Sometimes the giveaway is substitute parts, ABC corp will ony give the assembly line in china sufficint parts to make the legitimate products (plus a handful of spares for failures) so to make the counterfiets the factory has to source substitutes for custom parts, those substitutes may be slightly different from the slightly different from the legit parts.
AIUI the problem is that the US has H1B (and possiblly other visas but H1B seems to be the main one) visas that mean the immigrant can't easilly change jobs and if they lose their job they will most likely be sent home. This puts the employer in a much stronger position than they would be if they were employing an american citizen or permanent resident.
If the immigrant puts up with the shit for a sufficint number of years then AIUI the can generally become a permanent resident (and eventually if they wish a citizen).
Not really. If anything, the dispute should be between Cogent and Comcast directly, and either can choose to charge their subscribers more.
Or they can choose to leave the link(s) in a shitty congested state. Comcast like's this situation, shitty netflix (and any other internet video providers who happen to be behind cogent) means less competition for their cable TV products*. Cogent don't care much, their buisness is selling cheap connections not high quality ones.
Netflix can choose to respond to this by accepting the shitty congested link, by finding a different upstream provider with a less congested link to comcast (and likely also a higher price) or by cutting out the middleman and making their own deal with comcast. It seems they have chosen the latter.
* Which brings us to the real problem, too much vertical integration meaning the interests of ISPs don't align with the interests of their customers.
No sane ISP would agree to provide a guarantee to provide every packet you send to arbitary destinations since it would be impractical for them to provide that service. At best they will provide a gaurantee that it will reach the edge routers of their network without encountering congestion. At worst they will provide no guarantees at all as to what happens once it leaves your link to them and enters the core of their network.
Afaict microsoft's greatest threat in recent has not been that windows would be replaced on the desktop/laptop but that computing would shift away from desktops/laptops (where they hold a near monopoly) to smartphones/tablets (where they are an also-ran).
Afaict MS thought that by forcing metro (tablet interface+forced appstore) onto desktops and laptops they could both gain a footing in the tablet market and also get some of that easy money that apple enjoys from their appstore.
Unfortunately for MS it hasn't really worked out that way. the crippled RT was largely DOA. Regular win8 on tablets has become popular with some niches but is held back by the cost and bulk of x86 hardware.
Vote with your wallets and pick ISPs that do a better job with their peering arrangements.
The problem is here we have a case of a teir 1 provider vertically integrated with an access provider that carries local monopolies or effective monopolies in some areas. AIUI in the US it has been ruled that carriers don't have to share fiber or cable networks so the only networks that have real ISP competition are the crappy old DSL ones (IIRC in some areas these are even being actively removed to force people onto the single ISP fiber networks).
I disagee, the root of the problem is that a teir 1 provider is vertically integrated with a last mile provider that has monopoly or near-monopoly power in some areas.
If the last mile provider was not vertically integrated with a teir 1 it would be in their interests to set up good peering with large conent providers directly (as happens here in the UK) to avoid paying upstream transit bills but because they are verticially integrated with a teir 1 and don't have to buy transit from anyone it makes sense for them to play hardball on peering. Think about it? why are cogent involved here in the first place? Most likely because Verizon refused to peer with netflix.
Verizon and cogent are both transit free networks, if they depeer from each other then single homed customers of the two networks will simply not be able to communicate with each other.
Cogent may host much of Netflix, but they are by no means a Tier 1. This may no longer be the case - like I said, I have been out of the Tier 1 ISP world for years - but at least historically Cogent was known as a bottom feeder of the industry. They charged dirt-cheap rates but ran a crappy network and skimped on their upstream connections to cut costs.
AIUI cogent nowadays would be best described as a "wannabe tier 1", they are at the point where they no longer buy transit from anyone but it is widely suspected they rely on paid peering (kind of a middle ground between peering and transit) and they get into peering fights from time to time
but VZ has said, in effect, "you are not our 'peer'" and beyond a certain amount of peering bandwidth, you should start buying
And if customers understood what was going on and had access to a competitive markets of ISPs they would be saying to verizon "Sort out your little spat or we won't be your customers anymore".
Sadly many parts of the US are in the horrible situation of having their local monopoly vertically integrated with a teir 1 carrier. I believe some parts of europe have similar problems (here in the UK none of our local monopolies are vertically integrated with a teir 1).
Indeed it doesn't, if you want to run a quake server out of your laundry room with a public IPv4 IP you will likely have to pay for the privilage of having that public v4 IP.
Price is set by companies based on their assessment of supply and demand, not cost.
Price is set by companies in an attempt to maximise their profit, supply and demand is part of that but cost is also important.
They can afford to take a small loss on each console early in the generation where they are trying to build market share in the expectation they will make it up later through game sales and through having a better market postion later in the generation but there are limits to how big a loss can reasonablly be taken.
In principle* if you have n antennas then you can completely control the waveform at n locations of your choice. So the multi-antenna techniques will have to be combined with one or more traditional multiple access schemes (CDMA, FDMA, TDMA) to serve realistic numbers of users.
Still this has the potential to dramatically increase average throughput per antenna over more conventional techniques (at the cost of a LOT of processing power and very good interconnections between the cells)
* Practice is a bit messier because of real world imperfections
This will allow global routing tables to more accurately reflect the structure there is between ISPs, shrinking their size.
That was what the ivory tower guys thought, they came up with grand plans for heirachical routing that were never really adopted because they didn't respect the realities of the internet. They also came up with crazy ideas that network admins would find it acceptable to run multiple prefixes on their networks at once and that end hosts would somehow be able to determine which was the best to use for a given destination.
In reality no ISP or large company wants to tie their addresses to their current choice of upstream providers or manage multiple prefixes in paralell, so each ISP/large company is still going to end up with a route in the global routing table.
The table should still have less entries than the v4 one because most ISPs and large companies should have one block each rather than building up multiple blocks over time but it's not going to be as small as some people hoped.
Interesting, do you have a source for that figure? it sounds very cheap to me! At that price you would have to have a very low utilisation ratio to make it worth the cost of restructuring.
It seems merkel either doesn't know what she is talking about or is oversimplifiying.
The real answer depends on what you mean by "send an email from munich to paris"
If you mean a user using a mailserver based in munich sends an email to a user using a mailserver based in paris then the mail is unlikely to go anwhere near the US. Europe (unlike some other parts of the world) has a highly functiona internet infrastructure with good in-region peering.
However if you mean a user in munich sends an email to a user in paris then things get messier. Some people use email services run internally by their ISP or employer but a lot use email services run by big american corporations. That makes it relatively easy for the americans to spy on them (even if the email doesn't usually go through the US it's all too easy for the american parent company to interfere with their subsidaries in europe.
Protocols can be implemented in either software or hardware, which is chosen depends on various tradeoffs including cost, performance and flexibility.
In order to make a device that is IPv4 only able to run IPv6 is software changes. Carrier grade hardware will have firmware updates that add this functionality.
Able to run and able to run with decent performance and reliability are unfortunately not the same thing.
A high end router has a CPU that does various housekeeping tasks (updating routes, administrative access etc)and a forwarding engine that does the real work of matching packets to routes and pushing them out of the correct interface. If the forwarding engine is not designed for IPv6 then it is unlikely to be able to support it with a mere firmware upgrade.
AIUI some such routers "support" ipv6 but handle it with the CPU instead of the routing engine, so they work ok if IPv6 is a small proportion of total traffic but as the number of IPv6 packets in the mix grows they fall over.
So you are trying to say that routers have a hardware implementation of a map data structure instead of system memory, and that map is somehow mysteriously backed by non-standard computer memory..
Internet routing is not a simple map operation, it's a prefix match operation against a very large prefix table and it has to be performed at very high speed (potentially tens of millions of packets per second). For a large router at an ISP it's also likely to have poor locality so caches won't help as much as you would hope.
Further, you suppose that this hardware based map uses more power than a software based approach
More power than the same ammount of regular ram sure but regular ram isn't going to keep up.
Having said that the large size of the v6 address space means that addresses can be allocated in a way that brings us much closer to the ideal of one prefix per AS. There is also much less historical cruft, so I suspect that the IPv6 routing table will not be anwhere near as bad as the GP asserts.
I would also point out that the place you have the really big routing tables is on the core networks and yet most ISPs seem to have upgraded their core networks to IPv6 already, it's the access networks that are lagging.
I would think that if/when IPv4 addresses become prohibitively expensive for individual webservers then the hosting providers would start running such meta-servers as a service for their clients.
IP addresses are allocated regionally, the IANA handed out/8 blocks to the regional internet registries (RIRs). The RIRs are ARIN (US/canada), LACNIC (south/central america up to and including mexico), RIPE (europe and west asia), APNIC (east asia and australia) and afrinic (africa). The RIRs in turn allocate them either to LIRs (mainly ISPs) or sometimes directly to end users. The LIRs allocate them to end users as part of service offering.
The IANA reserved the last five/8's for allocation in a special final allocation, one to each RIR after all other/8 global unicast blocks had been allocated. A couple of years ago APNIC made an address space request that triggered that final allocation process leaving the IANA with no free blocks to hand out.
AIUI most if not all of the RIRs have announced special measures to be applied as their stocks of IPv4 addresses run low. APNIC and RIPE are now deep into special measures with new allocations only available in special circumstances (a small block for a new LIR, blocks for use by internet exchange points). ARIN are less deep into special measures with regular allocations still available though under tigher rules than before.
IPv6 is a cleaner soloution in the long term but it requires cooperation from basically everyone involved in the internet industry (software vendors, OS vendors, hardware vendors, carriers) to take it from the "cool toy" stage to the "replacement for IPv4" stage. To go v6 only the client software has to support it, the OS on the end device has to support it, the home gateway has to support it, the ISP has to support it and either a horrible translation mechanism needs to be deployed by the ISP or the server has to support it.
NAT is messier but solves the immediate problem for your organisation without requiring cooperation from anyone else.
Why should an increase in the money supply devalue the dollar in my pocket?
Producers want to get as much money as they can for the goods and services they produce. If at the current price point demand exceeds supply then prices will rise to balance supply with demand. Putting more money into the system increases demand at a given price point and so drives prices up.
The alternative to this is for the governent to force producers to sell at an artificially low price but that will lead to shortages.
Agree completely, but the problem is that to an outside observer it is impossible to know how many comparisons were actually done.
And even if the researchers are being honest about the number of comparisons THEY did it is very likely that multiple people will be independently working on the same problem. If all of those people individually only publish positive results then you get the same problem.
Slashdot Mirror
Sometimes the giveaway is substitute parts, ABC corp will ony give the assembly line in china sufficint parts to make the legitimate products (plus a handful of spares for failures) so to make the counterfiets the factory has to source substitutes for custom parts, those substitutes may be slightly different from the slightly different from the legit parts.
but they don't appear to be willing to do that.
AIUI the problem is that the US has H1B (and possiblly other visas but H1B seems to be the main one) visas that mean the immigrant can't easilly change jobs and if they lose their job they will most likely be sent home. This puts the employer in a much stronger position than they would be if they were employing an american citizen or permanent resident.
If the immigrant puts up with the shit for a sufficint number of years then AIUI the can generally become a permanent resident (and eventually if they wish a citizen).
Not really. If anything, the dispute should be between Cogent and Comcast directly, and either can choose to charge their subscribers more.
Or they can choose to leave the link(s) in a shitty congested state. Comcast like's this situation, shitty netflix (and any other internet video providers who happen to be behind cogent) means less competition for their cable TV products*. Cogent don't care much, their buisness is selling cheap connections not high quality ones.
Netflix can choose to respond to this by accepting the shitty congested link, by finding a different upstream provider with a less congested link to comcast (and likely also a higher price) or by cutting out the middleman and making their own deal with comcast. It seems they have chosen the latter.
* Which brings us to the real problem, too much vertical integration meaning the interests of ISPs don't align with the interests of their customers.
No sane ISP would agree to provide a guarantee to provide every packet you send to arbitary destinations since it would be impractical for them to provide that service. At best they will provide a gaurantee that it will reach the edge routers of their network without encountering congestion. At worst they will provide no guarantees at all as to what happens once it leaves your link to them and enters the core of their network.
Afaict microsoft's greatest threat in recent has not been that windows would be replaced on the desktop/laptop but that computing would shift away from desktops/laptops (where they hold a near monopoly) to smartphones/tablets (where they are an also-ran).
Afaict MS thought that by forcing metro (tablet interface+forced appstore) onto desktops and laptops they could both gain a footing in the tablet market and also get some of that easy money that apple enjoys from their appstore.
Unfortunately for MS it hasn't really worked out that way. the crippled RT was largely DOA. Regular win8 on tablets has become popular with some niches but is held back by the cost and bulk of x86 hardware.
Vote with your wallets and pick ISPs that do a better job with their peering arrangements.
The problem is here we have a case of a teir 1 provider vertically integrated with an access provider that carries local monopolies or effective monopolies in some areas. AIUI in the US it has been ruled that carriers don't have to share fiber or cable networks so the only networks that have real ISP competition are the crappy old DSL ones (IIRC in some areas these are even being actively removed to force people onto the single ISP fiber networks).
I disagee, the root of the problem is that a teir 1 provider is vertically integrated with a last mile provider that has monopoly or near-monopoly power in some areas.
If the last mile provider was not vertically integrated with a teir 1 it would be in their interests to set up good peering with large conent providers directly (as happens here in the UK) to avoid paying upstream transit bills but because they are verticially integrated with a teir 1 and don't have to buy transit from anyone it makes sense for them to play hardball on peering. Think about it? why are cogent involved here in the first place? Most likely because Verizon refused to peer with netflix.
Verizon and cogent are both transit free networks, if they depeer from each other then single homed customers of the two networks will simply not be able to communicate with each other.
A proper peering termination means that both parties would return to Internet transit to get to each other
But neither verizon or cogent buy transit so if they depeer each other then there will simply be no routes.
Cogent may host much of Netflix, but they are by no means a Tier 1. This may no longer be the case - like I said, I have been out of the Tier 1 ISP world for years - but at least historically Cogent was known as a bottom feeder of the industry. They charged dirt-cheap rates but ran a crappy network and skimped on their upstream connections to cut costs.
AIUI cogent nowadays would be best described as a "wannabe tier 1", they are at the point where they no longer buy transit from anyone but it is widely suspected they rely on paid peering (kind of a middle ground between peering and transit) and they get into peering fights from time to time
but VZ has said, in effect, "you are not our 'peer'" and beyond a certain amount of peering bandwidth, you should start buying
And if customers understood what was going on and had access to a competitive markets of ISPs they would be saying to verizon "Sort out your little spat or we won't be your customers anymore".
Sadly many parts of the US are in the horrible situation of having their local monopoly vertically integrated with a teir 1 carrier. I believe some parts of europe have similar problems (here in the UK none of our local monopolies are vertically integrated with a teir 1).
Indeed it doesn't, if you want to run a quake server out of your laundry room with a public IPv4 IP you will likely have to pay for the privilage of having that public v4 IP.
Price is set by companies based on their assessment of supply and demand, not cost.
Price is set by companies in an attempt to maximise their profit, supply and demand is part of that but cost is also important.
They can afford to take a small loss on each console early in the generation where they are trying to build market share in the expectation they will make it up later through game sales and through having a better market postion later in the generation but there are limits to how big a loss can reasonablly be taken.
In principle* if you have n antennas then you can completely control the waveform at n locations of your choice. So the multi-antenna techniques will have to be combined with one or more traditional multiple access schemes (CDMA, FDMA, TDMA) to serve realistic numbers of users.
Still this has the potential to dramatically increase average throughput per antenna over more conventional techniques (at the cost of a LOT of processing power and very good interconnections between the cells)
* Practice is a bit messier because of real world imperfections
going into well served areas means a lot of potentially angry customers who can't wait to switch
Do you think those who are currently stuck on dialup or 1mbps DSL are happy about the fact
This will allow global routing tables to more accurately reflect the structure there is between ISPs, shrinking their size.
That was what the ivory tower guys thought, they came up with grand plans for heirachical routing that were never really adopted because they didn't respect the realities of the internet. They also came up with crazy ideas that network admins would find it acceptable to run multiple prefixes on their networks at once and that end hosts would somehow be able to determine which was the best to use for a given destination.
In reality no ISP or large company wants to tie their addresses to their current choice of upstream providers or manage multiple prefixes in paralell, so each ISP/large company is still going to end up with a route in the global routing table.
The table should still have less entries than the v4 one because most ISPs and large companies should have one block each rather than building up multiple blocks over time but it's not going to be as small as some people hoped.
you can make about $20/IP
Interesting, do you have a source for that figure? it sounds very cheap to me! At that price you would have to have a very low utilisation ratio to make it worth the cost of restructuring.
It seems merkel either doesn't know what she is talking about or is oversimplifiying.
The real answer depends on what you mean by "send an email from munich to paris"
If you mean a user using a mailserver based in munich sends an email to a user using a mailserver based in paris then the mail is unlikely to go anwhere near the US. Europe (unlike some other parts of the world) has a highly functiona internet infrastructure with good in-region peering.
However if you mean a user in munich sends an email to a user in paris then things get messier. Some people use email services run internally by their ISP or employer but a lot use email services run by big american corporations. That makes it relatively easy for the americans to spy on them (even if the email doesn't usually go through the US it's all too easy for the american parent company to interfere with their subsidaries in europe.
IP is a protocol.
True
It is software, not hardware.
Protocols can be implemented in either software or hardware, which is chosen depends on various tradeoffs including cost, performance and flexibility.
In order to make a device that is IPv4 only able to run IPv6 is software changes. Carrier grade hardware will have firmware updates that add this functionality.
Able to run and able to run with decent performance and reliability are unfortunately not the same thing.
A high end router has a CPU that does various housekeeping tasks (updating routes, administrative access etc)and a forwarding engine that does the real work of matching packets to routes and pushing them out of the correct interface. If the forwarding engine is not designed for IPv6 then it is unlikely to be able to support it with a mere firmware upgrade.
AIUI some such routers "support" ipv6 but handle it with the CPU instead of the routing engine, so they work ok if IPv6 is a small proportion of total traffic but as the number of IPv6 packets in the mix grows they fall over.
So you are trying to say that routers have a hardware implementation of a map data structure instead of system memory, and that map is somehow mysteriously backed by non-standard computer memory..
Internet routing is not a simple map operation, it's a prefix match operation against a very large prefix table and it has to be performed at very high speed (potentially tens of millions of packets per second). For a large router at an ISP it's also likely to have poor locality so caches won't help as much as you would hope.
Further, you suppose that this hardware based map uses more power than a software based approach
More power than the same ammount of regular ram sure but regular ram isn't going to keep up.
Having said that the large size of the v6 address space means that addresses can be allocated in a way that brings us much closer to the ideal of one prefix per AS. There is also much less historical cruft, so I suspect that the IPv6 routing table will not be anwhere near as bad as the GP asserts.
I would also point out that the place you have the really big routing tables is on the core networks and yet most ISPs seem to have upgraded their core networks to IPv6 already, it's the access networks that are lagging.
I would think that if/when IPv4 addresses become prohibitively expensive for individual webservers then the hosting providers would start running such meta-servers as a service for their clients.
Your memory is fuzzy.
IP addresses are allocated regionally, the IANA handed out /8 blocks to the regional internet registries (RIRs). The RIRs are ARIN (US/canada), LACNIC (south/central america up to and including mexico), RIPE (europe and west asia), APNIC (east asia and australia) and afrinic (africa). The RIRs in turn allocate them either to LIRs (mainly ISPs) or sometimes directly to end users. The LIRs allocate them to end users as part of service offering.
The IANA reserved the last five /8's for allocation in a special final allocation, one to each RIR after all other /8 global unicast blocks had been allocated. A couple of years ago APNIC made an address space request that triggered that final allocation process leaving the IANA with no free blocks to hand out.
AIUI most if not all of the RIRs have announced special measures to be applied as their stocks of IPv4 addresses run low. APNIC and RIPE are now deep into special measures with new allocations only available in special circumstances (a small block for a new LIR, blocks for use by internet exchange points). ARIN are less deep into special measures with regular allocations still available though under tigher rules than before.
IPv6 is a cleaner soloution in the long term but it requires cooperation from basically everyone involved in the internet industry (software vendors, OS vendors, hardware vendors, carriers) to take it from the "cool toy" stage to the "replacement for IPv4" stage. To go v6 only the client software has to support it, the OS on the end device has to support it, the home gateway has to support it, the ISP has to support it and either a horrible translation mechanism needs to be deployed by the ISP or the server has to support it.
NAT is messier but solves the immediate problem for your organisation without requiring cooperation from anyone else.
Afaict cellular networks have been doing this for years but at least round here fixed line providers are only just starting to trial ISP level nat.
Why should an increase in the money supply devalue the dollar in my pocket?
Producers want to get as much money as they can for the goods and services they produce. If at the current price point demand exceeds supply then prices will rise to balance supply with demand. Putting more money into the system increases demand at a given price point and so drives prices up.
The alternative to this is for the governent to force producers to sell at an artificially low price but that will lead to shortages.
Agree completely, but the problem is that to an outside observer it is impossible to know how many comparisons were actually done.
And even if the researchers are being honest about the number of comparisons THEY did it is very likely that multiple people will be independently working on the same problem. If all of those people individually only publish positive results then you get the same problem.