It would appear that we have reached the limits of
what it is possible to achieve with computer technology,
although one should be careful with such statements,
as they tend to sound pretty silly in 5 years.
For the record: I have produced this quote around 20 years ago when similar statements about the "end of moore within 5-10 years" were made
You are entirely correct that "audit at the interface" is a good idea. The problem here is that the interface is hidden inside the CPU. And the "interface" is undisclosed. What if there is a broadcast packet that, when the PC is off, puts the ME into slave mode? Yea, come out of powerdown, do NOT enable video (keep the monitor blanked), send the harddisk contents to NSA... You can "see it happening" at the network interface (third meaning of that word) when it happens, but there might be a cryptographically secure way of preventing you from finding/probing that packet. In this case I mean with cryptographically secure that it is unfeasible to find it by brute force. Not that you can't see it come by on the net if it happens while you're looking. For such a feature: "all ME's report NOW" you can't have individually encrypted codes. So it would work for every one, e.g. a statiic say 256bit random password would be an example.
Anyway, This ME is embedded deep enough that it is difficult to probe at the interfaces. (you can't cut the trace that connects it to the rest of the system so that you can be sure it is turned off), and it has enough access to be able to do serious harm....
For example: Say the NSA secretly downloads your whole harddisk. But... you say: it's encrypted, don't worry! The ME has enough access that, when under control of "bad guys", it can halt your CPU just as it has obtained the key to your harddrive. Next time you boot the PC a mysterious packet flies off to some remote IP address...
> Both are illegal and they will commit double offense Ehh. Yeah, so what is different from the older tactic of promising to pop someones kneecaps? I'm told this is illegal too.
Shady lenders use shady tactics to force their clients to repay them. That's how it works. Now they have moved to the "internet" and "I know where to find you! (silently promises to pop kneecaps) " is no longer a threat to someone "far away" online, possibly through tor or whatever.
I find the new strategy less barbaric than the old one actually....
There are about 2 million sixteen year old boys in the USA (alone). Of these a bunch are interested in computers. Just because "that's a large enough group", I'm ignoring the 15 year olds, 17 year olds and the girls.
And one day, one of them will spot a uid=1234 in the URL and try what happens if you change that into uid=1235. According to current laws that is considered hacking, and the culprit needs to go to jail. And you're going to predict which one of the two hundred thousand computer-interested sixteen year olds is going to do that? Good luck!
Here in Holland a some students noted that if they ordered pizza from a certain shop, they got sent to a page: "You owe us $15.60, how are you going to pay?". And the URL clearly had that 15.60 visible. So they decided to change that to "0.10". So then the page said: "You owe us $0.10, how are you going to pay?". So they chose a payment method, paid $0.10 and.... they got redirected to the pizza-site where it said: Thank you for your payment, your pizza is on its way!
In the case of the free pizzas, the company who created that stupid "don't check the amount" code should be liable. Checking that the right amount was paid is elementary to a payment system. Similarly not only checking that a user is logged in, but also checking that he/she is logged in as the RIGHT user is elementary.
You cannot blame the guy who stumbled upon this issue for "hacking". Sure, getting almost-free pizzas for a year is a bit unethical. It would be nice to inform the maintainers of the issue, but since when is being "not nice" going to land you in jail? Well, I'll tell you: since they adopted those anti-hacking laws. And for those, it doesn't matter if you're nice. If you ARE nice and report it, they can (and often do) throw you in jail anyway.
Getting used to an SCS and getting better at using it efficiently is a hassle. So at one point in time, I decided to "learn" bitkeeper and use it.
After I had invested some time and effort in the suite, suddenly the licence is changed and it becomes incompatible with my use of the software.
Do you think I'm going to switch back after 11 years? Nah! Not me.
And actually git is a lot better. Linus wrote git to be good at a few things that are important to him and "his" little project (that isn't very little). Linus' intuition about what's important is good. That's what makes Linux and Git succesful.
Here in the Netherlands, there are a bunch of people who are on welfare. Or social security, or whatever you may call it. They are in the situation that when they start to work, they will lose their right to that. So the first $1500 they earn each month, they get to keep... almost nothing. Now you wonder why they hang around and do nothing? If, as a society, we provide everyone with a basic income. Disabled, without a job due to no fault of the person him/herself, too old to work or just plain lazy. Then for every dollar you earn, you get to keep most of it. No reason to cheat by working illegally, not a huge discontinuity when going from "can't work due to medial reasons" to "might be able to work part-time in another line of work".
Mind you.... The $2500 is way too much (at least here). The basic income should be enough that you can live in basic circumstances, but without many luxuries. If you want a car, work for it. If you want a vacation, work for it.
You're going on vacation or business trip and call a cab to go to the airport. You put your luggage in the trunk, have a pleasant conversation on the way to the airport and after you pay the driver, he says: Sorry sir, but the trunk-security-system detected [something dangerous that is NOT in your luggage], your luggage is being safely destroyed, have a nice day.
You agreed to the "terms and conditions" of the cab company that they may scan your luggage and proceed as appropriate when they detect something dangerous. That seems entirely plausible and acceptable, until it hits a false positive and they are destroying YOUR property.
Somehow, it seems that SOME people feel "safe" that something is being done. But taking people's property because of a perceived threat is wrong. It would be acceptable if they say: "Your transfer triggered a double check, it will go through with a couple of days delay". If someone really thinks this is a serious threat, then go ahead and arrest the guy.
This is similar to the discrimination situation. It's all fun-and-games until it happens to you and they pick on you.
Back in the old days, you needed to put information on dead trees and transport those dead trees to the people. Companies were willing to do that, for a fee.
Some research-fields are small, so doing a magazine-run (and editing) for a handful of people costs a lot of money. So the subscription fee is (sometimes) high.
Nowadays, the distribution need not cost much.
But the "editing" and "quality control" are parts that are still difficult in the internet-age.
Of course a prime ending in 9 will most often be followed by a prime ending in 1.
Primes are more-or-less "random": you can't easily predict the next prime. In every number-range primes have a certain density, and that density drops as numbers get larger. So "around 1000", the prime-density is higher than "around 10000".
So assuming the prime density is P and we have a prime p ending in nine, there is a P chance of p+2 (ending in 1) being prime. Then there (1-P)P chance of the next candidate being prime. This is a smaller number. The next number CANNOT be prime because it is divisible by five. Anyway, the chance of the next prime being p+10 is on the order of (1-P)^4.P.
Because the prime density is not all that low, even for numbers around a million, the fourth power of 1-P becomes small quickly.
Note that I'm using an adjusted prime density here. If there are x primes per stretch of 100 numbers on the number line, you have a prime density of x/100. The P I'm dealing with is only considering the 4-mod 10 options for primes. So P = 10/4. x/100 .
My mom has a vacation home and is retired.... She lives in one place for a few months and then in the other for a few months. So when she asks me "I need a new printer, what do you recommend?" my recommendation is: "Whatever you buy DO NOT, I repeat, DO NOT buy an inkjet". Guess what happened next?
Of course there is a waste ink part that needs emptying. So: "Waste ink receptacle full" is a reasonable error message. Designing it in such a way that it is (with some trouble) exchangeable should be quite possible without increasing cost.
But "having" that counter, the incentive is for the manufacturer to take big margins on when to call it "full".
My ISP for my server had an outage for a couple of hours. They then mailed their customer base that they'd refund a whole month of service. I'm guessing that mail went out to more than the affected customers: My server didn't go down. A couple of days later I got the invoice for that month. Full charge. Two days later: full credit. I'm guessing they did the right thing and credited everybody who got the mail, even if they sent it out to customers that weren't affected.
Wait.... let me get this straight.... Their BUSINESS service is down, "DOZENS of times", for over HALF AN HOUR, and then they give you a partial refund?
I get way better SERVICE (i.e. it works) on my residential account.
> When was the last time a telecom provider gave you a discount on your bill without you asking for it?
My cellphone was down for a day this summer. They apologized and offered a "free 500Mb of internet" for the September billing cycle.
I buy "internet service" on my phone with lots of margin. They charge like 100 times more for the "over your limit" traffic, so I rarely use the upper 500Mb of my "prepaid" 1G limit. It's cheaper to pay 12 months for 1Gb/month than to pay for 500Mb and go over the limit once a year.
So their "free 500Mb of internet traffic" is indeed completely free... for them.
Not just 97% of those with brain swelling dies, but 100%! Not only those with brain swelling die, but 100% of those exposed die! Even 100% of those NOT exposed die. Eventually.
The "1/hr" argument is in my "quantization". The larger the quantization, the larger the margins have to be to make sure you catch the bus. When I have an appointment in Rotterdam, I take the train to Schiedam and then the underground to the center. When I try to be at the station 5 minutes before the train leaves, two minutes extra "margin" make me catch the train earlier. I've given up on schedules and just go there and take the first train. When a train or bus goes once every 15 minutes you get there with 3-5 minutes to spare. If it goes every hour, you need about 5-10 minutes, and if it goes once a day you need to aim for being there 15 minutes before it leaves, because getting stuck in traffic for 5-10 minutes should be tolerable....
Self driving cars would still be an option in the country as long as you tolerate having a slightly longer lead time.
What is it that notorious car-drivers hate about public transport, so that they end up spending like 5-10 times more money than what public transport would have cost them?
That is the question that you need to answer before you can make predictions like in the slashdot summary.
Public transport has "quantization noise". If you leave for work 3 minutes later, you'll miss the bus, and end up at work 15 minutes later. Sure you can prevent that. Just be at the bus stop 5 minutes before the bus. This means you invest 5 minutes every day to prevent a 15 minute occasional delay.
Second: when you use public transport, you don't have "your own space". This means that if you leave your wallet, ipad or whatever lying around when you step out of the vehicle, barring exceptions, it is gone. If you have your own "transport space", you can also stock it with stuff you might need while travelling (e.g. my brother has mints in his car), of that you might end up needing at the destination (e.g. umbrella).
Third there is a cost issue. If you see the cost to you every time you move, that makes you consider it more. People ignore the: "the car needs xx petrol to drive this far, so it costs me at the least yy to make this trip". They see the filing up and paying whatever that takes as something that must be done, and then they pretend driving the car is free.
Those are the things the car-drivers will need to have satisfied before they can be converted. And if you don't convert the car-drivers you will not have the economy of scale to have "leaving out the steering column" make a difference.
The current topic will "solve" the quantization noise. You hit a button in your house a few minutes before you leave, and the system will get a self-drving-car on your driveway before you close the door. It does not solve the other problem. You cannot leave the umbrella in the car "in case it rains when I get there". You forget your stuff and the next occupant might take it. (or at least "where is my presentation" is not solved with a run to the car park).
The "self driving" and "semi-public-transport" ideas will only work if everybody gets to keep their private module. Those could be powered with say a small 1kW motor and have a 40km/h (25mph) speed limit. Then when they end up at the freeway, a bunch of them group together with a "power-unit" and they can travel on the freeway at high speed. (Pay more and the "maximum grouping delay" goes down).
The article claims that NTP is the cause of the leap second. NTP is just a protocol that handles keeping computer clocks in sync with each other and with the official time (UTC, IIRC).
If NTP handles leap seconds by increasing update frequency and then coming to the conclusion: "Whoa! my offset just went from 0.3 ms to 1000.25 ms, lets step the clock a second once we're sure this was not a fluke measurement". then that's a bad way of handling it in my opinion. (also suddenly speeding up does not provide a smooth-enough transition).
One of the things that is bad about this is that when normal operation can handle (the bandwidth of) most hosts updating every 1024 seconds, and a few hosts (just rebooted, just installed, sync lost, whatever), now all of a sudden a synchronized (pun intended!) attack will take place where many many hosts will increase their update frequency by several order-2 magnitudes.
For google, they internally have needs for synchronized clocks. Why I don't know, and I don't care. They have decided to handle the leap second in a more controlled way. It's actually not that hard. Just make sure that everything syncs off one level-0 server, and during the 20 hour period leading up to the leap second, add a variable number of microseconds to the exported time.
The thing that people-who-don't-know-better are suggesting is that the second will be the same all the time.
They think that nothing bad will come from "thirty years from now, the sun is in the south at 11:59:30" (assuming an average of 1 leap-second per year).
(I can't think of anything bad that would happen... but I know my limitations. It's probably annoying as hell to/some/ fields of research or something....)
Suppose I have a variable A that ends up with the value Y*X. (Y might be a difficult calculation). Next I want to calculate B = A/X .
This could happen for example when I'm doing physics calculations where the parameter X eventually cancels out.
Anyway, this will end up with B = Y if you do the math and cancel out the X. However, if you let a computer follow through with the calculations, when X=0, you'll end up with a variable A with the value zero. And if you assign a value of "1" to 0/0 you'll silently get a wrong result when Y != 1.
So: The computer should throw an error. There is no way a compiler can come up with a reasonable answer for the variable B.
If you want it your way, you write B = X?A/X:1; If you want get rid of that expression everywhere in your code, you get a few choices. In C++ you could probably define a "myfloat" that overloads the division operator. Or you could make a "mydiv" function.
Directional is an option for the satellites. But on the ground you'd have to be tracking all the time, and you'd have a dropout the moment one satellite goes away and another comes into view.
Oh, about the height. Suppose you're at 1000km. Then the area that can see the satellite at at least 45 degrees above the horizon is about 1000km in diameter. This covers an area of about 3 million square km. The earth is about 450 million square km. You'd need about 300 satellites to cover the earth with each spot getting on average two satellites (hopefully you can then arrange for every spot to get at least one satellite). If they need 4000 satellites they are apparently aiming for about 300km height! -> my calculation comes to 3000 satellites, they probably have a few in reserve, require a higher number of satellites to cover everywhere all the time etc.
Anyway, that would come to "too low" for the drag/orbit deay reasons..... So that would mean they are aiming for "the lowest the orbit-decay will alow".
Rereading my own message: Near: "You want them as close as possible." I forgot to mention: "because the required power to transmit to the satellite increases with the distance."
Older satellite internet systems used technologies borrowed from "TV broadcasting". What they effectively did is broadcast everybody's downlink via the satellite and everybody-for-himself had to use a land-line for the uplink. The idea being that you like having a big downlink pipe and it might be acceptable to have "only" 56k uplink.
All that is going overboard, as I understand things. Yes, people are going to transmit their uplink bits to the satellites: the stated goal of these projects is "internet everywhere", even where landlines don't come. So on most my calculations (I hope all of them:-) ) I calculated the round-trip delay via satellite.
On the other hand, I did not take the groundstation-to-destination delays into account. Those are on the order of 60ms minimum for a transatlantic link, for exactly the same reason that a transatlantic satellite link will take that amount of time: the light speed. (I just tried tracing packets to three American companies from Europe. All were reachable in less than 10ms (not enough to cross the ocean), with google performing worst from the three I tried: google, nbc, cnn. Apparently they all have servers serving european users here in europe).
Slashdot Mirror
.... John von Neumann said..... In 1947.
http://www.brainyquote.com/quo...
It would appear that we have reached the limits of
what it is possible to achieve with computer technology,
although one should be careful with such statements,
as they tend to sound pretty silly in 5 years.
For the record: I have produced this quote around 20 years ago when similar statements about the "end of moore within 5-10 years" were made
That should prevent forest fires for a long time, right?
You are entirely correct that "audit at the interface" is a good idea. The problem here is that the interface is hidden inside the CPU. And the "interface" is undisclosed. What if there is a broadcast packet that, when the PC is off, puts the ME into slave mode? Yea, come out of powerdown, do NOT enable video (keep the monitor blanked), send the harddisk contents to NSA...
You can "see it happening" at the network interface (third meaning of that word) when it happens, but there might be a cryptographically secure way of preventing you from finding/probing that packet. In this case I mean with cryptographically secure that it is unfeasible to find it by brute force. Not that you can't see it come by on the net if it happens while you're looking. For such a feature: "all ME's report NOW" you can't have individually encrypted codes. So it would work for every one, e.g. a statiic say 256bit random password would be an example.
Anyway, This ME is embedded deep enough that it is difficult to probe at the interfaces. (you can't cut the trace that connects it to the rest of the system so that you can be sure it is turned off), and it has enough access to be able to do serious harm....
For example: Say the NSA secretly downloads your whole harddisk. But... you say: it's encrypted, don't worry! The ME has enough access that, when under control of "bad guys", it can halt your CPU just as it has obtained the key to your harddrive. Next time you boot the PC a mysterious packet flies off to some remote IP address...
> Both are illegal and they will commit double offense
Ehh. Yeah, so what is different from the older tactic of promising to pop someones kneecaps? I'm told this is illegal too.
Shady lenders use shady tactics to force their clients to repay them. That's how it works. Now they have moved to the "internet" and "I know where to find you! (silently promises to pop kneecaps) " is no longer a threat to someone "far away" online, possibly through tor or whatever.
I find the new strategy less barbaric than the old one actually....
There are about 2 million sixteen year old boys in the USA (alone). Of these a bunch are interested in computers. Just because "that's a large enough group", I'm ignoring the 15 year olds, 17 year olds and the girls.
And one day, one of them will spot a uid=1234 in the URL and try what happens if you change that into uid=1235. According to current laws that is considered hacking, and the culprit needs to go to jail. And you're going to predict which one of the two hundred thousand computer-interested sixteen year olds is going to do that? Good luck!
Here in Holland a some students noted that if they ordered pizza from a certain shop, they got sent to a page: "You owe us $15.60, how are you going to pay?". And the URL clearly had that 15.60 visible. So they decided to change that to "0.10". So then the page said: "You owe us $0.10, how are you going to pay?". So they chose a payment method, paid $0.10 and.... they got redirected to the pizza-site where it said: Thank you for your payment, your pizza is on its way!
In the case of the free pizzas, the company who created that stupid "don't check the amount" code should be liable. Checking that the right amount was paid is elementary to a payment system. Similarly not only checking that a user is logged in, but also checking that he/she is logged in as the RIGHT user is elementary.
You cannot blame the guy who stumbled upon this issue for "hacking". Sure, getting almost-free pizzas for a year is a bit unethical. It would be nice to inform the maintainers of the issue, but since when is being "not nice" going to land you in jail? Well, I'll tell you: since they adopted those anti-hacking laws. And for those, it doesn't matter if you're nice. If you ARE nice and report it, they can (and often do) throw you in jail anyway.
Getting used to an SCS and getting better at using it efficiently is a hassle. So at one point in time, I decided to "learn" bitkeeper and use it. After I had invested some time and effort in the suite, suddenly the licence is changed and it becomes incompatible with my use of the software. Do you think I'm going to switch back after 11 years? Nah! Not me. And actually git is a lot better. Linus wrote git to be good at a few things that are important to him and "his" little project (that isn't very little). Linus' intuition about what's important is good. That's what makes Linux and Git succesful.
Here in the Netherlands, there are a bunch of people who are on welfare. Or social security, or whatever you may call it. They are in the situation that when they start to work, they will lose their right to that. So the first $1500 they earn each month, they get to keep... almost nothing. Now you wonder why they hang around and do nothing?
If, as a society, we provide everyone with a basic income. Disabled, without a job due to no fault of the person him/herself, too old to work or just plain lazy. Then for every dollar you earn, you get to keep most of it. No reason to cheat by working illegally, not a huge discontinuity when going from "can't work due to medial reasons" to "might be able to work part-time in another line of work".
Mind you.... The $2500 is way too much (at least here). The basic income should be enough that you can live in basic circumstances, but without many luxuries. If you want a car, work for it. If you want a vacation, work for it.
So let me make up an example story.
You're going on vacation or business trip and call a cab to go to the airport. You put your luggage in the trunk, have a pleasant conversation on the way to the airport and after you pay the driver, he says: Sorry sir, but the trunk-security-system detected [something dangerous that is NOT in your luggage], your luggage is being safely destroyed, have a nice day.
You agreed to the "terms and conditions" of the cab company that they may scan your luggage and proceed as appropriate when they detect something dangerous. That seems entirely plausible and acceptable, until it hits a false positive and they are destroying YOUR property.
Somehow, it seems that SOME people feel "safe" that something is being done. But taking people's property because of a perceived threat is wrong. It would be acceptable if they say: "Your transfer triggered a double check, it will go through with a couple of days delay". If someone really thinks this is a serious threat, then go ahead and arrest the guy.
This is similar to the discrimination situation. It's all fun-and-games until it happens to you and they pick on you.
Back in the old days, you needed to put information on dead trees and transport those dead trees to the people. Companies were willing to do that, for a fee.
Some research-fields are small, so doing a magazine-run (and editing) for a handful of people costs a lot of money. So the subscription fee is (sometimes) high.
Nowadays, the distribution need not cost much.
But the "editing" and "quality control" are parts that are still difficult in the internet-age.
Of course a prime ending in 9 will most often be followed by a prime ending in 1.
Primes are more-or-less "random": you can't easily predict the next prime. In every number-range primes have a certain density, and that density drops as numbers get larger. So "around 1000", the prime-density is higher than "around 10000".
So assuming the prime density is P and we have a prime p ending in nine, there is a P chance of p+2 (ending in 1) being prime. Then there (1-P)P chance of the next candidate being prime. This is a smaller number. The next number CANNOT be prime because it is divisible by five. Anyway, the chance of the next prime being p+10 is on the order of (1-P)^4.P.
Because the prime density is not all that low, even for numbers around a million, the fourth power of 1-P becomes small quickly.
Note that I'm using an adjusted prime density here. If there are x primes per stretch of 100 numbers on the number line, you have a prime density of x/100. The P I'm dealing with is only considering the 4-mod 10 options for primes. So P = 10/4. x/100 .
I wrote an elaborate comment the day before yesterday. I had to leave and when I came back to it yesterday, I finished it and hit submit. Poof! gone!
My mom has a vacation home and is retired.... She lives in one place for a few months and then in the other for a few months. So when she asks me "I need a new printer, what do you recommend?" my recommendation is: "Whatever you buy DO NOT, I repeat, DO NOT buy an inkjet". Guess what happened next?
Of course there is a waste ink part that needs emptying.
So: "Waste ink receptacle full" is a reasonable error message. Designing it in such a way that it is (with some trouble) exchangeable should be quite possible without increasing cost.
But "having" that counter, the incentive is for the manufacturer to take big margins on when to call it "full".
My ISP for my server had an outage for a couple of hours. They then mailed their customer base that they'd refund a whole month of service. I'm guessing that mail went out to more than the affected customers: My server didn't go down. A couple of days later I got the invoice for that month. Full charge. Two days later: full credit. I'm guessing they did the right thing and credited everybody who got the mail, even if they sent it out to customers that weren't affected.
Wait.... let me get this straight.... Their BUSINESS service is down, "DOZENS of times", for over HALF AN HOUR, and then they give you a partial refund?
I get way better SERVICE (i.e. it works) on my residential account.
> When was the last time a telecom provider gave you a discount on your bill without you asking for it?
My cellphone was down for a day this summer. They apologized and offered a "free 500Mb of internet" for the September billing cycle.
I buy "internet service" on my phone with lots of margin. They charge like 100 times more for the "over your limit" traffic, so I rarely use the upper 500Mb of my "prepaid" 1G limit. It's cheaper to pay 12 months for 1Gb/month than to pay for 500Mb and go over the limit once a year.
So their "free 500Mb of internet traffic" is indeed completely free... for them.
Not just 97% of those with brain swelling dies, but 100%!
Not only those with brain swelling die, but 100% of those exposed die!
Even 100% of those NOT exposed die. Eventually.
The "1/hr" argument is in my "quantization". The larger the quantization, the larger the margins have to be to make sure you catch the bus. When I have an appointment in Rotterdam, I take the train to Schiedam and then the underground to the center. When I try to be at the station 5 minutes before the train leaves, two minutes extra "margin" make me catch the train earlier. I've given up on schedules and just go there and take the first train. When a train or bus goes once every 15 minutes you get there with 3-5 minutes to spare. If it goes every hour, you need about 5-10 minutes, and if it goes once a day you need to aim for being there 15 minutes before it leaves, because getting stuck in traffic for 5-10 minutes should be tolerable....
Self driving cars would still be an option in the country as long as you tolerate having a slightly longer lead time.
What is it that notorious car-drivers hate about public transport, so that they end up spending like 5-10 times more money than what public transport would have cost them?
That is the question that you need to answer before you can make predictions like in the slashdot summary.
Public transport has "quantization noise". If you leave for work 3 minutes later, you'll miss the bus, and end up at work 15 minutes later. Sure you can prevent that. Just be at the bus stop 5 minutes before the bus. This means you invest 5 minutes every day to prevent a 15 minute occasional delay.
Second: when you use public transport, you don't have "your own space". This means that if you leave your wallet, ipad or whatever lying around when you step out of the vehicle, barring exceptions, it is gone. If you have your own "transport space", you can also stock it with stuff you might need while travelling (e.g. my brother has mints in his car), of that you might end up needing at the destination (e.g. umbrella).
Third there is a cost issue. If you see the cost to you every time you move, that makes you consider it more. People ignore the: "the car needs xx petrol to drive this far, so it costs me at the least yy to make this trip". They see the filing up and paying whatever that takes as something that must be done, and then they pretend driving the car is free.
Those are the things the car-drivers will need to have satisfied before they can be converted. And if you don't convert the car-drivers you will not have the economy of scale to have "leaving out the steering column" make a difference.
The current topic will "solve" the quantization noise. You hit a button in your house a few minutes before you leave, and the system will get a self-drving-car on your driveway before you close the door. It does not solve the other problem. You cannot leave the umbrella in the car "in case it rains when I get there". You forget your stuff and the next occupant might take it. (or at least "where is my presentation" is not solved with a run to the car park).
The "self driving" and "semi-public-transport" ideas will only work if everybody gets to keep their private module. Those could be powered with say a small 1kW motor and have a 40km/h (25mph) speed limit. Then when they end up at the freeway, a bunch of them group together with a "power-unit" and they can travel on the freeway at high speed. (Pay more and the "maximum grouping delay" goes down).
The article claims that NTP is the cause of the leap second. NTP is just a protocol that handles keeping computer clocks in sync with each other and with the official time (UTC, IIRC).
If NTP handles leap seconds by increasing update frequency and then coming to the conclusion: "Whoa! my offset just went from 0.3 ms to 1000.25 ms, lets step the clock a second once we're sure this was not a fluke measurement". then that's a bad way of handling it in my opinion. (also suddenly speeding up does not provide a smooth-enough transition).
One of the things that is bad about this is that when normal operation can handle (the bandwidth of) most hosts updating every 1024 seconds, and a few hosts (just rebooted, just installed, sync lost, whatever), now all of a sudden a synchronized (pun intended!) attack will take place where many many hosts will increase their update frequency by several order-2 magnitudes.
For google, they internally have needs for synchronized clocks. Why I don't know, and I don't care. They have decided to handle the leap second in a more controlled way. It's actually not that hard. Just make sure that everything syncs off one level-0 server, and during the 20 hour period leading up to the leap second, add a variable number of microseconds to the exported time.
The thing that people-who-don't-know-better are suggesting is that the second will be the same all the time.
They think that nothing bad will come from "thirty years from now, the sun is in the south at 11:59:30" (assuming an average of 1 leap-second per year).
(I can't think of anything bad that would happen... but I know my limitations. It's probably annoying as hell to /some/ fields of research or something....)
Suppose I have a variable A that ends up with the value Y*X. (Y might be a difficult calculation). Next I want to calculate B = A/X .
This could happen for example when I'm doing physics calculations where the parameter X eventually cancels out.
Anyway, this will end up with B = Y if you do the math and cancel out the X. However, if you let a computer follow through with the calculations, when X=0, you'll end up with a variable A with the value zero. And if you assign a value of "1" to 0/0 you'll silently get a wrong result when Y != 1.
So: The computer should throw an error. There is no way a compiler can come up with a reasonable answer for the variable B.
If you want it your way, you write B = X?A/X:1; If you want get rid of that expression everywhere in your code, you get a few choices. In C++ you could probably define a "myfloat" that overloads the division operator. Or you could make a "mydiv" function.
Directional is an option for the satellites. But on the ground you'd have to be tracking all the time, and you'd have a dropout the moment one satellite goes away and another comes into view.
Oh, about the height. Suppose you're at 1000km. Then the area that can see the satellite at at least 45 degrees above the horizon is about 1000km in diameter. This covers an area of about 3 million square km. The earth is about 450 million square km. You'd need about 300 satellites to cover the earth with each spot getting on average two satellites (hopefully you can then arrange for every spot to get at least one satellite). If they need 4000 satellites they are apparently aiming for about 300km height! -> my calculation comes to 3000 satellites, they probably have a few in reserve, require a higher number of satellites to cover everywhere all the time etc.
Anyway, that would come to "too low" for the drag/orbit deay reasons..... So that would mean they are aiming for "the lowest the orbit-decay will alow".
Rereading my own message: Near: "You want them as close as possible."
I forgot to mention: "because the required power to transmit to the satellite increases with the distance."
Older satellite internet systems used technologies borrowed from "TV broadcasting". What they effectively did is broadcast everybody's downlink via the satellite and everybody-for-himself had to use a land-line for the uplink. The idea being that you like having a big downlink pipe and it might be acceptable to have "only" 56k uplink.
All that is going overboard, as I understand things. Yes, people are going to transmit their uplink bits to the satellites: the stated goal of these projects is "internet everywhere", even where landlines don't come. So on most my calculations (I hope all of them :-) ) I calculated the round-trip delay via satellite.
On the other hand, I did not take the groundstation-to-destination delays into account. Those are on the order of 60ms minimum for a transatlantic link, for exactly the same reason that a transatlantic satellite link will take that amount of time: the light speed.
(I just tried tracing packets to three American companies from Europe. All were reachable in less than 10ms (not enough to cross the ocean), with google performing worst from the three I tried: google, nbc, cnn. Apparently they all have servers serving european users here in europe).