how you proceed. It is quite possible to achieve *less* performance than with a serial process, since read/write contention of the HDD without an appropriate level of caching will severely deteriorate performance.
Even then, the examples I give below are just the tip of the iceberg. Some tasks fundamentally rely on the results of previous steps to continue, and so there's no point having dozens of threads going if they are all waiting for the next chunk of the problem to be solved.
Slashdot Mirror
"It depends. The likely result would depend on why you didn't make it. Can verify you had to go to the hospital? Most businesses would give you a refund or credit. Something less drastic? Most businesses would give you a credit."
I really disagree with this idea. Why should someone else be able to arbiter whether my reason for not attending the concert is a "valid excuse"? In a similar way I never tell my boss why I want time off work: if I have leave available then they can either approve or deny my request. I refuse to allow others the power to adjudicate my own life.
"Scalping is illegal. End-of."
So? I think this is precisely the sort of thing that should never be "illegal" and is absolutely not a criminal matter.
If the ticket seller has a "terms of service" that re-selling tickets is against their policy, then they can take appropriate civil action against those that break their policy.
I thought the first sale doctrine was pretty strong in the US and most items could be resold without regard.
What is different about ticket sales? Can someone explain it to me?
Nearly every other country in the world prohibits or limits advertising of prescription medicine.
These countries still have a healthy pharmaceutical industry and obviously turn a profit for private companies. Remember, that sick people do need medicine and the pharmaceutical industry meets that need nicely and does a good job of it.
Chickens are modern dinosaurs and will gladly eat meat.
We have a few sturdy wooden boxes into which we regularly throw animal carcasses. The bodies are mostly scavenged from the road-side, others from shooting.
After a half week the body is crawling with maggots. We tip the whole lot out into the chicken coop. The little bastards go crazy for the wrigglers, but what may be surprising to many is that after a few days the bones will be stripped clean: the chickens happily eat the lot.
Very good for the chickens, and we feel good about keeping the neighborhood roads clear of road kill.
I too am an experimentalist and I was thinking about how to do it cheaply...
I think the simplest approach is to run the experiment twice, once with the device "on", and once with the device "off", and look for differences in the thrust, rather than trying to take a direct measurement of the thrust since as you correctly point out, there is heaps to take into account.
But I put scare quotes around "on" and "off" since how to do this will have an obvious impact on the device and the coupling to the device. Running power through cables will change their temperature, and thermal expansion may change the way they tug on the device. Same goes for the microwave guide tubes. My only idea worth sharing at this time is to set everything up as standard and then run it, and then replace one end of the cavity with a material transparent to microwaves and run it again. A high density plastic should do the trick, so replace 1mm of copper (or whatever the cavity is made from, brass? Aluminum? I don't think it matters...) with 10mm of plastic and the mass should be about the same. The only difference I would expect is that the cavity may run cooler in the second run. It should definitely be monitored for temperature, and if different by more than a degree (C) then some active cooling should be applied, which would then be the same for each run. Of course blowing a fan on it will probably produce more thrust than the device is generating :)
The thrust measurement can be directly measured against a very light spring scale provided the linear bearing has sufficiently low stiction. I'd recommend porous graphite air bearings to take the load of the device, most particularly because they have low stiction. The absolute position of the device can be read using an optical encoder, or a microscope viewing an appropriately fine vernier scale or grating: such a system should be able to measure a displacement of 10um with confidence: F = kD, so to measure a mN of thrust, the spring constant needs to be around 1000 N/m. I'm not a spring expert, but scouting around online this seems to be easily doable.
If I wasn't working so much at the moment, I'd even consider setting up the rig in the basement, since this is a cool experiment that with appropriate skill and care can actually be reproduced by an amateur, since no specific or crazy equipment should be needed.
"However I would likely not start my PhD thesis with it."
You've never written a thesis then. My own most definitely starts with a very basic and verified observable truth. The idea of a thesis is to build on knowledge and so the best place to start is to lay out what is considered to be accepted knowledge, with appropriate citations and references.
You're absolutely right, we can't.
But do not be surprised that some people want to talk about it. And certainly do not be surprised when foreign nations decide not to buy US hardware and software. I'm not saying Chinese hardware is any more innocent, that's not the point I'm making, the decision that the NSA made (was anyone consulted?) to subvert the security of every PC on the planet will have repercussions. And this will have significant impact on the IT economy of the US.
As an individual wanting to use a trusted computing platform, for my own computing needs and those of my small business, I now have to look far and wide for hardware I can trust. Personally I'm keeping a close eye on the Russian effort to shrug off x86 dependence.
Probably Israel.
If they are, then the poster is ignoring the atrocities being committed over the wall in Palestine. Once the young soldiers have worked off all their energy shooting Palestinians of course they won't feel like shooting each other, or their local civilians.
If gay people in a Christian society had to tolerate being asked "1. I don't like what you're doing so I'm going to try to "fix" you." a few times a year then I think they'd be pissed, but over all no more put-out than being asked for change by the train station.
Do not under-sell the discrimination that the Christian church has extended towards gay people. Discrimination that affects you daily, limits your freedom, impairs you financially, and alters your self esteem.
There is nothing redeeming about the Church's policy against gay people.
So yes, perhaps it isn't as bad as "hang, behead, stone, or kill you in some other gruesome way" but it's still inexcusable.
Sorry, we're well off topic. Small X-Box. Hmmm, didn't Sony do this with the Playstation?
I'm no expert, but my understanding is that the device fails to work on two levels.
1) The device samples a tiny drop of blood from the periphery of the body. Regardless of the sensitivity of the analyzing device, any blood sample taken in this way (big or small) will not represent the composition of the blood in the body. It may work for some factors (everyone has seen a diabetic test their finger, but even that test is only "close enough" for them to self medicate with it) but not for all, including some significant ones. Theranos claimed they could "correct" for this. It was an extraordinary claim that required extraordinary evidence that it would work. No such evidence has been presented.
2) The device works with a tiny drop of blood. Performing the analysis with such a small sample is problematic. Sure, it will work for some factors, since we have extremely sensitive tests for them, or they are abundant in the sample and easily detected. However, there are other factors that are difficult to detect, even with the largest and most sophisticated machines on the planet sampling an entire vial of blood. It was an extraordinary claim that their testing machine could sample a tiny drop of blood and perform reliable tests. No extraordinary evidence for this claim has ever been presented.
So that's my summary, feel free to read online for a while if you want to find some citations, but it's all been reported pretty well here on Slashdot and in the online media.
"My congressman is doing a fine job of protecting my job."
There's a lot more to good governance than this. Stop being so selfish.
You're absolutely right about the mesh network being insecure: but SO IS THE INTERNET (see NSA and Snowdon) and the physical and routing layers should always be regarded as untrusted.
The only solution is to use a secure VPN at all times. Of course you'll need an end point at some point, but I'd rather my end point was on a different continent, and ideally a country with minimal treaties with my own.
Of course performance for such a thing will be in the toilet. Ever used Tor? It's slow as well. But if faced with a choice between security and privacy (lumping them together) vs convenience, I know exactly which one I choose every time. Of course I'm not normal.
Like "hashtag" is just a stupid new word for "keyword".
"The International Data Corporation (IDC)"
Who the fuck are these guys? And why do they think they can comment on mobile phone technology and/or economics?
Holy shit, I stand 100% corrected! Thanks!
Guys, I was asking after "practical density". Of course the chip density is insane, just look at a 256 or 512GB micro SD card!!!!
There must be some practical issues getting in the way, whether it's heat, interconnects, or just plain dumb cost. However, what I was pondering, was when will we be able to buy a 10TB 3.5" form factor SSD? Just as we can for spinning rust.
Heat?
They would probably burn pretty hot, but yes, the possibilities are rather startling. It will be an interesting day when SSDs overtake HDDs on practical metrics such as data density, and even more so on price.
Any predictions on when that is going to happen?
Sounds great, choosing the correct level of atomicity is as much an art as a science. I was looking at an old project from five years ago and kept wondering "what the hell was I thinking" while looking at the multi-threading code. I've either gained some deeper insight over time, or (more likely) I've forgotten all of the domain-specific knowledge relevant to that project :)
"I try to break up the data into chunks that can each be operated on independently."
At first I thought you meant copies of the data, which was confusing since you then go on to talk about thread-safe code with mutex locking. My comment here is that if you truly break the data up into chunks then you don't need any locking or "thread-safety" since each thread has its own data. But I now realise you meant "work chunks" or "tasks": thread1 do rows 1-1000, thread2 do rows 1001-2000, etc... and of course as you outline, locking is the way to go there. As I wrote in my previous post, if you can separate the reads from the writes then you'll improve performance, but not all frameworks/languages/architectures can handle this.
Your approach sounds like you've rolled your own "temporary table" and it's great that its working well for you. My only spot of advice would be to model the failure mode carefully. It sounds like you're post-processing older data, and if there was a power failure then you'd only lose the latest job unit, which can then be re-calculated anew. If this scheme was acting on live data then the benefits of a relational database have now been lost: you may as well just run MongoDB in RAM with no backing store and cross your fingers!
Good luck with it, I have some coding to do myself, but my final comment is that I think it would be fair to say that writing software to utilise multiple cores is an excellent area to practice and develop in, as that is what the future will inevitably bring. I have some CS PhD friends working on "drop-in" parallel libraries, which perform truly magic stuff so that a serial program can be trans-piled into a parallel version. It sounds awesome, but they all tell me it's a long way away, if it will ever work at all....
"and do you really want to get on the same nightmare in 2020 with Win7"
It won't be 2020.
XP support ended in 2014. It's taken two years for a decent chunk of software to show up as unsupported on XP.
Win7 support will start to peter out in 2020, probably running until 2022 before things start to change. However, the landscape is going to change dramatically, and I wouldn't want to try and predict what things will be like 5 years from now. The current Win10 "continuous update" scheme of Microsoft may not work out so well, and software authors will look for other platforms to offer their wares. Virtualisation will continue to make a difference: so a software author may choose to dodge Win10 completely and offer their program in a self-contained Win7 VM. If Microsoft makes that impossible then we'll see a resurgence of interest in Wine and Mono. Getting all of that to work on a case-by-case basis can be tricky, but if a software developer simply provides a self-contained appliance with it all worked out then the user now has complete freedom to run the base OS of their choice.
So my guess is that anything could happen in the near future, so don't go claiming that in 2020 we'll all just miraculously shift over to Win10 regardless. My own software house now sees Win10 as a dead-end and are looking at all possible options to ensure that in the future we are not tied to its inevitable demise. But we're not doom and gloom about it, and personally I think this diversification is going to be a really good thing.
No.
Next question...
Sure, but if you tell me you've proven that the first million digits of pi don't repeat and you generated them with a computer, I can at least do some basic checks:
1) There are a million digits.
2) They don't repeat.
3) To the limit of the resources available to me I can run my own pi generation routine and ensure that my output matches yours.
That's how I was thinking about this proof, but I do realise it's different. Handling 200TB of boolean logic operations is kind of funky!! :)
You are right, and I'm not surprised in the least that you didn't see the performance gains you were hoping for.
Re-working a process to extract more performance from multiple concurrent operations depends more on the fundamental type of operation being performed than it ever does on the core count or the hardware being used.
Some tasks simply can't be processed in parallel, with some do so trivially. Most are somewhere in between, and deciding the best strategy depends on lots of factors.
You mentioned "spinning up the threads" and "scheduling them" but I can assure you that over a 10 minute time-frame those delays are trivial. Context-switching can take 10s of ms to process, which is significant for real-time high definition video, but not at all compared to your 10 minute time-frame. The most important detail will be how you handle disk access and shared memory resources. The "optimal" approach is to load one copy of the data from the HDD at the beginning, use a software framework that allows you to share access to the same data in memory, and set your ten threads in motion. They all read from a shared memory space, but write to their own output space. At the end the results should be aggregated and written back to the HDD in one go. Of course if the threads are reading and writing back from a common memory space then contention between the threads for write locks (depends a lot on the memory model, for argument sake I'll assume here writes can't be concurrent) will slow everything down. Going forwards, sharing access to the memory may not be possible in all cases, so if you have RAM to burn, make ten copies of the data and let each thread work on its own copy. This will be very fast, but takes ten times as much RAM. If RAM is limited, or there are so many data they must be constantly read and written from the HDD then you *must* be extremely careful about how you proceed. It is quite possible to achieve *less* performance than with a serial process, since read/write contention of the HDD without an appropriate level of caching will severely deteriorate performance.
Even then, the examples I give below are just the tip of the iceberg. Some tasks fundamentally rely on the results of previous steps to continue, and so there's no point having dozens of threads going if they are all waiting for the next chunk of the problem to be solved.
It's pretty fascinating stuff.
"With todays memory densities it really is beyond me that the option to put ECC memory on a $1700 CPU is disabled."
Pure marketing strategy. Not saying I agree with it, but the ones that support ECC can demand a higher price, so they are priced higher.
"Most of the time computers are used to shitpost on forums and look at porn/cat videos."
OR do all three at once!!! SHIT POST!!!! :)
"that is unreadable unless you know enough maths to not need to look it up."
And if you're looking it up on Wikipedia then you'll be caught in a never-ending loop of inadequate comprehension... :)