However, it does make sense for Vonage to at least try to play nice with the network - utilizing compression by default, etc.
Don't get me wrong - I'd be using Vonage myself if my phone company let me drop my local line but keep DSL. However, Vonage can't claim to have clean hands unless they take reasonable steps to keep network intrusion to a minimum.
The parent was simply pointing out that Vonage by default makes no attempt to compress the data much at all. You don't need the same kind of datarates for phone as you do from an mp3 file - if they just use a little compression they could drop network usage by quite a bit.
Uh, are you suggesting you want to ban overselling?
We could do that by law, but then nobody would sell broadband for less than $500/month. You simply can't afford to provide it for less than that right now.
What you get with your $25 DSL subscription is an agreement that they'll give you a lot of bandwidth as long as you only use it for generally-interactive network usage and not use protocols that tie up the network in a disproportionate manner.
ISPs don't normally police this strictly, but they do put their foot down when their costs go up substantially.
Take a poll of what most DSL customers want:
1. Allow Vonage - everybody's bill goes up by $10/month (including non-Vongage customers).
2. Allow Vonage with an extra $30/mo fee for only Vonage customers.
3. Ban Vonage.
Most people would pick #2 or 3.
Don't get me wrong - I sometimes play fast and loose with the ISP terms and conditions, but I try to be considerate and throttle network usage. If they complained I'd throttle further.
I don't expect the same SLA and TOS for a residential DSL line as a commercial T1. You only get what you pay for.
It is easy to say "well then, don't sell DSL if you can't afford the bandwidth" - but do we really want to not be able to buy DSL?
Imagine if sending email directly from client boxes was the norm.
This was the norm for me until recently when everybody and their uncle started blocking dynamic IPs...
There really should be some certificate-based whitelist system for mail servers. Who cares what your IP happens to be - are you a trustworthy mail server or not. Allow users to apply and grant a graylist certificate of some sort. Then after a trial period whitelist them, and then blacklist them the instant there is abuse. The certificate could be granted via a mechanism like cacert.org - email verification. That would keep the certs free, and would work as a domain verification system...
If they do not have a monopoly, then who cares. It is their business.
I'm not sure this business model could work.
Suppose the deal is that the public allows a cable company to come in a wire up a city in exchange for a 10 year monopoly.
After 10 years, in theory somebody else can come in and run their own wires, but nobody would ever do so unless offered a monopoly of their own. The reason is that although the incumbent company may be charging very high prices at the moment, they'd drop them the instant a competitor actually got up and running. Since the incumbent starts out with 100% market share, the competitor wouldn't be able to pay for their loans on all that cable they just ran and they'd go out of business. At this point, the incumbent buys their cables since they're the only company that can profitably use them, and they raise their prices again.
Wires on the ground tend to be a natural monopoly, and consequently they tend to require regulation. I'm open to other scenarios that would eliminate the need for regulation, but in most cases it is a necessary evil...
The simulation might need to increase its recording rate of people's minds leading up the the event. Or whatever.
Only if the masters of such a universe made a mistake or intended this result.
If you are simulating a universe you can make as many measurements as you want, at whatever simulated-time-intervals that you want.
For example, suppose I make a simulator that simulates a perfect electronic oscillator (no noise, etc.). I can have it generate a sine wave at a frequency of 1E9999 Hz with perfect accuracy. I can sample that waveform at a rate of 1E999999999999 Hz. I can do all this on an 8088 (assuming that it had enough storage to hold the desired number of samples).
How can I do this? Simple - I never said I could do it in realtime. I could run my simulation over 100 simulated hours with perfect accuracy (well, within arbitrary precision, at least). Sure, it might take me 100,000 years to run the simulation, but I'd get the data and it wouldn't perturb the simulation at all.
Likewise - if our universe is simulated there is no reason to believe that it is simulated in realtime. It is possible that 1000 years in the parent universe pass in the time it takes for one second to pass in our "simulated" universe. If they need to collect more data they can just slow down the simulation accordingly, and not sacrifice accuracy. If you were running a simulation of a scale the size of our universe and the apparent precision on the order of a plank length/time, then why would you mess up the laws of physics your simulation simply to collect a little more data for a short interval? The "butterfly effect" would make the long-term results of your simulation worthless after you've done this even once.
If somebody were taking care to simulate an entire universe you'd have to assume that there would only be apparent errors in the simulation if the simulation designer intended them to be there...
And this is the fault. Randi does not have a set guideline, nor does a disinterested party have control of either the funds or the testing procedures.
How can you have a set guideline for:
1. Proof of the existance of telepathy. 2. Proof of the ability to levitate objects. 3. Proof of the ability to predict that something "bad" will happen "soon". 4. etc...
His system is that you tell him what you claim to be able to do, and then you agree with him on what constitutes the guidelines. It is hard to have a "standard" set of guidelines for a huge variety of potential claims.
If a psychic is upset they should simply publish what the unfair proposed guidelines were, and let people draw their own conclusions...
That's one way to do it. Or, if you don't like wasting time, just go with the straight microsoft ticket. Install all the software, run windows update once (ok probably twice as you need to install some things sepperatetly) and you are on your way.
Uh, windowsupdate only updates windows, and a few MS programs.
If I have a Quicken security problem, windowsupdate won't fix it.
If kmymoney has a security problem, Red Hat will distribute a patch for it.
So, we're comparing the number of patches to the MS Windows OS, to the number of patches for just about every free piece of software available for linux...
World of Warcraft PvP server players have started compiling lists of known farmers working in shifts and are going around pounding them flat whenever they can (more power to em).
Probably a losing battle unless it is done by the server admins (ie they just massacre them by the legion in a puff of logic).
Otherwise it just raises the cost of doing business. You'll have a field of farmers flanked by mercenaries armed to the teeth. If they can pay farmers, then they can pay guards...
If you can do this, you guys can all come over to my place for a front seat view of the resulting fireworks.
I think I'll take a back seat. The asteroid impact calculator says that it will have an impact energy of 594 megatons, which is probably something I'd try to stay far away from.
It will, however, be quite effective at getting rid of that Best Buy.
Money does matter to the kin and to the dying, who care about their families.
This is of course a very controversial topic.
My observation is that most people put a dollar figure on life. However, anybody who openly talks about this is looked down upon. It is acceptable to be a bean counter, it just isn't acceptable to actually say that you are one, or that bean counting was the reason for a decision you made.
Kind of like a company I know somebody at. They were doing some testing to marginally improve the safety of their products and to better comply with industry-wide regulations. Two managers got in a discussion, and one brought up cost-benefit analysis. The superior chastised him - safety should come first no matter what the cost. And yet, the project had a budget, and if any body suggested increasing it tenfold they'd be fired. So, there was a cost-benefit analysis, it just was never discussed or analyzed to make sure it was done right.
The problem with not talking about the costs of saving lives is that we end up misallocating resources. We'll spend billions of dollars on medicare so that old people live a few months longer on hospital beds, but we won't spend any money on decent foster homes, so that abused children don't have to live with their abusers. Which is the bigger health problem - people who are going to die in a year no matter what, or people who have a whole life ahead of them who could be destined for greatness or the ghetto? And which costs more, trying to push back death two weeks for somebody diagnosed with a terminal disease, or creating some incentives for half-decent families to adopt kids who would otherwise be abandoned to abuse?
Don't get me wrong - I'm all for reasearch to improve the quality and length of life. However, cutting-edge treatments simply aren't for everyone. We should be researching how cure difficult diseases, and we should let the rich pay for cutting-edge treatment if they so desire. As more doctors become educated in these treatments they will become mainstream, and available to everyone (possibly with government assistance).
There is a price in human life even when we try to save lives - if we don't count the costs we may harm more than we help...
Uh, Merck, despite losing probably 20+% of its revenue over the last six months is still running a profit. It may go through some rough times with upcoming patent expirations, but it is still expected to make it through reasonably intact. (It just doesn't have the world's greatest stock growth potential right now.)
On the other hand, the US Federal government has been losing money for decades, even though it can raise prices (taxes) anytime it feels like it...
Most companies are reasonably thrifty. Ones that are having financial issues (like Merck) tend to be even moreso.
After all, those photons are moving (oddly enough) at the speed of light
I think he was talking about the neutrinos - which are moving very close to the speed of light (from our frame of reference).
In fact - they're moving so close to light speed that I don't think anybody has come up with an experiment that can measure how much slower than c they are actually travelling. After all, if people knew they weren't quite at light speed it wouldn't have been such a big discovery when it was found that they have mass. The only way this was discovered was because of oscillations, which imply mass. However, it does not imply the amount of mass, which is already known to be extrememly small (just not how small).
So, my guess is that VERY little time passes in the neutrino's frame before it hits earth. It probably is a VERY small fraction of a second.
1. Yes - if your filesystem code crashes, you could end up with a dirty filesystem.
2. Yes - if your hard drive code crashes, you could end up with a dirty hard drive.
But:
3. No - if your webcam driver crashes, you won't end up with a dirty hard drive.
Right now with linux, if a kernel-level driver of any kind panics, the whole thing goes down the tubes.
Certainly a little compartmentalization can't possibly hurt. It won't fix every problem, but it does prevent a small problem in a non-essential driver from taking down the whole system.
As you point out, it will still be critical for some pieces of code to just work without bugs at all. However, the amount of that code can be reduced in a microkernel design.
Also - I don't think TWAIN is windows-specific. I seem to recall using TWAIN on a Mac many a year ago...
Agreed - however that is purely a matter of getting the browser maintainers to add their cert to the list of trusted root CAs.
For MS that might take some time. I would have hoped that Firefox would be more enlightened, but they've been talking about it for 1.5 years now without actually doing much about it...
I would suspect designing and building an orbital telescope to equal Hubble would cost many billions of dollars.
Why design anything?
Why not just build a new one just like the existing Hubble? Maybe upgrade a few parts where it is cheap to do so, but nothing that would require extensive re-integration.
Rather than designing fancy super-upgradable telescopes why not just make one that does a single job, launch it, and then start on the next one. If one fails, just send up an EXACT copy. It can't cost nearly as much just to rebuild the same thing, can it?
kontact/kmail will only do it with gnupg 1.9.x. This is considered alpha code, and this really isn't the kind of application you want to use alpha code for...
When I went to grad school the first thing I did was enable my remote shell accounts to use s/key, and I printed out 100 passwords and kept them on me.
Universities tend to be notorious for that sort of thing. My main concern was actually sniffing, though.
Speaker cables are unbalanced. There is also no such thing as "send and receive" on speaker cables.
It is true that speaker cables are generally not prone to interference, but that is because there is no signal amplification once they get to their destination.
Illustration:
Your signal comes in from the DVD player - a few mW in total. If there is a lot of noise nearby, it could be noticable, since noise is probably in the uW-mW range. Your amp then boosts whatever input it gets to 1000W and sends it to the speakers. Now if you add a few uW of noise it isn't noticable at all, but any noise that came in via the DVD input got boosted by the amp and is quite noticable.
A balanced signal is one which is sent via twisted pair where each wire in the pair is 180 degrees out of phase with respect to the other, and where the sum of the two signals is ground. This is noise resistant, since incoming noise will be in-phase in each cable, and when the signals are inverted and combined the noise will cancel out.
Unbalanced signals are still differential - you can just take an unbalanced microphone signal and put it into a speaker and you will get sound (albeit virtually inaudible). In fact, a speaker will function as a very poor microphone - if you put two speakers back-to-back on the same line, if you speak into one it vibrates the diaphram, and induces current in the wire, which then travels to the other speaker and causes its diaphram to move.
Where you do benefit in running balanced signals to speakers is when the speakers are powered - since you're now applying gain to the signal.
I'm pretty sure that just about every balanced system out there uses at least three wires. Ethernet uses only 2 for each way (for 10BaseT), but that's probably because the plug in the wall provides a ground reference. Is anyone aware of a battery-operated device that successfully uses an ethernet interface with no other cables attached?
Disclaimer - I'm not an EE. And if somebody who really is one cares to illuminate this discussion, please feel free to do so. However, kindly provide more than "you're wrong"...
Suppose your two 44.1 kHz reference clocks on two soundcards are off by 10 Hz. That is 10 samples per second of drift. For a 5-minute song that is only 3000 samples drift, or 68ms (maybe noticable).
For a 20-minute movie soundtrack or classical work you now have 12000 samples drift, which is over 1/4th of a second. You don't need golden ears to hear that - it will sound like an echo coming from the other room.
If you're playing the sound from a movie you're watching on TV, the final lines of the 120 minute movie will be delayed by 1.6 seconds in the other room. That will be VERY noticable.
Scale all times in accordance with actual sound card frequency ref error. If it is only 1 Hz, then the movie will still be noticable at 160ms, but the rest should be fine.
With multicast you should be better - just the 1ms of delay associated with network latency, which you won't hear. Occassionally a sound card might cut out for 1/10th of a second if it gets too far ahead of the data stream, and then it will sync up with the stream rate. Sound cards slower than the stream rate will continue to fall farther and farther behind, however, as long as they are caching the network data.
Of course, if you multicast the whole stream at network speeds, then everything will get sent in 25 seconds, and everything will go back out of sync again...
However, you also benefit from much-more-tailored build processes. I'm not just talking tailored to your CPU - but also to your preferences.
Do you prefer KDE or GNOME? You can tell gentoo your preference and it will build apps accordingly, not compiling in needless support for the other, and consequently eliminating depedencies that you don't care for.
That and the package management system works rather well - you could get many benefits from using gentoo and only installing binary packages...
When you're getting down to the single nm range, we're no longer talking UV-lithography - you need X-rays. It probably will also get harder to come up with photosensitive dyes that respond to X-rays...
Sure, and many of us know how to go about recovering hard drives, etc. Still, we all run backups and RAID since it is much easier just to hot-swap a drive, or restore a backup.
I think his point is that while all admins should understand the inner-workings of their systems, they shouldn't expect to have to dive into them day-to-day on every box they manage...
Slashdot Mirror
However, it does make sense for Vonage to at least try to play nice with the network - utilizing compression by default, etc.
Don't get me wrong - I'd be using Vonage myself if my phone company let me drop my local line but keep DSL. However, Vonage can't claim to have clean hands unless they take reasonable steps to keep network intrusion to a minimum.
The parent was simply pointing out that Vonage by default makes no attempt to compress the data much at all. You don't need the same kind of datarates for phone as you do from an mp3 file - if they just use a little compression they could drop network usage by quite a bit.
Uh, are you suggesting you want to ban overselling?
We could do that by law, but then nobody would sell broadband for less than $500/month. You simply can't afford to provide it for less than that right now.
What you get with your $25 DSL subscription is an agreement that they'll give you a lot of bandwidth as long as you only use it for generally-interactive network usage and not use protocols that tie up the network in a disproportionate manner.
ISPs don't normally police this strictly, but they do put their foot down when their costs go up substantially.
Take a poll of what most DSL customers want:
1. Allow Vonage - everybody's bill goes up by $10/month (including non-Vongage customers).
2. Allow Vonage with an extra $30/mo fee for only Vonage customers.
3. Ban Vonage.
Most people would pick #2 or 3.
Don't get me wrong - I sometimes play fast and loose with the ISP terms and conditions, but I try to be considerate and throttle network usage. If they complained I'd throttle further.
I don't expect the same SLA and TOS for a residential DSL line as a commercial T1. You only get what you pay for.
It is easy to say "well then, don't sell DSL if you can't afford the bandwidth" - but do we really want to not be able to buy DSL?
Imagine if sending email directly from client boxes was the norm.
This was the norm for me until recently when everybody and their uncle started blocking dynamic IPs...
There really should be some certificate-based whitelist system for mail servers. Who cares what your IP happens to be - are you a trustworthy mail server or not. Allow users to apply and grant a graylist certificate of some sort. Then after a trial period whitelist them, and then blacklist them the instant there is abuse. The certificate could be granted via a mechanism like cacert.org - email verification. That would keep the certs free, and would work as a domain verification system...
If they do not have a monopoly, then who cares. It is their business.
I'm not sure this business model could work.
Suppose the deal is that the public allows a cable company to come in a wire up a city in exchange for a 10 year monopoly.
After 10 years, in theory somebody else can come in and run their own wires, but nobody would ever do so unless offered a monopoly of their own. The reason is that although the incumbent company may be charging very high prices at the moment, they'd drop them the instant a competitor actually got up and running. Since the incumbent starts out with 100% market share, the competitor wouldn't be able to pay for their loans on all that cable they just ran and they'd go out of business. At this point, the incumbent buys their cables since they're the only company that can profitably use them, and they raise their prices again.
Wires on the ground tend to be a natural monopoly, and consequently they tend to require regulation. I'm open to other scenarios that would eliminate the need for regulation, but in most cases it is a necessary evil...
The simulation might need to increase its recording rate of people's minds leading up the the event. Or whatever.
Only if the masters of such a universe made a mistake or intended this result.
If you are simulating a universe you can make as many measurements as you want, at whatever simulated-time-intervals that you want.
For example, suppose I make a simulator that simulates a perfect electronic oscillator (no noise, etc.). I can have it generate a sine wave at a frequency of 1E9999 Hz with perfect accuracy. I can sample that waveform at a rate of 1E999999999999 Hz. I can do all this on an 8088 (assuming that it had enough storage to hold the desired number of samples).
How can I do this? Simple - I never said I could do it in realtime. I could run my simulation over 100 simulated hours with perfect accuracy (well, within arbitrary precision, at least). Sure, it might take me 100,000 years to run the simulation, but I'd get the data and it wouldn't perturb the simulation at all.
Likewise - if our universe is simulated there is no reason to believe that it is simulated in realtime. It is possible that 1000 years in the parent universe pass in the time it takes for one second to pass in our "simulated" universe. If they need to collect more data they can just slow down the simulation accordingly, and not sacrifice accuracy. If you were running a simulation of a scale the size of our universe and the apparent precision on the order of a plank length/time, then why would you mess up the laws of physics your simulation simply to collect a little more data for a short interval? The "butterfly effect" would make the long-term results of your simulation worthless after you've done this even once.
If somebody were taking care to simulate an entire universe you'd have to assume that there would only be apparent errors in the simulation if the simulation designer intended them to be there...
And this is the fault. Randi does not have a set guideline, nor does a disinterested party have control of either the funds or the testing procedures.
How can you have a set guideline for:
1. Proof of the existance of telepathy.
2. Proof of the ability to levitate objects.
3. Proof of the ability to predict that something "bad" will happen "soon".
4. etc...
His system is that you tell him what you claim to be able to do, and then you agree with him on what constitutes the guidelines. It is hard to have a "standard" set of guidelines for a huge variety of potential claims.
If a psychic is upset they should simply publish what the unfair proposed guidelines were, and let people draw their own conclusions...
That's one way to do it. Or, if you don't like wasting time, just go with the straight microsoft ticket. Install all the software, run windows update once (ok probably twice as you need to install some things sepperatetly) and you are on your way.
Uh, windowsupdate only updates windows, and a few MS programs.
If I have a Quicken security problem, windowsupdate won't fix it.
If kmymoney has a security problem, Red Hat will distribute a patch for it.
So, we're comparing the number of patches to the MS Windows OS, to the number of patches for just about every free piece of software available for linux...
Makes sense - I was unaware that one of those cables was not grounded...
World of Warcraft PvP server players have started compiling lists of known farmers working in shifts and are going around pounding them flat whenever they can (more power to em).
Probably a losing battle unless it is done by the server admins (ie they just massacre them by the legion in a puff of logic).
Otherwise it just raises the cost of doing business. You'll have a field of farmers flanked by mercenaries armed to the teeth. If they can pay farmers, then they can pay guards...
If you can do this, you guys can all come over to my place for a front seat view of the resulting fireworks.
I think I'll take a back seat. The asteroid impact calculator says that it will have an impact energy of 594 megatons, which is probably something I'd try to stay far away from.
It will, however, be quite effective at getting rid of that Best Buy.
Money does matter to the kin and to the dying, who care about their families.
This is of course a very controversial topic.
My observation is that most people put a dollar figure on life. However, anybody who openly talks about this is looked down upon. It is acceptable to be a bean counter, it just isn't acceptable to actually say that you are one, or that bean counting was the reason for a decision you made.
Kind of like a company I know somebody at. They were doing some testing to marginally improve the safety of their products and to better comply with industry-wide regulations. Two managers got in a discussion, and one brought up cost-benefit analysis. The superior chastised him - safety should come first no matter what the cost. And yet, the project had a budget, and if any body suggested increasing it tenfold they'd be fired. So, there was a cost-benefit analysis, it just was never discussed or analyzed to make sure it was done right.
The problem with not talking about the costs of saving lives is that we end up misallocating resources. We'll spend billions of dollars on medicare so that old people live a few months longer on hospital beds, but we won't spend any money on decent foster homes, so that abused children don't have to live with their abusers. Which is the bigger health problem - people who are going to die in a year no matter what, or people who have a whole life ahead of them who could be destined for greatness or the ghetto? And which costs more, trying to push back death two weeks for somebody diagnosed with a terminal disease, or creating some incentives for half-decent families to adopt kids who would otherwise be abandoned to abuse?
Don't get me wrong - I'm all for reasearch to improve the quality and length of life. However, cutting-edge treatments simply aren't for everyone. We should be researching how cure difficult diseases, and we should let the rich pay for cutting-edge treatment if they so desire. As more doctors become educated in these treatments they will become mainstream, and available to everyone (possibly with government assistance).
There is a price in human life even when we try to save lives - if we don't count the costs we may harm more than we help...
Uh, Merck, despite losing probably 20+% of its revenue over the last six months is still running a profit. It may go through some rough times with upcoming patent expirations, but it is still expected to make it through reasonably intact. (It just doesn't have the world's greatest stock growth potential right now.)
On the other hand, the US Federal government has been losing money for decades, even though it can raise prices (taxes) anytime it feels like it...
Most companies are reasonably thrifty. Ones that are having financial issues (like Merck) tend to be even moreso.
After all, those photons are moving (oddly enough) at the speed of light
I think he was talking about the neutrinos - which are moving very close to the speed of light (from our frame of reference).
In fact - they're moving so close to light speed that I don't think anybody has come up with an experiment that can measure how much slower than c they are actually travelling. After all, if people knew they weren't quite at light speed it wouldn't have been such a big discovery when it was found that they have mass. The only way this was discovered was because of oscillations, which imply mass. However, it does not imply the amount of mass, which is already known to be extrememly small (just not how small).
So, my guess is that VERY little time passes in the neutrino's frame before it hits earth. It probably is a VERY small fraction of a second.
I think his point is that:
1. Yes - if your filesystem code crashes, you could end up with a dirty filesystem.
2. Yes - if your hard drive code crashes, you could end up with a dirty hard drive.
But:
3. No - if your webcam driver crashes, you won't end up with a dirty hard drive.
Right now with linux, if a kernel-level driver of any kind panics, the whole thing goes down the tubes.
Certainly a little compartmentalization can't possibly hurt. It won't fix every problem, but it does prevent a small problem in a non-essential driver from taking down the whole system.
As you point out, it will still be critical for some pieces of code to just work without bugs at all. However, the amount of that code can be reduced in a microkernel design.
Also - I don't think TWAIN is windows-specific. I seem to recall using TWAIN on a Mac many a year ago...
Agreed - however that is purely a matter of getting the browser maintainers to add their cert to the list of trusted root CAs.
For MS that might take some time. I would have hoped that Firefox would be more enlightened, but they've been talking about it for 1.5 years now without actually doing much about it...
Well, we just wasted the last 12 months not doing anything about it.
Really - if we had just planned ahead, we'd have HOP up and running already...
I would suspect designing and building an orbital telescope to equal Hubble would cost many billions of dollars.
Why design anything?
Why not just build a new one just like the existing Hubble? Maybe upgrade a few parts where it is cheap to do so, but nothing that would require extensive re-integration.
Rather than designing fancy super-upgradable telescopes why not just make one that does a single job, launch it, and then start on the next one. If one fails, just send up an EXACT copy. It can't cost nearly as much just to rebuild the same thing, can it?
Uh, that sounds awfully high.
Naturally fermented wines don't go about about 12% alcohol content since that is bacteriocidal.
If your fluids were more than 8% alcohol there would be no need to enbalm drunks who died in accidents...
Name one of the modern clients that won't do it.
kontact/kmail will only do it with gnupg 1.9.x. This is considered alpha code, and this really isn't the kind of application you want to use alpha code for...
When I went to grad school the first thing I did was enable my remote shell accounts to use s/key, and I printed out 100 passwords and kept them on me.
Universities tend to be notorious for that sort of thing. My main concern was actually sniffing, though.
Speaker cables are unbalanced. There is also no such thing as "send and receive" on speaker cables.
It is true that speaker cables are generally not prone to interference, but that is because there is no signal amplification once they get to their destination.
Illustration:
Your signal comes in from the DVD player - a few mW in total. If there is a lot of noise nearby, it could be noticable, since noise is probably in the uW-mW range. Your amp then boosts whatever input it gets to 1000W and sends it to the speakers. Now if you add a few uW of noise it isn't noticable at all, but any noise that came in via the DVD input got boosted by the amp and is quite noticable.
A balanced signal is one which is sent via twisted pair where each wire in the pair is 180 degrees out of phase with respect to the other, and where the sum of the two signals is ground. This is noise resistant, since incoming noise will be in-phase in each cable, and when the signals are inverted and combined the noise will cancel out.
Unbalanced signals are still differential - you can just take an unbalanced microphone signal and put it into a speaker and you will get sound (albeit virtually inaudible). In fact, a speaker will function as a very poor microphone - if you put two speakers back-to-back on the same line, if you speak into one it vibrates the diaphram, and induces current in the wire, which then travels to the other speaker and causes its diaphram to move.
Where you do benefit in running balanced signals to speakers is when the speakers are powered - since you're now applying gain to the signal.
I'm pretty sure that just about every balanced system out there uses at least three wires. Ethernet uses only 2 for each way (for 10BaseT), but that's probably because the plug in the wall provides a ground reference. Is anyone aware of a battery-operated device that successfully uses an ethernet interface with no other cables attached?
Disclaimer - I'm not an EE. And if somebody who really is one cares to illuminate this discussion, please feel free to do so. However, kindly provide more than "you're wrong"...
Suppose your two 44.1 kHz reference clocks on two soundcards are off by 10 Hz. That is 10 samples per second of drift. For a 5-minute song that is only 3000 samples drift, or 68ms (maybe noticable).
For a 20-minute movie soundtrack or classical work you now have 12000 samples drift, which is over 1/4th of a second. You don't need golden ears to hear that - it will sound like an echo coming from the other room.
If you're playing the sound from a movie you're watching on TV, the final lines of the 120 minute movie will be delayed by 1.6 seconds in the other room. That will be VERY noticable.
Scale all times in accordance with actual sound card frequency ref error. If it is only 1 Hz, then the movie will still be noticable at 160ms, but the rest should be fine.
With multicast you should be better - just the 1ms of delay associated with network latency, which you won't hear. Occassionally a sound card might cut out for 1/10th of a second if it gets too far ahead of the data stream, and then it will sync up with the stream rate. Sound cards slower than the stream rate will continue to fall farther and farther behind, however, as long as they are caching the network data.
Of course, if you multicast the whole stream at network speeds, then everything will get sent in 25 seconds, and everything will go back out of sync again...
However, you also benefit from much-more-tailored build processes. I'm not just talking tailored to your CPU - but also to your preferences.
Do you prefer KDE or GNOME? You can tell gentoo your preference and it will build apps accordingly, not compiling in needless support for the other, and consequently eliminating depedencies that you don't care for.
That and the package management system works rather well - you could get many benefits from using gentoo and only installing binary packages...
When you're getting down to the single nm range, we're no longer talking UV-lithography - you need X-rays. It probably will also get harder to come up with photosensitive dyes that respond to X-rays...
Sure, and many of us know how to go about recovering hard drives, etc. Still, we all run backups and RAID since it is much easier just to hot-swap a drive, or restore a backup.
I think his point is that while all admins should understand the inner-workings of their systems, they shouldn't expect to have to dive into them day-to-day on every box they manage...