I can't really see how it costs the utility more to NOT deliver power. I can see how they have fixed costs that are the same either way.
Plus, if they cared about not ramping up and down generators then the contract wouldn't just be for a total quantity used, but would also stipulate some kind of continuous draw. I can't see how they got out of ramping up and down generators if MS just burned through an extra $70k of power in a day. Every day before then they weren't using quite as much, and then on the last day instead of using the same amount as every day prior they pulled out the stops and put a huge load on the grid. That means peaker plants and the works all kicking off. How does it cost MORE if they don't just keep doing what they did every day prior?
I'll take you at your word (I'm a chemist, not a particle physicist, though I do know enough about Boltzman distributions that I could probably work it out).
Agreed that the velocity distribution might not be right - initially this stuff all started out as a mostly uniform cloud of gas with particles falling towards the nearest clump of mass, with particles that are farther picking up more speed - should be possible with some assumptions and calculus to work out what the distribution is assuming no collisions at all, and then it would move from that to a Boltzman distribution as collisions happen. That said, I'm sure there is SOME interaction among particles. The mean free path could be rather long though - and all of the particles are moving fast so collisions will shoot off some X-Rays and then everything is still moving along.
Again, I'm not a physicist, though I might be if I had the time.:) One lifetime just isn't enough.
It isn't irrelevant, but there are matters of degree here. The Pi is a relatively open platform that would be a great educational tool.
Would I love to have an even MORE open platform - absolutely!
Right now I am typing this on an FOSS browser on an FOSS operating system. However, there are a few bits here and there that are proprietary, as is much of the hardware and firmware in my system. I'd prefer that stuff to be FOSS, but since no such computer is currently available, I make do with what I have.
Even RMS lives with this kind of compromise - since to do otherwise is to deny himself the use of a computer entirely.
It likely only got hotter with time - as it fell towards the galaxy and picked up speed. In order to lose temperature the gas molecules have to actually interact with something. There isn't anything for them to interact with - these things are flying through intergalactic space basically flying around the galaxy.
Put hot water in a thermos and it stays hot for a while. Put hot water in intergalactic space and it stays hot for much longer. However, the water molecules still interact and release photons out into space radiatively. Ionized hydrogen is just loose protons flying through space - they don't just emit photons, and neither do free electrons as far as I'm aware. When they occasionally collide then you get photons released as x rays, which is what gets imaged.
I think most of the mass of the universe actually exists as these clouds of gas surrounding superclusters.
I don't think it has been this hot for the entire life of the universe - it actually hot hotter with time most likely. This was likely gas found in intergalactic space that fell gravitationally towards the milky way. After falling for billions of years it is moving really fast. However, the gas is so sparse that there really aren't any collisions to speak of. Sure, if a particle hits a star or planet or something that will stop it, but chances are this stuff is hitting our atmosphere all the time, but for every particle that hits the earth trillions pass all around us flying through the solar system and the space beyond at incredible velocity.
I wouldn't even be concerned as much about taking down prod (the dev would likely get it back up quickly). I'd be more concerned that you have no idea what is on your prod server any longer, and when you test the next upgrade you test it against what you thought was in prod, and THAT takes down prod after the tests all pass. Or you need a new server and now one of your servers is different from all of the others. Or that dev gets let go and nobody can figure out why the server isn't working as expected.
At work we actually had to pay to redevelop an enhancement because apparently we never obtained the source code to a component of our software that was running in production. No doubt the source was emailed back and forth umpteen years ago, but it never got checked in, and somebody probably just emailed an executable to include in the install package.
For this reason it isn't even enough to have somebody else do the install - you actually need good practices for how you make install packages in the first place and how your installers work. A process held together with duct tape isn't any better just because somebody else runs it.
What we do at my company is allow the developers to work with the project managers and deploy their applications out to a test environment for client facing review and acceptance as often as they like...snip...Once it's been decided that a production deployment is ready then the actual task of deploying the application is assigned to whoever wants it (usually the team lead) since the process of deploying to production is identical to deploying to test in our environments.
Uh, you hope the process is the same. How do you know the first team followed the same procedures as the second?
By all means deploy all the prototypes you want, but it isn't "Testing" unless the exact same process is used to install and control the test servers as will be used for the production servers.
My thoughts exactly. Unless the development team wants to end up doing tier 1 or tier 2 production support as well, it's in their best interests to create tools and documentation to deploy the application.
No, unless the people who PAY the development team wants them to do production support for all eternity and be up the creek whenever somebody leaves it is in THEIR best interests to deploy the application on the basis of documentation only.
Even that isn't enough - the deployment process needs to be rigorous and standardized, so that when you lose 1000 servers you're not digging out 1000 separate binders of install scripts done 47 different ways. Instead you're just firing off your automated install scripts en masse.
You don't need to have the developers watch the configuration team install the software. That is because it is all automated, and they install it on a production-equivalent test system, and you then get to test everything they installed. If the install is bad you wipe it and start all over again. That isn't a big deal since your tests are automated too.
If it installs fine on test it will install fine on production, since the only thing done differently is to point the script someplace else.
If your install procedure isn't scripted to this degree, then you really have no idea what you have in production, and if something happens to it whatever you do to fix it will not have it end up the same way, since you have no idea what it was to begin with.
Too many companies treat servers like sculptures crafted over weeks by a master craftsman. They should be injection molded pieces of plastic whose tolerances are way higher than they need to be.
Agreed. Alas, I suspect that at most big corporations things are fairly broken.
We run into these kinds of problems at work. I have no confidence at all that we have any production-equivalent environments. We also use incremental install processes such that building a new server is like replaying the entire development history of the project (maybe skipping steps that we hope are safe to skip).
Companies just aren't willing to pay for GOOD configuration management teams. If everything were checked into a good source management system, and server builds were fully automated, and servers could be cloned from saved images, then you'd know that if you ever had a disaster or needed an extra server it would be a 15 minute operation.
Most companies probably don't realize it, but if they had a real disaster they'd be really up the creek.
Yeah, I commented elsewhere that my Apple data was apparently a bit dated (don't really follow them that closely). Their earlier models had much longer upgrade support than their newer ones. So, Android and Apple are converging, but Android still has a way to go, and it really is even a competition only for the Nexus models.
1-1.5 years on a device you replace every two years sounds nice, unless you don't buy your phone the day it comes out. Most people end up buying phones that already have considerable mileage. For example, I'd really prefer to upgrade now, but I'm delaying to get a new Nexus, since the Galaxy Nexus already has quite a bit of time on the odometer.
I didn't say that it wasn't. I said that it got its last update about six months after it was discontinued, and it may or may not get another. It is current, but that only means that it was supported for about 1.5 years after introduction.
Yes. First, I was generally rounding - 1.5, 1.75, whatever.
You also forgot the ADP. That one was abandoned about a year after it came out.
I think phones should receive updates for two years from the day that they're LAST sold, so that means the Galaxy Nexus is at 0 days and counting, the Nexus S is at about 9 months, and the Nexus One got its last update about 9 months after it sold, but that was a minor release (Gingerbread is older than Sept 2011 I believe). So, haggling over a few months is missing the point.
Agreed. When I go to invest in a kickstarter project the only thing I have to go on is reputation and the promise of a return. That return is usually VERY limited. So, reputation is almost everything.
Kickstarter is basically useless for somebody with a good idea but a week resume. A venture capitalist can discuss plans in detail and put somebody on the inside to see how money is being spent and give you the money in drabs as you make progress, and they stand to reap a ton of rewards if all goes well. Note of those things apply to Joe Smith who wants to donate $20 to something on the internet.
Based on that article it sounds like this material isn't really ideal for making high-power magnets. I know more about the practical use of these magnets than their materials, so I'll have to defer to a physicist to comment on that. Superconductors don't like strong magnetic fields in general (Google the Josephson Junction), so a higher critical temperature alone doesn't make a superconductor more useful when building magnets.
In fact, I think modern magnets incorporate copper in the wire perhaps at the cost of superconductivity simply so that the wires don't completely melt if the magnet quenches. These magnets can cost millions of dollars to build, so accident-tolerance is a useful property. Of course, going back to the topic of this article, if the magnet does quench there goes a small fortune in liquid helium...
Superconductor research is advancing rapidly, with new advances leading, step by step, towards room-temperature superconductors.
Sure, but it is foolish to squander what we have and need RIGHT NOW in the hope that we won't need it later. Better to conserve now, and then just have extra helium to spare later if it isn't needed.
My understanding is that the higher-temperature superconductors are not capable of withstanding the large magnetic fields required in an MRI. While most people understand that superconductivity is sensitive to high temperatures, it is also sensitive to the presence of magnetic fields (obviously an important consideration in a magnet!).
Superconducting electromagnets have been around for decades now and are somewhat ubiquitous. However, the designs haven't really changed in any revolutionary sense. They might change in the future, but it seems more wise to plan on incremental improvement.
If a guy comes in with an earache and needs antibiotics that is probably a 15 minute office visit, and reimbursed cheaply.
If a guy comes in for his quarterly visit and he is a diabetic and has had heart surgery, then it is probably worth spending more so that the doctor can take an hour to make sure everything is being taken care of to avoid more costly problems later. However, there is nothing to stop the doctor from charging insurance for the hour, but wrapping the whole thing up in 30 minutes.
Fair points - my iOS update data is out of date for sure. They updated the 3G about 5 months after it was discontinued, and that was it (though that was 2.3 years after introduction). The Nexus S got its most recent update about six months after it was discontinued - I'm skeptical that it will get another, but it might. Granted, it was also on the market for a lot less time than most iPhones are.
Seems that iOS and the nexus phones are converging on update support. Now if only the rest of the Android models would play ball...
Is helium really the only way to cool MRI machines? I'll bet that, if prices rise enough, somebody will invent a new method not involving helium.
It is, at least right now. You're talking 4K - liquid helium is REALLY cold. You need superconducting magnets that can operate in multiple Tesla magnetic fields. If you read up on superconductivity you'll find that magnetic fields disable superconductors - so you need temperatures as low as you can get them to counteract that.
Trust me, the helium is already plenty expensive at the scales they operate on. If somebody had another way to cool these things down that was remotely practical it would be adopted all over the place.
That said, they already do try to recycle it. The problem is the recycled stuff tends to get used in balloons since it is cheaper to resell for that use than to reprocess it, largely because the cost of helium is artificially low.
Yup, but it could be recycled for additional medical use instead of vented into the atmosphere where it ends up floating into space...
The natural sources of helium are somewhere around 2% pure. Party balloons are something like 90%, and medical use is over 99%. If you can make 99% helium from 2% helium, you can certainly make it from 90% helium.
I think we need to reevaluate what's wrong and focus research towards re-engineering MRI machines or use different mediums to cool these differently.
Even at today's artificially low rates, helium is still QUITE expensive for things like MRIs, as is the associated maintenance/downtime/etc - and the small risk of destroying the million dollar magnet every time you fuss with it.
If somebody could think of a better way to cool them, trust me, they would.
Do you have any concept of how cold liquid Helium is? At atmospheric pressure we're taking around 4K - four degrees above absolute zero. In some applications it is operated under reduced pressure to get it down to super-critical temperatures lower than that. If you stuck an MRI magnet in deep space it would be warmed up by cosmic background radiation to a higher temperature than that (though you probably could practically operate an MRI at the ambient temperature of DEEP space, it would take an age for the magnet to cool down on its own and you'd need huge radiators to disperse the heat generated by the procedure to keep it that way - plus it isn't terribly convenient for patients to have to travel past the Ort cloud for their test).
Yes, but better to have the price rise now to a reasonable level that encourages conservation, than sell it all off at subsidized prices and then once it runs out we're paying through the noses to retrofit everything in a panic as half the hospital MRIs shut down in the meantime.
If prices are expected to go up it only makes sense to gradually work them in than to force everybody to go cold turkey.
Right now I'm sure there is some recycling but with the prices so cheap it isn't that high a priority - cheaper to just buy more new gas, er liquid.
I think you're missing the point. Sure, the article is talking about deployment rates, but the bigger issue is that there is even an update to be deployed. With iOS you get updates for years after getting your device, period. With Android you're lucky to get an update while the thing is still being sold in stores. The actual deployments are reasonably fast (though generally slower with Android - which I think is sane from a QA perspective). They just never happen unless you buy a Nexus device, and even then only for 1.5 years after they are FIRST sold.
If you bought a brand new Nexus phone two years ago, you wouldn't even be running ICS, let alone Jelly Bean. A brand new Nexus phone two years ago would have been the Nexus One. If you bought a brand new Nexus phone a year before that you wouldn't have gotten a single software update - that would have been the ADP and it never even got Eclair.
Slashdot Mirror
I can't really see how it costs the utility more to NOT deliver power. I can see how they have fixed costs that are the same either way.
Plus, if they cared about not ramping up and down generators then the contract wouldn't just be for a total quantity used, but would also stipulate some kind of continuous draw. I can't see how they got out of ramping up and down generators if MS just burned through an extra $70k of power in a day. Every day before then they weren't using quite as much, and then on the last day instead of using the same amount as every day prior they pulled out the stops and put a huge load on the grid. That means peaker plants and the works all kicking off. How does it cost MORE if they don't just keep doing what they did every day prior?
Yeah, but they're WAY more exciting than a AA battery if you short the leads. :)
I'll take you at your word (I'm a chemist, not a particle physicist, though I do know enough about Boltzman distributions that I could probably work it out).
Agreed that the velocity distribution might not be right - initially this stuff all started out as a mostly uniform cloud of gas with particles falling towards the nearest clump of mass, with particles that are farther picking up more speed - should be possible with some assumptions and calculus to work out what the distribution is assuming no collisions at all, and then it would move from that to a Boltzman distribution as collisions happen. That said, I'm sure there is SOME interaction among particles. The mean free path could be rather long though - and all of the particles are moving fast so collisions will shoot off some X-Rays and then everything is still moving along.
Again, I'm not a physicist, though I might be if I had the time. :) One lifetime just isn't enough.
It isn't irrelevant, but there are matters of degree here. The Pi is a relatively open platform that would be a great educational tool.
Would I love to have an even MORE open platform - absolutely!
Right now I am typing this on an FOSS browser on an FOSS operating system. However, there are a few bits here and there that are proprietary, as is much of the hardware and firmware in my system. I'd prefer that stuff to be FOSS, but since no such computer is currently available, I make do with what I have.
Even RMS lives with this kind of compromise - since to do otherwise is to deny himself the use of a computer entirely.
It likely only got hotter with time - as it fell towards the galaxy and picked up speed. In order to lose temperature the gas molecules have to actually interact with something. There isn't anything for them to interact with - these things are flying through intergalactic space basically flying around the galaxy.
Put hot water in a thermos and it stays hot for a while. Put hot water in intergalactic space and it stays hot for much longer. However, the water molecules still interact and release photons out into space radiatively. Ionized hydrogen is just loose protons flying through space - they don't just emit photons, and neither do free electrons as far as I'm aware. When they occasionally collide then you get photons released as x rays, which is what gets imaged.
I think most of the mass of the universe actually exists as these clouds of gas surrounding superclusters.
I don't think it has been this hot for the entire life of the universe - it actually hot hotter with time most likely. This was likely gas found in intergalactic space that fell gravitationally towards the milky way. After falling for billions of years it is moving really fast. However, the gas is so sparse that there really aren't any collisions to speak of. Sure, if a particle hits a star or planet or something that will stop it, but chances are this stuff is hitting our atmosphere all the time, but for every particle that hits the earth trillions pass all around us flying through the solar system and the space beyond at incredible velocity.
I wouldn't even be concerned as much about taking down prod (the dev would likely get it back up quickly). I'd be more concerned that you have no idea what is on your prod server any longer, and when you test the next upgrade you test it against what you thought was in prod, and THAT takes down prod after the tests all pass. Or you need a new server and now one of your servers is different from all of the others. Or that dev gets let go and nobody can figure out why the server isn't working as expected.
At work we actually had to pay to redevelop an enhancement because apparently we never obtained the source code to a component of our software that was running in production. No doubt the source was emailed back and forth umpteen years ago, but it never got checked in, and somebody probably just emailed an executable to include in the install package.
For this reason it isn't even enough to have somebody else do the install - you actually need good practices for how you make install packages in the first place and how your installers work. A process held together with duct tape isn't any better just because somebody else runs it.
What we do at my company is allow the developers to work with the project managers and deploy their applications out to a test environment for client facing review and acceptance as often as they like...snip...Once it's been decided that a production deployment is ready then the actual task of deploying the application is assigned to whoever wants it (usually the team lead) since the process of deploying to production is identical to deploying to test in our environments.
Uh, you hope the process is the same. How do you know the first team followed the same procedures as the second?
By all means deploy all the prototypes you want, but it isn't "Testing" unless the exact same process is used to install and control the test servers as will be used for the production servers.
My thoughts exactly. Unless the development team wants to end up doing tier 1 or tier 2 production support as well, it's in their best interests to create tools and documentation to deploy the application.
No, unless the people who PAY the development team wants them to do production support for all eternity and be up the creek whenever somebody leaves it is in THEIR best interests to deploy the application on the basis of documentation only.
Even that isn't enough - the deployment process needs to be rigorous and standardized, so that when you lose 1000 servers you're not digging out 1000 separate binders of install scripts done 47 different ways. Instead you're just firing off your automated install scripts en masse.
You don't need to have the developers watch the configuration team install the software. That is because it is all automated, and they install it on a production-equivalent test system, and you then get to test everything they installed. If the install is bad you wipe it and start all over again. That isn't a big deal since your tests are automated too.
If it installs fine on test it will install fine on production, since the only thing done differently is to point the script someplace else.
If your install procedure isn't scripted to this degree, then you really have no idea what you have in production, and if something happens to it whatever you do to fix it will not have it end up the same way, since you have no idea what it was to begin with.
Too many companies treat servers like sculptures crafted over weeks by a master craftsman. They should be injection molded pieces of plastic whose tolerances are way higher than they need to be.
Agreed. Alas, I suspect that at most big corporations things are fairly broken.
We run into these kinds of problems at work. I have no confidence at all that we have any production-equivalent environments. We also use incremental install processes such that building a new server is like replaying the entire development history of the project (maybe skipping steps that we hope are safe to skip).
Companies just aren't willing to pay for GOOD configuration management teams. If everything were checked into a good source management system, and server builds were fully automated, and servers could be cloned from saved images, then you'd know that if you ever had a disaster or needed an extra server it would be a 15 minute operation.
Most companies probably don't realize it, but if they had a real disaster they'd be really up the creek.
Yeah, I commented elsewhere that my Apple data was apparently a bit dated (don't really follow them that closely). Their earlier models had much longer upgrade support than their newer ones. So, Android and Apple are converging, but Android still has a way to go, and it really is even a competition only for the Nexus models.
1-1.5 years on a device you replace every two years sounds nice, unless you don't buy your phone the day it comes out. Most people end up buying phones that already have considerable mileage. For example, I'd really prefer to upgrade now, but I'm delaying to get a new Nexus, since the Galaxy Nexus already has quite a bit of time on the odometer.
I didn't say that it wasn't. I said that it got its last update about six months after it was discontinued, and it may or may not get another. It is current, but that only means that it was supported for about 1.5 years after introduction.
Yes. First, I was generally rounding - 1.5, 1.75, whatever.
You also forgot the ADP. That one was abandoned about a year after it came out.
I think phones should receive updates for two years from the day that they're LAST sold, so that means the Galaxy Nexus is at 0 days and counting, the Nexus S is at about 9 months, and the Nexus One got its last update about 9 months after it sold, but that was a minor release (Gingerbread is older than Sept 2011 I believe). So, haggling over a few months is missing the point.
Agreed. When I go to invest in a kickstarter project the only thing I have to go on is reputation and the promise of a return. That return is usually VERY limited. So, reputation is almost everything.
Kickstarter is basically useless for somebody with a good idea but a week resume. A venture capitalist can discuss plans in detail and put somebody on the inside to see how money is being spent and give you the money in drabs as you make progress, and they stand to reap a ton of rewards if all goes well. Note of those things apply to Joe Smith who wants to donate $20 to something on the internet.
Not sure that was true "in the first place" - I can't imagine there was that much demand for helium back in the 50s.
Government regulation of markets isn't automatically a bad thing. However, it certainly is the case that it can lead to harm down the road.
Based on that article it sounds like this material isn't really ideal for making high-power magnets. I know more about the practical use of these magnets than their materials, so I'll have to defer to a physicist to comment on that. Superconductors don't like strong magnetic fields in general (Google the Josephson Junction), so a higher critical temperature alone doesn't make a superconductor more useful when building magnets.
In fact, I think modern magnets incorporate copper in the wire perhaps at the cost of superconductivity simply so that the wires don't completely melt if the magnet quenches. These magnets can cost millions of dollars to build, so accident-tolerance is a useful property. Of course, going back to the topic of this article, if the magnet does quench there goes a small fortune in liquid helium...
Superconductor research is advancing rapidly, with new advances leading, step by step, towards room-temperature superconductors.
Sure, but it is foolish to squander what we have and need RIGHT NOW in the hope that we won't need it later. Better to conserve now, and then just have extra helium to spare later if it isn't needed.
My understanding is that the higher-temperature superconductors are not capable of withstanding the large magnetic fields required in an MRI. While most people understand that superconductivity is sensitive to high temperatures, it is also sensitive to the presence of magnetic fields (obviously an important consideration in a magnet!).
Superconducting electromagnets have been around for decades now and are somewhat ubiquitous. However, the designs haven't really changed in any revolutionary sense. They might change in the future, but it seems more wise to plan on incremental improvement.
There is probably some truth to this.
If a guy comes in with an earache and needs antibiotics that is probably a 15 minute office visit, and reimbursed cheaply.
If a guy comes in for his quarterly visit and he is a diabetic and has had heart surgery, then it is probably worth spending more so that the doctor can take an hour to make sure everything is being taken care of to avoid more costly problems later. However, there is nothing to stop the doctor from charging insurance for the hour, but wrapping the whole thing up in 30 minutes.
Fair points - my iOS update data is out of date for sure. They updated the 3G about 5 months after it was discontinued, and that was it (though that was 2.3 years after introduction). The Nexus S got its most recent update about six months after it was discontinued - I'm skeptical that it will get another, but it might. Granted, it was also on the market for a lot less time than most iPhones are.
Seems that iOS and the nexus phones are converging on update support. Now if only the rest of the Android models would play ball...
Is helium really the only way to cool MRI machines? I'll bet that, if prices rise enough, somebody will invent a new method not involving helium.
It is, at least right now. You're talking 4K - liquid helium is REALLY cold. You need superconducting magnets that can operate in multiple Tesla magnetic fields. If you read up on superconductivity you'll find that magnetic fields disable superconductors - so you need temperatures as low as you can get them to counteract that.
Trust me, the helium is already plenty expensive at the scales they operate on. If somebody had another way to cool these things down that was remotely practical it would be adopted all over the place.
That said, they already do try to recycle it. The problem is the recycled stuff tends to get used in balloons since it is cheaper to resell for that use than to reprocess it, largely because the cost of helium is artificially low.
Yup, but it could be recycled for additional medical use instead of vented into the atmosphere where it ends up floating into space...
The natural sources of helium are somewhere around 2% pure. Party balloons are something like 90%, and medical use is over 99%. If you can make 99% helium from 2% helium, you can certainly make it from 90% helium.
I think we need to reevaluate what's wrong and focus research towards re-engineering MRI machines or use different mediums to cool these differently.
Even at today's artificially low rates, helium is still QUITE expensive for things like MRIs, as is the associated maintenance/downtime/etc - and the small risk of destroying the million dollar magnet every time you fuss with it.
If somebody could think of a better way to cool them, trust me, they would.
Do you have any concept of how cold liquid Helium is? At atmospheric pressure we're taking around 4K - four degrees above absolute zero. In some applications it is operated under reduced pressure to get it down to super-critical temperatures lower than that. If you stuck an MRI magnet in deep space it would be warmed up by cosmic background radiation to a higher temperature than that (though you probably could practically operate an MRI at the ambient temperature of DEEP space, it would take an age for the magnet to cool down on its own and you'd need huge radiators to disperse the heat generated by the procedure to keep it that way - plus it isn't terribly convenient for patients to have to travel past the Ort cloud for their test).
Yes, but better to have the price rise now to a reasonable level that encourages conservation, than sell it all off at subsidized prices and then once it runs out we're paying through the noses to retrofit everything in a panic as half the hospital MRIs shut down in the meantime.
If prices are expected to go up it only makes sense to gradually work them in than to force everybody to go cold turkey.
Right now I'm sure there is some recycling but with the prices so cheap it isn't that high a priority - cheaper to just buy more new gas, er liquid.
I think you're missing the point. Sure, the article is talking about deployment rates, but the bigger issue is that there is even an update to be deployed. With iOS you get updates for years after getting your device, period. With Android you're lucky to get an update while the thing is still being sold in stores. The actual deployments are reasonably fast (though generally slower with Android - which I think is sane from a QA perspective). They just never happen unless you buy a Nexus device, and even then only for 1.5 years after they are FIRST sold.
If you bought a brand new Nexus phone two years ago, you wouldn't even be running ICS, let alone Jelly Bean. A brand new Nexus phone two years ago would have been the Nexus One. If you bought a brand new Nexus phone a year before that you wouldn't have gotten a single software update - that would have been the ADP and it never even got Eclair.