It would not be practical for ISPs to start chasing direct peering deals with every minor content provider out there. They are singling Netflix out because Netflix alone accounts for ~33% of all prime-time traffic almost entirely through a single peer.
I would be surprised if pressure for direct peering spread much beyond the top three prime-time services in any given market. Google is ahead of all the peering disputes with their ISP/IX-hosted content caches and you do not see anyone complaining about this giving Google's content an edge over CDNs and smaller content distributors. Congestion? Those caches bypass all transit when they get a hit.
Diff. Speed with the tree: 40km/h Diff. Speed with child going in the same direction as the car: 30km/h
The child has a lower relative speed to the car so the car decides to crash in the child instead of the tree.
So, speed differential alone is not acceptable - the car's passengers will almost certainly survive crashing in the tree largely uninjured but the cyclist is likely to get severely injured or worse.
I'm a cyclist and usually do 25-30km/h. I would hate to know I have now become a deliberately preferred crash target by self-driving car algorithms.
their software will not always be able to calculate everything in the few milliseconds or so you might have to make such decisions.
Before worrying about how quickly the software and processor can get whatever they need to do done, I would worry about sensor latency, obstructions and failures: the best software and processing power in the world does you no good if they do not receive accurate data in time.
The idea of basic driving becoming increasingly dependent on delicate and often extremely expensive to repair/replace sensors remaining in perfect working order makes me nervous. We used to have cars that would keep running until the engine self-destructed from abuse to cars that might refuse to start because there is dust on the body or some other trivial thing.
It would also solve the issue of being unable to see out of the cockpit.
If you radically diffuse incoming light, that would also radically diffuse other incoming light as well somewhat like trying to see through a thick cloud. This still leaves you effectively blind unless your shield has a known transfer function that can be reversed by a camera and signal processing... but then your opponent may be able to deduct your shield's transfer function and re-focus their energy beams accordingly.
Another problem is what happens to your plasma shield when it gets hit by a high-power laser or other energy beam: unless your plasma shield is at temperatures and pressures approaching those required for nuclear fusion (what happens if external energy causes the shield's plasma to actually reach critical? I imagine it would cause some severe local disruption), a powerful enough laser would be able to locally super-heat the plasma (thin it and knock some material out of the field) and punch through it faster than flow rate can cool the hot-spot down - if you can afford pumping gigawatts into your shields and other systems, your opponents probably have the technology to pump gigawatts in their lasers.
Plasma shields sound nice in theory but in practice, they might not be practical much beyond protection from space junk and relatively crude/low-energy weapons.
If your civilization is advanced enough to have inter-stellar starships, the simplest way of dumping a few tons of nuclear waste without having to worry about environmental impact would be to load the material in a rocket/torpedo and fire it at the most convenient star you come across.
The only application developers who need to bother with binary packaging are those who decide to use the Native Development Kit.
All other developers who stick to the Android Development Kit's recommendations to NOT use the the NDK are just Java bytecode that gets recompiled to whatever the device needs either when the APK is initially installed (ART in Android 4.4+) or on-the-fly (JIT/Dalvik) as the application launches/runs.
You can get improvements on the same lithography size by using different materials, different dopants, different dopant concentrations, different circuit geometry, different layer thicknesses, different voltages, improving the overall process, finer clock and power gating, faster dynamic voltage switching, etc.
Shrinking circuitry is not the only way to reduce both static and dynamic power. It is merely the most commonly known factor since it has a 30+ years history of making transistors faster, cheaper, more power-efficient, etc.
Nuclear tests excluded, Chernobyl and Fukushima are the only two nuclear incidents with wide-ranging environmental effects. Chernobyl is mainly due to operator error and reckless disregard for even the most basic safety instructions (postmortem investigation found no evidence of any control rods in the core despite the manufacturer's manual saying a minimum of 20 or so control rods must be inserted at all times), trying to rush a 12+ hours ramp-down procedure in less than four hours while most of Fukushima's problems were due to poor plant layout, running all power feeds in the same conduits and putting most generators in floodable zones, making it impossible to restore power to cooling pumps in a timely fashion.
TMI was an exercise in dumb design with the status light bound to the switch instead of a position sensor to give feedback about the actual valve's state. Relying on thermodynamic tables and flow rates to estimate reactor coolant level which could have been directly observed by a simple float sensor did not help confused operators figure out what was happening either.
Those old "unsafe" plants may not have been as close to intrinsically safe as MSRs but they did not fail on a whim either. Most of those failures could have happened with MSRs too if MSR designers had not learned from others' past mistakes.
For every byte Netflix provides to Comcast, there is a paying Comcast customer who requested it.
Not really: Comcast subscribers are not paying for any particular quantity of Netflix or any other content and may not even be subscribers of. Flat monthly rates are based on bulk average costs plus markups where the low-volume users "subsidize" the high-volume users so the ISPs can meet their gross profit margin target.
As for "Comcast not providing transit to Netflix," keep in mind that this whole thing started with Comcast not liking how thin L3 was stretching their (Comcast-L3) peering agreement to avoid paying to deliver Netflix's traffic. If L3 had decided to re-negotiate their Comcast peering, L3 would have passed the bill to Netflix and Netflix would have passed it down to their subscribers, which is fundamentally what happened with the direct connect agreement Netflix ended up signing with Comcast to bypass L3 altogether.
Personally, I like the idea of popular high-bandwidth services eating a bigger chunk of the downstream costs so non-subscribers might not have to. It would also give content provider incentives to use new forms of more cost-efficient distribution methods.
CDNs are paid by content/service providers to get data from A to B, usually on a sender-pays basis. If the volume of data coming from them causes peering/transit to go imbalanced beyond what the agreements allows, it is not that hard to imagine the destination network being displeased about getting asked to eat the bill.
The main problem with this is most ISPs are unlikely to pass the bucks they get from CDNs back to their subscribers.
Managing QoS on a per-sub, per-service basis would quickly become a nightmare.
Simple PHB QoS on the other hand is supported by every quarter-decent router and can easily scale beyond Tbps but for it to be viable on its own, you need to be able to trust the source of QoS-tagged traffic and agree on the meaning and use of whichever QoS tag(s) get used to avoid expensive traffic policing.
By letting the deals be made by content providers and ISPs, you lose a lot of the transparency.
Equally important would be how much it costs.
Having to setup QoS on a per-subscriber, per-service basis carries a lot more administrative overhead and routing resource cost than setting it up as a network-wide policy - at the very least, the ISP would need to setup SPI firewalls on their transit links or somewhere downstream to pick which packets have to be prioritized on a per-subscriber basis, which can easily be more expensive than setting up a direct peering agreement with associated links and hardware.
Would you prefer ISPs investing most of your direct-paid QoS fees in SPI equipment for priority tagging or ISPs investing your indirectly paid fees in adding capacity to accommodate content/service-provider-paid peering arrangements for the services you are paying for? The peering option is cheaper overall, should be more reliable and faster. The peering deal also has the benefit of not being tied to your specific ISP account so you can access your online services from your neighbors, friends, family, etc. with full benefits as long as their ISP also has such a peering agreement, which is going to be handy if you often watch Netflix or similar on your portable devices outside your home.
Allowing people to request and pay for preferential treatment of their traffic might be a nice idea but network-wide agreements for popular high-bandwidth or time-critical services is nicer IMO.
If your old iPad was not still a good-enough device for you, you would not be using it anymore or would at the very least be itching to get a newer iPad.
There is no universal definition of how good is good enough. "Good enough" does not have to be an Android or Windows-based device and it does not need to be a low-end device either. It just needs to let you do whatever you expect the device to do well enough that you do not feel like anything better/newer is currently worth the expense.
First, those who aren't heavy computer computer users; the grandmothers of the world who check their email once a day.
In my immediate family, I am the only one who really needs a PC. The rest of them rarely do anything more compute-intensive than watching cats on youtube or facebook. Among my friends, about half of them could manage with only a tablet and most of the other half uses their PCs mainly for gaming... that leaves 10-20% of the people I know genuinely needing a PC for something other than gaming.
Isn't saying your three years old iPad being enough for your needs exactly the same as saying you already own a good-enough device?
That was exactly my point. At the rate performance vs price is improving, most people who do not already own good-enough soon will. The annual upgrade fever is cooling down, now the high-end market is transitioning to something closer to a replacement economy but there are still plenty of sales to be had by raising the entry-level bar like the N7v1 did two years ago.
If the ISPs let their regular peering degrade to the point a substantial number of people have to buy traffic priority for their over-the-top services to work, this still effectively become an additional tax/fee on top of those specific services and the costs of handling priority routing on a customer-by-customer, service-by-service basis will be much higher than prioritizing a frequently requested service across the board and charging the peer that traffic comes from for it.
The EU's network neutrality explicitly allow content providers to sign agreements for enhanced service quality:
"Providers of content, applications and services and providers of electronic communications to the public should therefore be free to conclude specialised services agreements on defined levels of quality of service as long as such agreements do not substantially impair the general quality of internet access services."
Once technology becomes "good enough" for a substantial chunk of the market and a substantial chunk of the market already owns such a "good enough" device, people are no longer so eager to spend globs of cash on incrementally better devices. The threshold for "good enough" is now starting to move down the price point ladder so interest in premium-priced models will likely fade in the near-future - it becomes difficult to justify spending over $500 on a tablet when you can get most of the same features on $150-250 models.
Well, not providing premium traffic does not work either since ISPs simply end up in peering disputes over lopsided transit balances that entitle them to seek compensation for excess traffic and let links to those peers go to congestion hell if the peers refuse to pay up.
If Netflix and company had to pay premiums to L3 and others for L3 and company to pay dues to ISPs for their transit imbalance, Netflix and company would likely request preferential treatment from transit providers and CDNs to get their fair share of those widened pipes to ISPs they are paying extras for.
Direct peering achieves almost the exact same thing but without the middlemen. This is no different from peering with CDNs or signing up for hosting Google cache nodes.
For most websites who serve relatively low-bandwidth content or a relatively small number of people, this probably won't have that much of an effect - it is only a tiny percentage of all websites that have aggregate peak bandwidth high enough for direct peering to make any sense bothering with and even the previous network neutrality bill would not have prevented that.
Even the European Union which many look at for being more pro-consumer than almost anywhere else in the world has a network neutrality bill that allows direct peering deals to enhance performance, quality of service and reliability of popular online services as long as it does not interfere with or otherwise degrade other services.
If you relied on VoIP, would you like the option to pay maybe $1/month extra to have a 1Mbps fully-QoS'd channel to guarantee that your VoIP traffic always gets through no matter how badly intermediate networks between your modem and VoIP provider might be? That's one of the use-cases the EUP offered as a justification for having to allow some degree of traffic prioritization.
As long as ISPs are not allowed to intentionally degrade non-premium traffic on the back of direct-peering deals, I see no fundamental problem with it.
AMD serves the desktop/server-with-big-TV-Display market very well and should not be questioned.
The conventional PC market is already undergoing a substantial shake-down with people embracing smartphones and tablets in bulk. At the rate Android platforms are evolving, we are only two or three years away from the average phone and tablet being able to handle just about anything the average person might want to throw at them. For many people, current devices have already passed the good-enough milestone. In my immediate family, I am the only one who genuinely needs a PC - both of my sisters and their boyfriends do just about all their online stuff using their phones and own a computer or laptop they hardly ever use anymore and neither of my parents use the PCs I gave them for anything much beyond basic web browsing.
your anaolgy does not work. PCs and mini-computers were fundementally different, applications written for one would, generally, not work on the other.
Same goes for PCs and tablets but that is not preventing people from shifting significant chunks of their everyday computing from their PCs to their phones and tablets. The more powerful mobile devices become, the more applications will get adapted to them and the fewer reasons people will have to stick to conventional PCs just like how companies shifted the bulk of their workloads from minis to micros as micros became both more capable and much cheaper.
Freezing aside, another and probably by far the biggest reason ethanol is a common (and often mandatory) fuel additive is to reduce emissions - many countries and states require that automotive gasoline contains 5-10% ethanol.
Slashdot Mirror
It would not be practical for ISPs to start chasing direct peering deals with every minor content provider out there. They are singling Netflix out because Netflix alone accounts for ~33% of all prime-time traffic almost entirely through a single peer.
I would be surprised if pressure for direct peering spread much beyond the top three prime-time services in any given market. Google is ahead of all the peering disputes with their ISP/IX-hosted content caches and you do not see anyone complaining about this giving Google's content an edge over CDNs and smaller content distributors. Congestion? Those caches bypass all transit when they get a hit.
You said picking the crash target is easy: simply use the lowest speed differential.
Car's absolute speed: 40km/h
Tree's absolute speed: 0km/h
Child's absolute speed on bicycle: 10km/h
Diff. Speed with the tree: 40km/h
Diff. Speed with child going in the same direction as the car: 30km/h
The child has a lower relative speed to the car so the car decides to crash in the child instead of the tree.
So, speed differential alone is not acceptable - the car's passengers will almost certainly survive crashing in the tree largely uninjured but the cyclist is likely to get severely injured or worse.
I'm a cyclist and usually do 25-30km/h. I would hate to know I have now become a deliberately preferred crash target by self-driving car algorithms.
their software will not always be able to calculate everything in the few milliseconds or so you might have to make such decisions.
Before worrying about how quickly the software and processor can get whatever they need to do done, I would worry about sensor latency, obstructions and failures: the best software and processing power in the world does you no good if they do not receive accurate data in time.
The idea of basic driving becoming increasingly dependent on delicate and often extremely expensive to repair/replace sensors remaining in perfect working order makes me nervous. We used to have cars that would keep running until the engine self-destructed from abuse to cars that might refuse to start because there is dust on the body or some other trivial thing.
You can "remove" it by installing an aftermarket launcher screen like Nova.
Just about every "default" app and function can be overridden with aftermarket apps.
It would also solve the issue of being unable to see out of the cockpit.
If you radically diffuse incoming light, that would also radically diffuse other incoming light as well somewhat like trying to see through a thick cloud. This still leaves you effectively blind unless your shield has a known transfer function that can be reversed by a camera and signal processing... but then your opponent may be able to deduct your shield's transfer function and re-focus their energy beams accordingly.
Another problem is what happens to your plasma shield when it gets hit by a high-power laser or other energy beam: unless your plasma shield is at temperatures and pressures approaching those required for nuclear fusion (what happens if external energy causes the shield's plasma to actually reach critical? I imagine it would cause some severe local disruption), a powerful enough laser would be able to locally super-heat the plasma (thin it and knock some material out of the field) and punch through it faster than flow rate can cool the hot-spot down - if you can afford pumping gigawatts into your shields and other systems, your opponents probably have the technology to pump gigawatts in their lasers.
Plasma shields sound nice in theory but in practice, they might not be practical much beyond protection from space junk and relatively crude/low-energy weapons.
If your civilization is advanced enough to have inter-stellar starships, the simplest way of dumping a few tons of nuclear waste without having to worry about environmental impact would be to load the material in a rocket/torpedo and fire it at the most convenient star you come across.
There is also the issue of emissions - two-strokes engines usually fare much worse than four-strokes ones.
This seems to go against everything the push for hybrids and all-electric vehicles stand for.
The only application developers who need to bother with binary packaging are those who decide to use the Native Development Kit.
All other developers who stick to the Android Development Kit's recommendations to NOT use the the NDK are just Java bytecode that gets recompiled to whatever the device needs either when the APK is initially installed (ART in Android 4.4+) or on-the-fly (JIT/Dalvik) as the application launches/runs.
You can get improvements on the same lithography size by using different materials, different dopants, different dopant concentrations, different circuit geometry, different layer thicknesses, different voltages, improving the overall process, finer clock and power gating, faster dynamic voltage switching, etc.
Shrinking circuitry is not the only way to reduce both static and dynamic power. It is merely the most commonly known factor since it has a 30+ years history of making transistors faster, cheaper, more power-efficient, etc.
Nuclear tests excluded, Chernobyl and Fukushima are the only two nuclear incidents with wide-ranging environmental effects. Chernobyl is mainly due to operator error and reckless disregard for even the most basic safety instructions (postmortem investigation found no evidence of any control rods in the core despite the manufacturer's manual saying a minimum of 20 or so control rods must be inserted at all times), trying to rush a 12+ hours ramp-down procedure in less than four hours while most of Fukushima's problems were due to poor plant layout, running all power feeds in the same conduits and putting most generators in floodable zones, making it impossible to restore power to cooling pumps in a timely fashion.
TMI was an exercise in dumb design with the status light bound to the switch instead of a position sensor to give feedback about the actual valve's state. Relying on thermodynamic tables and flow rates to estimate reactor coolant level which could have been directly observed by a simple float sensor did not help confused operators figure out what was happening either.
Those old "unsafe" plants may not have been as close to intrinsically safe as MSRs but they did not fail on a whim either. Most of those failures could have happened with MSRs too if MSR designers had not learned from others' past mistakes.
For every byte Netflix provides to Comcast, there is a paying Comcast customer who requested it.
Not really: Comcast subscribers are not paying for any particular quantity of Netflix or any other content and may not even be subscribers of. Flat monthly rates are based on bulk average costs plus markups where the low-volume users "subsidize" the high-volume users so the ISPs can meet their gross profit margin target.
As for "Comcast not providing transit to Netflix," keep in mind that this whole thing started with Comcast not liking how thin L3 was stretching their (Comcast-L3) peering agreement to avoid paying to deliver Netflix's traffic. If L3 had decided to re-negotiate their Comcast peering, L3 would have passed the bill to Netflix and Netflix would have passed it down to their subscribers, which is fundamentally what happened with the direct connect agreement Netflix ended up signing with Comcast to bypass L3 altogether.
Personally, I like the idea of popular high-bandwidth services eating a bigger chunk of the downstream costs so non-subscribers might not have to. It would also give content provider incentives to use new forms of more cost-efficient distribution methods.
CDNs are paid by content/service providers to get data from A to B, usually on a sender-pays basis. If the volume of data coming from them causes peering/transit to go imbalanced beyond what the agreements allows, it is not that hard to imagine the destination network being displeased about getting asked to eat the bill.
The main problem with this is most ISPs are unlikely to pass the bucks they get from CDNs back to their subscribers.
It depends what sort of QoS we are talking about.
Managing QoS on a per-sub, per-service basis would quickly become a nightmare.
Simple PHB QoS on the other hand is supported by every quarter-decent router and can easily scale beyond Tbps but for it to be viable on its own, you need to be able to trust the source of QoS-tagged traffic and agree on the meaning and use of whichever QoS tag(s) get used to avoid expensive traffic policing.
But what is really important is who pays.
By letting the deals be made by content providers and ISPs, you lose a lot of the transparency.
Equally important would be how much it costs.
Having to setup QoS on a per-subscriber, per-service basis carries a lot more administrative overhead and routing resource cost than setting it up as a network-wide policy - at the very least, the ISP would need to setup SPI firewalls on their transit links or somewhere downstream to pick which packets have to be prioritized on a per-subscriber basis, which can easily be more expensive than setting up a direct peering agreement with associated links and hardware.
Would you prefer ISPs investing most of your direct-paid QoS fees in SPI equipment for priority tagging or ISPs investing your indirectly paid fees in adding capacity to accommodate content/service-provider-paid peering arrangements for the services you are paying for? The peering option is cheaper overall, should be more reliable and faster. The peering deal also has the benefit of not being tied to your specific ISP account so you can access your online services from your neighbors, friends, family, etc. with full benefits as long as their ISP also has such a peering agreement, which is going to be handy if you often watch Netflix or similar on your portable devices outside your home.
Allowing people to request and pay for preferential treatment of their traffic might be a nice idea but network-wide agreements for popular high-bandwidth or time-critical services is nicer IMO.
If your old iPad was not still a good-enough device for you, you would not be using it anymore or would at the very least be itching to get a newer iPad.
There is no universal definition of how good is good enough. "Good enough" does not have to be an Android or Windows-based device and it does not need to be a low-end device either. It just needs to let you do whatever you expect the device to do well enough that you do not feel like anything better/newer is currently worth the expense.
First, those who aren't heavy computer computer users; the grandmothers of the world who check their email once a day.
In my immediate family, I am the only one who really needs a PC. The rest of them rarely do anything more compute-intensive than watching cats on youtube or facebook. Among my friends, about half of them could manage with only a tablet and most of the other half uses their PCs mainly for gaming... that leaves 10-20% of the people I know genuinely needing a PC for something other than gaming.
Isn't saying your three years old iPad being enough for your needs exactly the same as saying you already own a good-enough device?
That was exactly my point. At the rate performance vs price is improving, most people who do not already own good-enough soon will. The annual upgrade fever is cooling down, now the high-end market is transitioning to something closer to a replacement economy but there are still plenty of sales to be had by raising the entry-level bar like the N7v1 did two years ago.
If the ISPs let their regular peering degrade to the point a substantial number of people have to buy traffic priority for their over-the-top services to work, this still effectively become an additional tax/fee on top of those specific services and the costs of handling priority routing on a customer-by-customer, service-by-service basis will be much higher than prioritizing a frequently requested service across the board and charging the peer that traffic comes from for it.
The EU's network neutrality explicitly allow content providers to sign agreements for enhanced service quality:
"Providers of content, applications and services and providers of electronic communications to the public should therefore be free to conclude specialised services agreements on defined levels of quality of service as long as such agreements do not substantially impair the general quality of internet access services."
Once technology becomes "good enough" for a substantial chunk of the market and a substantial chunk of the market already owns such a "good enough" device, people are no longer so eager to spend globs of cash on incrementally better devices. The threshold for "good enough" is now starting to move down the price point ladder so interest in premium-priced models will likely fade in the near-future - it becomes difficult to justify spending over $500 on a tablet when you can get most of the same features on $150-250 models.
Well, not providing premium traffic does not work either since ISPs simply end up in peering disputes over lopsided transit balances that entitle them to seek compensation for excess traffic and let links to those peers go to congestion hell if the peers refuse to pay up.
If Netflix and company had to pay premiums to L3 and others for L3 and company to pay dues to ISPs for their transit imbalance, Netflix and company would likely request preferential treatment from transit providers and CDNs to get their fair share of those widened pipes to ISPs they are paying extras for.
Direct peering achieves almost the exact same thing but without the middlemen. This is no different from peering with CDNs or signing up for hosting Google cache nodes.
For most websites who serve relatively low-bandwidth content or a relatively small number of people, this probably won't have that much of an effect - it is only a tiny percentage of all websites that have aggregate peak bandwidth high enough for direct peering to make any sense bothering with and even the previous network neutrality bill would not have prevented that.
Even the European Union which many look at for being more pro-consumer than almost anywhere else in the world has a network neutrality bill that allows direct peering deals to enhance performance, quality of service and reliability of popular online services as long as it does not interfere with or otherwise degrade other services.
If you relied on VoIP, would you like the option to pay maybe $1/month extra to have a 1Mbps fully-QoS'd channel to guarantee that your VoIP traffic always gets through no matter how badly intermediate networks between your modem and VoIP provider might be? That's one of the use-cases the EUP offered as a justification for having to allow some degree of traffic prioritization.
As long as ISPs are not allowed to intentionally degrade non-premium traffic on the back of direct-peering deals, I see no fundamental problem with it.
They may be able to support their kids economically but at 50+ years old, they may have a hard time with the stamina to keep up with teen-aged kids.
If I had kids, I would prefer dealing with them in my 30s while my own health is still unlikely to become a problem.
AMD serves the desktop/server-with-big-TV-Display market very well and should not be questioned.
The conventional PC market is already undergoing a substantial shake-down with people embracing smartphones and tablets in bulk. At the rate Android platforms are evolving, we are only two or three years away from the average phone and tablet being able to handle just about anything the average person might want to throw at them. For many people, current devices have already passed the good-enough milestone. In my immediate family, I am the only one who genuinely needs a PC - both of my sisters and their boyfriends do just about all their online stuff using their phones and own a computer or laptop they hardly ever use anymore and neither of my parents use the PCs I gave them for anything much beyond basic web browsing.
your anaolgy does not work. PCs and mini-computers were fundementally different, applications written for one would, generally, not work on the other.
Same goes for PCs and tablets but that is not preventing people from shifting significant chunks of their everyday computing from their PCs to their phones and tablets. The more powerful mobile devices become, the more applications will get adapted to them and the fewer reasons people will have to stick to conventional PCs just like how companies shifted the bulk of their workloads from minis to micros as micros became both more capable and much cheaper.
It is almost exactly the same pattern.
Freezing aside, another and probably by far the biggest reason ethanol is a common (and often mandatory) fuel additive is to reduce emissions - many countries and states require that automotive gasoline contains 5-10% ethanol.