The AOL comparison is germane - they also built their own international network and achieved no-cost ("tier 1") peering status (where their interconnects with major services providers like Sprint, MCI and AT&T do not attract per megabit charges).
AOL started to do this around 2000, and were transit-free around 2003 IIRC. Google is probably thinking it's worth their while doing this, whether or not they move into the consumer internet market, especially given the low cost of bandwidth in the US right now.
There's more than one cable system linking US with Europe, it just happens that several carriers (Above.Net being one) only have capacity through TAT-14.
Other carriers have working circuits on TAT-14 and another link (e.g. Apollo, Tyco, AC-1, Gemini) and may have some degraded service (depending on whether their transatlantic links are less than twice the size of their peak demand). FranceTelecom OpenTransit is an example of one of them.
Interestingly, not many EU ISPs use TAT-14 North route, since it has a propagation delay of around 110ms (which is 40ms or so more than TAT-14 South from the UK and more than most other transatlantic cables)
Most ISPs in Europe that I can see are fine. Certainly the big international transit ISPs (Sprint, L3, C&W, MCI et al) aren't showing any more trouble than normal.
At the risk of being accused of Karma whoring, This page and This wired article from the late 90s are are good summary and a great story about undersea cables, respectively, despite being a little out of date.
There are two components of the cost of assigning IP addresses.
One is the administrative cost of someone picking your address, writing in the database it belongs to you and giving it to you.
The other is the cost of acquiring the addresses, which outside the ARIN region is pretty independant of the block size (RIPE/APNIC assign more blocks as you fill the old one, and don't charge outside the annual fee).
IPv6 makes the latter happen far less frequently, but does nothing for the former. The large ISP I work for assigns addresses based on need charging a fixed amount that is not based on the size - this model is fairly common in Europe.
Disclaimer/boast: I work for one of the top-5 US ISPs. (the one that used to own optus).
Most of our big broadband provider customers are on pay-per megabit, not flat rate. Besides, a flat rate circuit is only flat-rate until you need a larger circuit, at which point it costs more money.
Routers and circuits to reach peering points are also not free, although the cost of peering is usually far lower than the cost of transit.
Don't make the mistake of saying that because the marginal cost of bandwidth is zero this makes the average cost zero - this is only true till you have to upgrade.
It's also increasingly less true (although it once was) to say the US providers don't bear any of the costs for international connectivity. An increasing number of US majors ISPs (e.g. AT&T, Sprint, UUnet, C&W, Qwest) have extensive European and Asia PAC networks, and in some cases, regional ISPs like NTT or France Telecom get settlement-free peering on one or both sides of the water.
The cost per megabit of bandwidth is falling rapidly (up to 3% a month on some routes), but it isn't free - if broadband providers think they will make more money by charging per megabit, taking into account competition, cost of measuring the usage, cost of billing & billing disputes (should you pay when you get DoS'd?), they will do.
The cost of billing, measuring usage et al are relatively similar across countries - the cost of bandwidth is MUCH higher in AU than US, so this stuff looks more attractive. Of course, capping protects your network from the 1% of abusers for a lower price, so it looks even more attractive.
Any arguments that bandwidth providers should be legally forced to operate all-you-can-eat bandwidth restaurants for gluttons are unlikely to fly at any regulators, even if the company has a monopoly.
Why ISPs aren't using something like this is reasonably simple - the router hardware, and the metering systems necessary to track usage and implement this are very expensive.
COPS, diffserv and other priority mechanisms are somewhat available but not widely deployed on ISP/service provider networks.
I'm certainly not convinced that replacing my router hardware with systems that can do this (and hiring the people that can support it) are financially viable yet.
Until then, flat rate with overuse detection and prevention remains among the most cost effective methods of providing consumer broadband.
Of course, widespread overuse might well encourage people to deploy these types of solutions on their network - so maybe they really are driving the technology economy.
I called SkyInet.net in Manila (whose servers were being used to distribute the second part of the virus, payload still unknown) about 10AM CET (8AM UTC) this morning. I guess it was about 8PM there, and the lady on 24hr support sounded VERY harasssed. Still, nice to get a proactive response out of them quickly - they deleted the files within 1/2 an hour.
One of the goals of Internet 2 that is useful (IMO, anyway) is a test platform for very high-speed applications.
It's all very well developing some very shiny technologies that 'should work' when they have enough WAN bandwidth, but it's another thing entirely to do real-world testing on them.
Things like developing the protocols to send HDTV over the network need a real live network like this (to test human factors in development as much as technical), so it's really not just "to take down roadblocks from the first Internet, like heavy traffic and slow interfaces, and speed things up for college researchers sharing information".
Given what happened to the shares of the AT&T companies after they were broken up, is there any reason to believe that he _wouldn't_ want the company broken up into four pieces?
From my experience, no matter how good you are, people just aren't going to pay a 20yr old a full consultant rate, no matter how good he is technically, or how much experience he has. You don't say where in the world you are, but if you're not in a major US IT city, $20 isn't all that bad. That said, you'll probably snag a better paid job just because of the extra experience you got while working at your current job, even if it was only for a few months. Enjoy it while it lasts, I'm sure the world will wake up soon, and coding will become a clerical level job.
£14k for a CS/IT graduate from a decent UK university is very low, regardless of where in the country you work. People from my CS course who were looking for money when they graduated last year got £22k+ in London, and some of them got nearly £30k. The strangest thing to me was to see how tolerant the vast majority of them were of awful salaries - it seems like those people who actually bothered to hunt for a decent salary got _far_ more than those who took traditional (i.e. defense industry, IT consultancy) jobs. Crap banking contractors get £30/hr for their skills (which equates to not far off £100k). I've seen DECENT banking contractors at over £100/hr.
Slashdot Mirror
So do Johnson and Allchin have 'president' and 'co-president' nametags then? Whose is whose?
The AOL comparison is germane - they also built their own international network and achieved no-cost ("tier 1") peering status (where their interconnects with major services providers like Sprint, MCI and AT&T do not attract per megabit charges).
AOL started to do this around 2000, and were transit-free around 2003 IIRC. Google is probably thinking it's worth their while doing this, whether or not they move into the consumer internet market, especially given the low cost of bandwidth in the US right now.
There's more than one cable system linking US with Europe, it just happens that several carriers (Above.Net being one) only have capacity through TAT-14.
Other carriers have working circuits on TAT-14 and another link (e.g. Apollo, Tyco, AC-1, Gemini) and may have some degraded service (depending on whether their transatlantic links are less than twice the size of their peak demand). FranceTelecom OpenTransit is an example of one of them.
Interestingly, not many EU ISPs use TAT-14 North route, since it has a propagation delay of around 110ms (which is 40ms or so more than TAT-14 South from the UK and more than most other transatlantic cables)
Most ISPs in Europe that I can see are fine. Certainly the big international transit ISPs (Sprint, L3, C&W, MCI et al) aren't showing any more trouble than normal.
At the risk of being accused of Karma whoring, This page and This wired article from the late 90s are are good summary and a great story about undersea cables, respectively, despite being a little out of date.
There are two components of the cost of assigning IP addresses.
One is the administrative cost of someone picking your address, writing in the database it belongs to you and giving it to you.
The other is the cost of acquiring the addresses, which outside the ARIN region is pretty independant of the block size (RIPE/APNIC assign more blocks as you fill the old one, and don't charge outside the annual fee).
IPv6 makes the latter happen far less frequently, but does nothing for the former.
The large ISP I work for assigns addresses based on need charging a fixed amount that is not based on the size - this model is fairly common in Europe.
Disclaimer/boast: I work for one of the top-5 US ISPs. (the one that used to own optus).
Most of our big broadband provider customers are on pay-per megabit, not flat rate.
Besides, a flat rate circuit is only flat-rate until you need a larger circuit, at which point it costs more money.
Routers and circuits to reach peering points are also not free, although the cost of peering is usually far lower than the cost of transit.
Don't make the mistake of saying that because the marginal cost of bandwidth is zero this makes the average cost zero - this is only true till you have to upgrade.
It's also increasingly less true (although it once was) to say the US providers don't bear any of the costs for international connectivity. An increasing number of US majors ISPs (e.g. AT&T, Sprint, UUnet, C&W, Qwest) have extensive European and Asia PAC networks, and in some cases, regional ISPs like NTT or France Telecom get settlement-free peering on one or both sides of the water.
The cost per megabit of bandwidth is falling rapidly (up to 3% a month on some routes), but it isn't free - if broadband providers think they will make more money by charging per megabit, taking into account competition, cost of measuring the usage, cost of billing & billing disputes (should you pay when you get DoS'd?), they will do.
The cost of billing, measuring usage et al are relatively similar across countries - the cost of bandwidth is MUCH higher in AU than US, so this stuff looks more attractive. Of course, capping protects your network from the 1% of abusers for a lower price, so it looks even more attractive.
Any arguments that bandwidth providers should be legally forced to operate all-you-can-eat bandwidth restaurants for gluttons are unlikely to fly at any regulators, even if the company has a monopoly.
Why ISPs aren't using something like this is reasonably simple - the router hardware, and the metering systems necessary to track usage and implement this are very expensive.
COPS, diffserv and other priority mechanisms are somewhat available but not widely deployed on ISP/service provider networks.
I'm certainly not convinced that replacing my router hardware with systems that can do this (and hiring the people that can support it) are financially viable yet.
Until then, flat rate with overuse detection and prevention remains among the most cost effective methods of providing consumer broadband.
Of course, widespread overuse might well encourage people to deploy these types of solutions on their network - so maybe they really are driving the technology economy.
I called SkyInet.net in Manila (whose servers were being used to distribute the second part of the virus, payload still unknown) about 10AM CET (8AM UTC) this morning.
I guess it was about 8PM there, and the lady on 24hr support sounded VERY harasssed. Still, nice to get a proactive response out of them quickly - they deleted the files within 1/2 an hour.
http://x42.com/urlcalc/ looks a lot like prior art to me.
One of the goals of Internet 2 that is useful (IMO, anyway) is a test platform for very high-speed applications.
It's all very well developing some very shiny technologies that 'should work' when they have enough WAN bandwidth, but it's another thing entirely to do real-world testing on them.
Things like developing the protocols to send HDTV over the network need a real live network like this (to test human factors in development as much as technical), so it's really not just "to take down roadblocks from the first Internet, like heavy traffic and slow interfaces, and speed things up for college researchers sharing information".
Given what happened to the shares of the AT&T companies after they were broken up, is there any reason to believe that he _wouldn't_ want the company broken up into four pieces?
From my experience, no matter how good you are, people just aren't going to pay a 20yr old a full consultant rate, no matter how good he is technically, or how much experience he has. You don't say where in the world you are, but if you're not in a major US IT city, $20 isn't all that bad. That said, you'll probably snag a better paid job just because of the extra experience you got while working at your current job, even if it was only for a few months. Enjoy it while it lasts, I'm sure the world will wake up soon, and coding will become a clerical level job.
£14k for a CS/IT graduate from a decent UK university is very low, regardless of where in the country you work. People from my CS course who were looking for money when they graduated last year got £22k+ in London, and some of them got nearly £30k. The strangest thing to me was to see how tolerant the vast majority of them were of awful salaries - it seems like those people who actually bothered to hunt for a decent salary got _far_ more than those who took traditional (i.e. defense industry, IT consultancy) jobs. Crap banking contractors get £30/hr for their skills (which equates to not far off £100k). I've seen DECENT banking contractors at over £100/hr.