Speaking of crazy technology, would it be possible to build an audiophile system where basically you used very high quality microphones similar to what they used to record orchestras etc. Then the system plays back stuff through reasonably decent speakers[1] in the room, records the result using the microphones, and keeps "practising" till the playback is as close to perfect as possible (the microphones produce close to the same values as the studio put[2] in the recording).
The advantage of this "practising" approach is the filtering or changes do not have to start "after" each sound, they can actually start "before" the sound - in anticipation of it. This allows your time-domain filters a lot more flexibility;).
The disadvantage of course is it could take a while before the system gets good at a particular performance in a particular room (and weather - humidity, pressure, will affect it). And you probably have to be sitting very close to where the microphones are for the performance. Of course with some intelligence and heuristics you could still have it play a new piece of music adequately - but for best performance you need to let it practice for a while.
It might be easier to do a variation of the concept with very fancy headphones (you may need a dummy head with mikes).
[1] The speakers could also be digitally controlled - the position of the diaphragms could be measured optically and a very powerful amplifier would "slam" the diaphragm to the desired position at very high speeds. If this can be done accurately 192000 times a second your ears would be the low pass filter (some animals might not be impressed but too bad;) ).
FWIW I've been thinking about this stuff since the 1980s and 1990s, but I've been working in IT not audio... The technology level probably wasn't good enough back then anyway.
[2] Yes I know the studio will usually do a lot of processing, but if their release version sounds the same to you as it does to them then that's "high fidelity" right?
That's why I don't install AV software on my PC. I'm less likely to screw up than AV vendors are. Seriously. My own PCs have NEVER been infected by a virus. And yes I know how to check, and I know how to upload suspicious stuff to VirusTotal, and I know how to run browsers with different user accounts from my main account. Whereas the AV vendors make this sort of screw up every few years. So it's no point for me to slow down my computer with AV software. The sort of malware that would infect me would probably not be detected by their stuff anyway.
BUT I do install AV software on other people's PCs. Since they do screw up more often. Despite that my sister somehow still managed to get her PC infected, and the AV software (Avira) just wouldn't detect or clean it...
I don't put AV software on production servers either unless PHBs etc require it. In my experience if you do things right, AV software is more likely to cause you problems than a virus.
How's that a deterrent? Email addresses are a dime a thousand...
The more obvious reason is they are still keeping your account and data around, and they aren't bothering to set things up so that new accounts can be created with email addresses from old deleted accounts.
If you go the delete route, they claim to delete your data, but more importantly, you can never use that email address again to open a facebook account.
Sounds a bit contradictory to me... I suppose they could hash your email address instead of storing it. But I doubt they do that.
The thing is if he could make Iron Man suits so quickly and had something like Jarvis, he could actually make an army of robots and drones with iron man tech, and they could still be under his control via Jarvis and similar AIs. If I were a multi billionaire with such advanced obedient AI tech, I'd have my own private "security force".
But yeah it's not all about realism. Otherwise Peter Parker would be a rich man just by selling his webbing formula and web shooter technology- not a stupid journalist (which is why I found the organic webshooter thing more plausible).
FWIW MSSQL defaults to 255 worker threads, which is likely to be more than the number of logical CPUs in most servers.
If you're the OP AC, you can try reduce your max worker threads to "n+1 logical CPUs" on a 1500 connection test DB server and see if the DB performs better.
I doubt it will. The thing is a thread of execution is a useful concept for a programmer - you set up a thread to handle each task and let the OS worry about multiplexing efficiently across logical/physical/whatever CPUs. Same goes for processes. Trying to squeeze out more performance by doing the OS's job should only be reserved for very specific cases. One day the rules could change (hot-pluggable/asymmetric CPUs etc), you'd then have to modify the app if you did the OS's job, whereas if you didn't you just wait for OS people to issue an update, and then you get the performance increase.
Similarly some DBs have an option to bypass the OS filesystem stuff (e.g. Oracle RAW device). But I don't think most sane admins would want to do that. Having the OS take care of it allows you to do a lot of nice stuff.
Without the conditions I state, HFT would not actually be trading but just a fancy way of disguising the transfer of money from nonHFT/nonfavoured traders to favoured HFT traders.
After all I can easily make money in a casino/market if: 1) My big bad bets are rolled back 2) I get a bailout (and keep my bonus +commisions) if I screw up really big time - betting using other people's money! 3) I get to see other people's planned moves before their move takes effect AND change my mind accordingly.
A vacuum updates the frozen TID, which is a trivial operation and allows a subsequent TID to safely wrap around.
What if you have at least one outstanding transaction/connection? Can vacuum update the frozen TID then?
For example if you have a transaction that's open for a few weeks and happen to have 4 billion transactions during that time.
I believe perl DBI/DBD in AUTOCOMMIT OFF mode starts a new transaction immediately after you commit or rollback. So if you have an application using that library that is idling for weeks a transaction would presumably be open for the entire time- since it would be connected to the database and already have started a transaction. This is probably also true for some python and ruby db libraries.
This could also happen if someone starts up psql, starts a transaction and goes on vacation;).
Ideally, any single service/application should NEVER have more threads than there are n+1 logical CPUs.
In the ideal world you'll never have more nonparallelizable tasks than you have CPUs.
However in the real world you often do. It is usually better for the application developers to focus on having their application solve the application related problems, and let the OS take care of the multitasking and other OS related problems.
A process per client also means that if a process crashes it is less likely to affect other clients. And if there are memory leaks for whatever weird/stupid reason, if you close that process, it frees up the mem and does not affect other processes. Compare Google Chrome and Firefox - Chrome often actually uses more memory for a given set of pages, but just close the unwanted tabs and windows and the mem is freed up, with Firefox sometimes due to plugins, bugs etc you have to close the ENTIRE browser to free up the memory. You can usually get away with having a DB process quit every now and then, but people are more likely to notice if you keep restarting the entire DB.
Of course if you do ever need 1000 or more simultaneous DB connections you probably need a different solution than a single server running plain Postgresql.
The rate at which these decisions are being made indicates that it is not going through a human mind.
Why's that a problem? Index trackers don't involve human minds either.
I'm fine with HFT. My only conditions are: 1) No rollbacks for HFT trades[1]. You screw up you eat the loss. 2) If bailouts are needed for whatever reason (your company loses billions of other people's money), the traders involved (if any) and the bosses go to jail for 20 years. 3) The exchange only allows you to see what everyone else sees. No "peeking at other people's cards".
If they still do HFT with these conditions we might eventually see an improvement in algorithms, software quality and testing.
[1] Rollbacks are only allowed if it's not your fault e.g. the casino aka exchange screws up big time (slow downs don't count, going down doesn't count, exchange treating 1=2 counts, exchange treating buy as sell counts.).
I don't normally use toilet paper (at least for cleaning butts - seems unhygienic to me). My worry would be running out of clean water and soap...
As for the topic - the big issue is if you allow speculators to corner the market on wheat or other food commodity, raise the prices and profit from it. The finance bunch have enough money and leverage to do that. And I believe they may have already done so before.
Even if that happens the rich countries will have food. The poor countries won't.
They eventually gave you your cash back, but how many people would do what you did and fight them for the money?
It's just a way of stealing lots of money from very many people. The telcos get a cut, so their bosses don't care.
If you stole even 20 bucks from someone, they call the cops on you and you'd be in trouble, but the Telcos and their partners get away with stealing from thousands and thousands of people.
Actually different teachers around the world could put up their videos on the same topics.
And the students can go figure out which teachers they understand better.
Then teachers can spend more time on trying to teach the students who still have problems understanding stuff. Or figuring out if the students really understand stuff or even have mastered the topic.
Might take another 20-50 years before that'll happen.
Slashdot Mirror
Speaking of crazy technology, would it be possible to build an audiophile system where basically you used very high quality microphones similar to what they used to record orchestras etc. Then the system plays back stuff through reasonably decent speakers[1] in the room, records the result using the microphones, and keeps "practising" till the playback is as close to perfect as possible (the microphones produce close to the same values as the studio put[2] in the recording).
;).
;) ).
The advantage of this "practising" approach is the filtering or changes do not have to start "after" each sound, they can actually start "before" the sound - in anticipation of it. This allows your time-domain filters a lot more flexibility
The disadvantage of course is it could take a while before the system gets good at a particular performance in a particular room (and weather - humidity, pressure, will affect it). And you probably have to be sitting very close to where the microphones are for the performance. Of course with some intelligence and heuristics you could still have it play a new piece of music adequately - but for best performance you need to let it practice for a while.
It might be easier to do a variation of the concept with very fancy headphones (you may need a dummy head with mikes).
[1] The speakers could also be digitally controlled - the position of the diaphragms could be measured optically and a very powerful amplifier would "slam" the diaphragm to the desired position at very high speeds. If this can be done accurately 192000 times a second your ears would be the low pass filter (some animals might not be impressed but too bad
FWIW I've been thinking about this stuff since the 1980s and 1990s, but I've been working in IT not audio... The technology level probably wasn't good enough back then anyway.
[2] Yes I know the studio will usually do a lot of processing, but if their release version sounds the same to you as it does to them then that's "high fidelity" right?
That's why I don't install AV software on my PC. I'm less likely to screw up than AV vendors are. Seriously. My own PCs have NEVER been infected by a virus. And yes I know how to check, and I know how to upload suspicious stuff to VirusTotal, and I know how to run browsers with different user accounts from my main account. Whereas the AV vendors make this sort of screw up every few years. So it's no point for me to slow down my computer with AV software. The sort of malware that would infect me would probably not be detected by their stuff anyway.
BUT I do install AV software on other people's PCs. Since they do screw up more often. Despite that my sister somehow still managed to get her PC infected, and the AV software (Avira) just wouldn't detect or clean it...
I don't put AV software on production servers either unless PHBs etc require it. In my experience if you do things right, AV software is more likely to cause you problems than a virus.
Yeah. "Random Salesman CEO Spouts Nonsense Showing His Lack Of Clue".
How's that a deterrent? Email addresses are a dime a thousand...
The more obvious reason is they are still keeping your account and data around, and they aren't bothering to set things up so that new accounts can be created with email addresses from old deleted accounts.
If you go the delete route, they claim to delete your data, but more importantly, you can never use that email address again to open a facebook account.
Sounds a bit contradictory to me... I suppose they could hash your email address instead of storing it. But I doubt they do that.
Yeah, in the bright new future the unemployed peasants will have jobs tending farms on Facebook.
They'll sell their produce to the farmers in Iowa and elsewhere...
What would you consider a reasonable time frame for travel, and how many hours or days do you plan to spend looking at each planet?
There are very many planets out there...
If possible I'd have robotic ships do the exploring and then providing a shortlist of interesting ones.
Not the same. You have to pay mercenaries. Privateers get their profits from who they attack.
Privateers were basically pirates that a Government officially allows to rob ships of another country.
Your phone may last longer than the company that made it.
You could reduce the risk of theft by making your phone very very distinctive in a permanent way.
Ugly helps - example: http://www.androidauthority.com/good-guy-samsung-gives-free-custom-s3-lucky-facebook-fan-exchange-dragon-drawing-110858/
Low fence value, I bet the thief's girlfriend won't accept it either.
It's not 100% of course.
Yeah I found Jarvis rather noteworthy too.
The thing is if he could make Iron Man suits so quickly and had something like Jarvis, he could actually make an army of robots and drones with iron man tech, and they could still be under his control via Jarvis and similar AIs. If I were a multi billionaire with such advanced obedient AI tech, I'd have my own private "security force".
But yeah it's not all about realism. Otherwise Peter Parker would be a rich man just by selling his webbing formula and web shooter technology- not a stupid journalist (which is why I found the organic webshooter thing more plausible).
When can I buy one? This would make my welding so much easier
Before you buy one I recommend you find out what the minimum, maximum and typical latency is.
High latency would make welding harder not easier. Plus give some people motion sickness...
Upcoming Atom, Haswell? Meh... I'm still waiting for the Roswell CPUs...
FWIW MSSQL defaults to 255 worker threads, which is likely to be more than the number of logical CPUs in most servers.
If you're the OP AC, you can try reduce your max worker threads to "n+1 logical CPUs" on a 1500 connection test DB server and see if the DB performs better.
I doubt it will. The thing is a thread of execution is a useful concept for a programmer - you set up a thread to handle each task and let the OS worry about multiplexing efficiently across logical/physical/whatever CPUs. Same goes for processes. Trying to squeeze out more performance by doing the OS's job should only be reserved for very specific cases. One day the rules could change (hot-pluggable/asymmetric CPUs etc), you'd then have to modify the app if you did the OS's job, whereas if you didn't you just wait for OS people to issue an update, and then you get the performance increase.
Similarly some DBs have an option to bypass the OS filesystem stuff (e.g. Oracle RAW device). But I don't think most sane admins would want to do that. Having the OS take care of it allows you to do a lot of nice stuff.
If you have special software it might even be less error prone and faster than paper.
You can make ipads water resistant too:
http://brentsgeekyramblings.blogspot.com/2012/02/liquipel-ipad-test.html
And phones of course:
http://www.pcworld.com/article/243827/hzo_nanotechnology_seal_keeps_smartphones_from_drowning.html
But if someone sits/steps on it...
To me it goes past indifference to declare a match when there is none.
She's definitely a very dangerous criminal in my eyes.
Someone told me that 1) is implemented already.
Without the conditions I state, HFT would not actually be trading but just a fancy way of disguising the transfer of money from nonHFT/nonfavoured traders to favoured HFT traders.
After all I can easily make money in a casino/market if:
1) My big bad bets are rolled back
2) I get a bailout (and keep my bonus +commisions) if I screw up really big time - betting using other people's money!
3) I get to see other people's planned moves before their move takes effect AND change my mind accordingly.
A vacuum updates the frozen TID, which is a trivial operation and allows a subsequent TID to safely wrap around.
What if you have at least one outstanding transaction/connection? Can vacuum update the frozen TID then?
For example if you have a transaction that's open for a few weeks and happen to have 4 billion transactions during that time.
I believe perl DBI/DBD in AUTOCOMMIT OFF mode starts a new transaction immediately after you commit or rollback. So if you have an application using that library that is idling for weeks a transaction would presumably be open for the entire time- since it would be connected to the database and already have started a transaction. This is probably also true for some python and ruby db libraries.
This could also happen if someone starts up psql, starts a transaction and goes on vacation ;).
Ideally, any single service/application should NEVER have more threads than there are n+1 logical CPUs.
In the ideal world you'll never have more nonparallelizable tasks than you have CPUs.
However in the real world you often do. It is usually better for the application developers to focus on having their application solve the application related problems, and let the OS take care of the multitasking and other OS related problems.
A process per client also means that if a process crashes it is less likely to affect other clients. And if there are memory leaks for whatever weird/stupid reason, if you close that process, it frees up the mem and does not affect other processes. Compare Google Chrome and Firefox - Chrome often actually uses more memory for a given set of pages, but just close the unwanted tabs and windows and the mem is freed up, with Firefox sometimes due to plugins, bugs etc you have to close the ENTIRE browser to free up the memory. You can usually get away with having a DB process quit every now and then, but people are more likely to notice if you keep restarting the entire DB.
Of course if you do ever need 1000 or more simultaneous DB connections you probably need a different solution than a single server running plain Postgresql.
The rate at which these decisions are being made indicates that it is not going through a human mind.
Why's that a problem? Index trackers don't involve human minds either.
I'm fine with HFT. My only conditions are:
1) No rollbacks for HFT trades[1]. You screw up you eat the loss.
2) If bailouts are needed for whatever reason (your company loses billions of other people's money), the traders involved (if any) and the bosses go to jail for 20 years.
3) The exchange only allows you to see what everyone else sees. No "peeking at other people's cards".
If they still do HFT with these conditions we might eventually see an improvement in algorithms, software quality and testing.
[1] Rollbacks are only allowed if it's not your fault e.g. the casino aka exchange screws up big time (slow downs don't count, going down doesn't count, exchange treating 1=2 counts, exchange treating buy as sell counts.).
I don't normally use toilet paper (at least for cleaning butts - seems unhygienic to me). My worry would be running out of clean water and soap...
As for the topic - the big issue is if you allow speculators to corner the market on wheat or other food commodity, raise the prices and profit from it. The finance bunch have enough money and leverage to do that. And I believe they may have already done so before.
Even if that happens the rich countries will have food. The poor countries won't.
Your problem is there are too many frigging AIs around. The machines have won...
Bees have been co-evolving with bacteria and fungi for millions of years in a battle to make a microbe-proof honey. The bees won.
It's not microbe proof - spores of clostridium botulinum have been found in honey.
Wheat has a good shelf life if stored in dry cool conditions. Sugar stores for an even longer time.
They eventually gave you your cash back, but how many people would do what you did and fight them for the money?
It's just a way of stealing lots of money from very many people. The telcos get a cut, so their bosses don't care.
If you stole even 20 bucks from someone, they call the cops on you and you'd be in trouble, but the Telcos and their partners get away with stealing from thousands and thousands of people.
Actually different teachers around the world could put up their videos on the same topics.
And the students can go figure out which teachers they understand better.
Then teachers can spend more time on trying to teach the students who still have problems understanding stuff. Or figuring out if the students really understand stuff or even have mastered the topic.
Might take another 20-50 years before that'll happen.