"300bps acoustic coupler using a Z80 based computer with 16k RAM and a tape recorder as secondary storage device."
You had the 16K expansion pack? No fair! I had the same beast, but with only 1K of RAM, there wasn't any point in trying to load those tapes in, the apps were too big to fit in memory. All I could do was write one-page BASIC apps.
Eventually my Dad got his hands on a spare 16K piggyback cartridge and let me have his old one, so I could play MINOTAUR.
I concur. One of the reasons I switched from Mac OS 8 to Windows 95 was the multitasking being smoother. When I saw Linux in 1998 I was totally blown away and moved to that (I really wanted to go to BeOS, but it was obviously going nowhere). Somewhere over the last decade, something has gone awry with Linux. I still use it, but it -really sucks- when I kick off a low-priority backup with rsync and can no longer browse the web, play audio, or even use the command line. Waiting over a minute for 'top' to load is a travesty. Doing the same thing used to work Just Fine back with Linux 2.4.
I feel like they squeezed-out so much performance for server tasks that there are no time slices left for actually using a desktop, or there's something fundamentally wonky with the scheduling, or they're prioritizing disk cache over apps that should stay in RAM. Whatever it is, it's definitely related to I/O, and it makes using Linux really un-fun.
Of course, it used to be that X was given scheduler priority by the distributions, 'improvements' to the kernel ended that practice. Maybe the two are related?
If I recall, the K6 was pretty weak, clock-for-clock. The K6-2 was better, but still not as good as comparable Intel offerings of the time. The K6-III was pretty awesome, but only if you couldn't afford a Pentium III, and the Athlon totally dominated for quite a while, probably until the Core 2 architecture arrived.
I think what pissed me off the most about my K6-2 was that the L2 cache was on the motherboard. The K6-2+ and K6-III were much faster because they had it on-die.
There probably were folks who thought it was a bad bill, but voted for it anyway because it bought them leverage on (what they felt were) more important issues.
I'm a bit of a state house watcher, and I've heard politicians stand up and speak against bills five minutes before voting for them. Basically, if the chairman of the committee favors something and you don't, but it's going to pass anyway, you curry favor with the chairman by letting him submit the bill to the floor with 'unanimous approval', thereby increasing the chances of getting your own issue heard by the now appeased chairman in the future. In the end, you get the same result you would have if you opposed the thing, but the next time you need something, you're more likely to get it.
That or the HVAC might have been out. Our state legislature seems to decide completely on-the-fly that 'today is going to be the last day of session'. They typically suspend public hearings and pass 300 pieces of legislation that night. Why would you suspend public hearings and do 80% of your work on one coffee-fueled all-nighter? Well, the committee rooms don't have air conditioning, suits are really hot, and most of the legislature is a bit portly. Once the summer heat starts penetrating the marble walls, there's no stopping it until late October, so they 'go Nike' on democracy's ass and Just Do It.
I'd actually really like this. I think Linux kernel development is going to stay on 2.6 for another five years or more, since the product is essentially 'complete', and now it's just making incremental changes to keep up with the times.
What I'd like to see is a 'Linux 3.0' tree that starts at 'zero code'. Here are a few zany ideas from my non-compsci-trained brain:
1. Developed in a way that can be compiled as a microkernel -or- as a 'hybrid' like Apple's Mach. Full-blown computers would probably be better off running a microkernel, while portable/embedded devices would likely benefit from the 'hybrid' approach. 2. Has driver classes that are factored-out to minimize duplication of code. Just a thought: What if there was one kernel-level 'serial' data stream type that pretty much all I/O could inherit, various drivers would set parameters on the I/O that would tune it for the actual hardware being interfaced with. Serial links to devices and virtual disks in VMs might need in-order raw transfers, SATA drives might benefit from an intelligent algorithm spitting out blocks that were tuned to the physical interface, IP links and buffered I/O need to dynamically adapt their window sizes. All that logic should sit somewhere where it's accessible to other components that might want it. 3. There's got to be a way to get graphics drivers into the system, then have X programs run inside an X11 runtime a layer above that. I guess KMS and DRI already handle a lot of that, but tighter integration is a must. 4. Build the kernel to be The Best Damn Kernel Ever. If POSIX compliance is a need that prevents you from making The Best Damn Kernel Ever, offload POSIX to a runtime like Windows does. With an architecture like that, you could integrate stuff like WINE closer to the kernel, and portable/embedded devices could interface directly with the slimmer, simpler kernel interfaces. 5. Virtualization needs to be right in the core. The microkernel's 'servers' should be active keepers of the system, scanning through memory to find duplicate pages, actively profiling and optimizing I/O and scheduling, checking for hardware faults, and providing an interface for drivers to pipe data through looking for malicious code fragments. If Trojan.Whatever is traversing my network port and I have the NIC driver using the hypervisor as a virus scanner, I want the hypervisor to quarantine the traffic and send a message to other things on the system that handle it according to settings or inform me. Having that happen on the 'application' level like it is today is just asking for rootkits.
Re:For all the humor...
on
Muscle Mice
·
· Score: 1
I can still ride, but bumpy roads or jumping a curb puts me out of commission for two days. So does sprinting to catch up with my dog, or playing with the doggie pull-toy.
I just hope whatever caused my body to lose bone mass has stopped, there's no way to tell except for checking on it every few years.
Re:For all the humor...
on
Muscle Mice
·
· Score: 1
Thanks for the support. Right now it's not clear if it will stop where it is or get worse. Still, it sucks to have osteopenia when you're young.
I only found out because I got a slipped disc in my neck. My doctor asked what I did to mess things up so bad, and the answer was 'nothing at all', which triggered all sorts of tests to find out if I had nutritional or hormonal deficiencies, or even cancer. So far it looks like None of the Above, which is a good thing, I think.
For all the humor in the title, there's hopefully just as much promise.
My doctor recently told me that my twenty-something year old skeleton is basically aged like a geriatric's. The implications long term are not good. If they can make stem cells grow bone and muscle, I might not spend my fifties fighting infections in a wheelchair. It's bad enough not being able to ride a bike before I'm 30.
As a stockholder, it's in my interests that the company -not- screw up again, and that they do right by the Gulf communities.
I get a voting stake on who sits on the board.
I -take their money-.
All you do on the sidelines as a normal consumer is buy the product.
Look, I did the math and researched oil spills of this magnitude before I bought in. I don't think this spill is as bad as most environmentalists want it to be, and I'm overall quite pleased with BP's restitution and cleanup. In the end, it's going to be half a missed shrimping season and a basic cleanup of the few places where oil did make it to shore.
I bought-in like mad when the stock started going below $40/share. I made a bunch of loot, and I'm just a regular working Joe. There's a reason stocks are 'public', it means anyone can buy them.
It drives me mad when people say that investing is something for 'people with lots of power and money', It's not rocket-science, and anyone can do it. Investing has netted me about $15K of spending cash in the last two years, all on a middle-class salary.
Anyone with half a brain knows that institutional investors were dumping BP stock because their clients were angry at them and it was driving the stock down; the fiscal fundamentals would put the shares back over $40 once things stabilized. I'm still holding, though. I suspect a lot of the $20B put aside won't get used and BP will resume dividends again soon. Once that happens, there's no reason the stock won't bump up another 30%.
It shows that sex and drug use are now the activities of the most popular of the crowd, the trendsetters that everyone looks up to. When I was in high school, the weird kids were the ones smoking, drinking, and having sex the earliest. Sure, the 'cool kids' and jocks caught on later, in junior and senior year, but the only 14 year-olds having sex (that I know of) in my high school were the antisocial goths and Nine Inch Nails fans.
"How does it make any sense to show off that we're "within range" by launching right off our coast, when they could announce a test off their own coast and let us track it, without the danger of touching off nuclear Armageddon?"
I agree that this is most likely one of our own, but the logic of sneaking in close and launching in the opposite direction is a pretty safe thing to do. The first telemetry data will show that the missile isn't a threat (since it's headed in the other direction). A launch 'from home' towards us would be much more likely to inspire immediate retaliation. Plus, it wouldn't show that you can sneak up close and have your 1500-mile range missiles cover our whole territory.
I hope you're wrong. It seems to me that from 12 nautical miles out and farther, anyone could pop up and still be legally within 'international waters'.
If the Chinese popped-up 35 miles out and we did anything besides issue a friendly escort, we'd be picking a fight.
There's no reason for us to demonstrate our well-known SLBM capabilities to our own country's most populated county. I think it's highly likely that this is one of our 'peer nations' showing-off that we're entirely 'within range'.
Trust me, they saw this at Beale, where they have a huge radar system designed to see incoming ICBMs. If this wasn't an exercise or a test launch (both of which would have likely happened a little more out of the way), you won't hear much more about it. Neither the military nor the media would tell Americans if China, Russia, or India was playing around right off our beaches, the reality that we could all easily become iridescent chalky dust at the drop of a hat distracts from the main objectives: Fighting bad guys with AK-47s to secure cheap energy, and buying stuff on credit.
I think I have an idea where this might be heading...
You can fit 4 Mac Mini servers into 1U now, they just need a tray that diverts the heat from them. That's a killer opportunity right there. You get much more oomph from four Mac Minis than you do from an XServe.
http://www.apple.com/macmini/specs.html, yup, you could definitely fit these one-high, and side-by-side into a 1U tray, with room for connectors and venting. I think trays might even be deep enough to hold six of them.
X programs on Ubuntu clients and servers will still have to run in X, but X will be running inside Wayland. Think of it like running Xming inside Windows. I can still 'host' and 'join' X sessions to and from it.
I think this might actually be a good move, X is overkill for 98% of users. A smaller, slimmer, faster-moving display foundation that still allows the 'legacy' X stuff to run in a runtime would be great.
I don't think that we can determine if there actually -is- anything in his/her hand. Take a walk around a city sometime and you'll see plenty of loonies holding the sides of their heads, talking back to the voices they 'hear' inside their mind. I'm guessing schizophrenia manifested itself the same way back then as it does now.
Hell, there could have been a loud noise nearby. I make that move when an ambulance passes by.
I run VirtualBox on a 3GHz 64-bit Core 2 Linux host. Both Debian and Ubuntu have a frustrating long-term bug that makes sound choppy and unusable for Windows guests. That said, I still run it (since I'm doing 'systems' stuff and no multimedia).
My current system is actually pretty awesome:
1. The host Linux OS has a Samba share that is joined via winbind to... 2. A guest running Windows Server 2008 R2.
Clients are my other machines, and a slew of VMs running XP, Windows 7, and Linux.
I get bare-metal file server performance from Linux and a real Active Directory and WDS deployment server to test Windows stuff for work.
In my state the tax on rolling papers alone is $0.17 each. Drums of tobacco are taxed at 80% of the wholesale cost. It adds up to a very expensive habit, even to roll your own.
In a perfect world, you would call up your state rep/senator and tell them about this crazy idea:
A small group of hip technical folks sit on a committee that sets a 'standard' config once a year. The config just has to hit the bases for 'basic office machine', no advanced 3D, no huge hard drive. This is the machine that 90% of your state's desktop hardware budget will go to, and the schools and municipalities can get on board too. Put the config out to bid, and let vendors fight over the price. You'll end up getting cheaper desktops and cutting a huge amount of IT bullcrap from your state and local government.
I've been to way too many poor public schools that have 'regular office PCs' with $400 3D cards in them to trust the way we do it now. Those vendors might not be convincing when you're buying a PC for yourself, but when you're spending someone else's money, it's easy to get the big HD, expensive CPU, and top-tier video card.
I know folks who thought it was super-smart to roll their own cigarettes. In the end, they pay a huge tax on the tobacco, and a huge tax on papers, and spend a fair amount of their time trying to roll cigarettes in windy new-england weather. In the end, the pre-packaged cigarettes cost about the same or slightly more than the 'rollies', but they're clearly a 'win' when you add the time it takes to hand-roll to the equation.
We have a department at my work that builds their own machines. It's clear that their reasons aren't financial, though. It ends up costing them about the same (since they need staff to maintain hardware that we just RMA), but they have much better -control- of their stuff, which has a lot of value in some corner-cases.
Correct me if I'm wrong, but isn't one head on one side of one low-density platter a lot more reliable than today's highest areal density disks and twice the mechanical pieces?
I always buy 'one generation behind', prices are low and bugs are worked out. I spec machines for our labs, and I always suggest the lowest speed CPU at the current fab process, the smallest hard drive, and maxed-out RAM that matches the FSB. For 'general purpose' office computers, CPU is basically a non-issue, neither are graphics. Hard drive -latency- and insufficient RAM are root of the bottlenecks users (used to) gripe about.
Slashdot Mirror
"300bps acoustic coupler using a Z80 based computer with 16k RAM and a tape recorder as secondary storage device."
You had the 16K expansion pack? No fair! I had the same beast, but with only 1K of RAM, there wasn't any point in trying to load those tapes in, the apps were too big to fit in memory. All I could do was write one-page BASIC apps.
Eventually my Dad got his hands on a spare 16K piggyback cartridge and let me have his old one, so I could play MINOTAUR.
I concur. One of the reasons I switched from Mac OS 8 to Windows 95 was the multitasking being smoother. When I saw Linux in 1998 I was totally blown away and moved to that (I really wanted to go to BeOS, but it was obviously going nowhere). Somewhere over the last decade, something has gone awry with Linux. I still use it, but it -really sucks- when I kick off a low-priority backup with rsync and can no longer browse the web, play audio, or even use the command line. Waiting over a minute for 'top' to load is a travesty. Doing the same thing used to work Just Fine back with Linux 2.4.
I feel like they squeezed-out so much performance for server tasks that there are no time slices left for actually using a desktop, or there's something fundamentally wonky with the scheduling, or they're prioritizing disk cache over apps that should stay in RAM. Whatever it is, it's definitely related to I/O, and it makes using Linux really un-fun.
Of course, it used to be that X was given scheduler priority by the distributions, 'improvements' to the kernel ended that practice. Maybe the two are related?
If I recall, the K6 was pretty weak, clock-for-clock. The K6-2 was better, but still not as good as comparable Intel offerings of the time. The K6-III was pretty awesome, but only if you couldn't afford a Pentium III, and the Athlon totally dominated for quite a while, probably until the Core 2 architecture arrived.
I think what pissed me off the most about my K6-2 was that the L2 cache was on the motherboard. The K6-2+ and K6-III were much faster because they had it on-die.
There probably were folks who thought it was a bad bill, but voted for it anyway because it bought them leverage on (what they felt were) more important issues.
I'm a bit of a state house watcher, and I've heard politicians stand up and speak against bills five minutes before voting for them. Basically, if the chairman of the committee favors something and you don't, but it's going to pass anyway, you curry favor with the chairman by letting him submit the bill to the floor with 'unanimous approval', thereby increasing the chances of getting your own issue heard by the now appeased chairman in the future. In the end, you get the same result you would have if you opposed the thing, but the next time you need something, you're more likely to get it.
That or the HVAC might have been out. Our state legislature seems to decide completely on-the-fly that 'today is going to be the last day of session'. They typically suspend public hearings and pass 300 pieces of legislation that night. Why would you suspend public hearings and do 80% of your work on one coffee-fueled all-nighter? Well, the committee rooms don't have air conditioning, suits are really hot, and most of the legislature is a bit portly. Once the summer heat starts penetrating the marble walls, there's no stopping it until late October, so they 'go Nike' on democracy's ass and Just Do It.
That's a really horrible way to set policy, if that's what you're implying.
Let X = Number of people who benefit from Bacon Cheeseburgers.
Let Y = Number of people who die from Bacon Cheeseburgers.
Obviously, we should ban Bacon Cheeseburgers, right?
I'd actually really like this. I think Linux kernel development is going to stay on 2.6 for another five years or more, since the product is essentially 'complete', and now it's just making incremental changes to keep up with the times.
What I'd like to see is a 'Linux 3.0' tree that starts at 'zero code'. Here are a few zany ideas from my non-compsci-trained brain:
1. Developed in a way that can be compiled as a microkernel -or- as a 'hybrid' like Apple's Mach. Full-blown computers would probably be better off running a microkernel, while portable/embedded devices would likely benefit from the 'hybrid' approach.
2. Has driver classes that are factored-out to minimize duplication of code. Just a thought: What if there was one kernel-level 'serial' data stream type that pretty much all I/O could inherit, various drivers would set parameters on the I/O that would tune it for the actual hardware being interfaced with. Serial links to devices and virtual disks in VMs might need in-order raw transfers, SATA drives might benefit from an intelligent algorithm spitting out blocks that were tuned to the physical interface, IP links and buffered I/O need to dynamically adapt their window sizes. All that logic should sit somewhere where it's accessible to other components that might want it.
3. There's got to be a way to get graphics drivers into the system, then have X programs run inside an X11 runtime a layer above that. I guess KMS and DRI already handle a lot of that, but tighter integration is a must.
4. Build the kernel to be The Best Damn Kernel Ever. If POSIX compliance is a need that prevents you from making The Best Damn Kernel Ever, offload POSIX to a runtime like Windows does. With an architecture like that, you could integrate stuff like WINE closer to the kernel, and portable/embedded devices could interface directly with the slimmer, simpler kernel interfaces.
5. Virtualization needs to be right in the core. The microkernel's 'servers' should be active keepers of the system, scanning through memory to find duplicate pages, actively profiling and optimizing I/O and scheduling, checking for hardware faults, and providing an interface for drivers to pipe data through looking for malicious code fragments. If Trojan.Whatever is traversing my network port and I have the NIC driver using the hypervisor as a virus scanner, I want the hypervisor to quarantine the traffic and send a message to other things on the system that handle it according to settings or inform me. Having that happen on the 'application' level like it is today is just asking for rootkits.
I can still ride, but bumpy roads or jumping a curb puts me out of commission for two days. So does sprinting to catch up with my dog, or playing with the doggie pull-toy.
I just hope whatever caused my body to lose bone mass has stopped, there's no way to tell except for checking on it every few years.
Thanks for the support. Right now it's not clear if it will stop where it is or get worse. Still, it sucks to have osteopenia when you're young.
I only found out because I got a slipped disc in my neck. My doctor asked what I did to mess things up so bad, and the answer was 'nothing at all', which triggered all sorts of tests to find out if I had nutritional or hormonal deficiencies, or even cancer. So far it looks like None of the Above, which is a good thing, I think.
For all the humor in the title, there's hopefully just as much promise.
My doctor recently told me that my twenty-something year old skeleton is basically aged like a geriatric's. The implications long term are not good. If they can make stem cells grow bone and muscle, I might not spend my fifties fighting infections in a wheelchair. It's bad enough not being able to ride a bike before I'm 30.
As a stockholder, it's in my interests that the company -not- screw up again, and that they do right by the Gulf communities.
I get a voting stake on who sits on the board.
I -take their money-.
All you do on the sidelines as a normal consumer is buy the product.
Look, I did the math and researched oil spills of this magnitude before I bought in. I don't think this spill is as bad as most environmentalists want it to be, and I'm overall quite pleased with BP's restitution and cleanup. In the end, it's going to be half a missed shrimping season and a basic cleanup of the few places where oil did make it to shore.
http://www.dailymail.co.uk/news/article-492804/The-uninvited-guest-Chinese-sub-pops-middle-U-S-Navy-exercise-leaving-military-chiefs-red-faced.html
Looks like the Chinese are getting handy with these sabres, eh?
I bought-in like mad when the stock started going below $40/share. I made a bunch of loot, and I'm just a regular working Joe. There's a reason stocks are 'public', it means anyone can buy them.
It drives me mad when people say that investing is something for 'people with lots of power and money', It's not rocket-science, and anyone can do it. Investing has netted me about $15K of spending cash in the last two years, all on a middle-class salary.
Anyone with half a brain knows that institutional investors were dumping BP stock because their clients were angry at them and it was driving the stock down; the fiscal fundamentals would put the shares back over $40 once things stabilized. I'm still holding, though. I suspect a lot of the $20B put aside won't get used and BP will resume dividends again soon. Once that happens, there's no reason the stock won't bump up another 30%.
It shows that sex and drug use are now the activities of the most popular of the crowd, the trendsetters that everyone looks up to. When I was in high school, the weird kids were the ones smoking, drinking, and having sex the earliest. Sure, the 'cool kids' and jocks caught on later, in junior and senior year, but the only 14 year-olds having sex (that I know of) in my high school were the antisocial goths and Nine Inch Nails fans.
"How does it make any sense to show off that we're "within range" by launching right off our coast, when they could announce a test off their own coast and let us track it, without the danger of touching off nuclear Armageddon?"
I agree that this is most likely one of our own, but the logic of sneaking in close and launching in the opposite direction is a pretty safe thing to do. The first telemetry data will show that the missile isn't a threat (since it's headed in the other direction). A launch 'from home' towards us would be much more likely to inspire immediate retaliation. Plus, it wouldn't show that you can sneak up close and have your 1500-mile range missiles cover our whole territory.
I hope you're wrong. It seems to me that from 12 nautical miles out and farther, anyone could pop up and still be legally within 'international waters'.
If the Chinese popped-up 35 miles out and we did anything besides issue a friendly escort, we'd be picking a fight.
Bingo!
There's no reason for us to demonstrate our well-known SLBM capabilities to our own country's most populated county. I think it's highly likely that this is one of our 'peer nations' showing-off that we're entirely 'within range'.
Trust me, they saw this at Beale, where they have a huge radar system designed to see incoming ICBMs. If this wasn't an exercise or a test launch (both of which would have likely happened a little more out of the way), you won't hear much more about it. Neither the military nor the media would tell Americans if China, Russia, or India was playing around right off our beaches, the reality that we could all easily become iridescent chalky dust at the drop of a hat distracts from the main objectives: Fighting bad guys with AK-47s to secure cheap energy, and buying stuff on credit.
I think I have an idea where this might be heading...
You can fit 4 Mac Mini servers into 1U now, they just need a tray that diverts the heat from them. That's a killer opportunity right there. You get much more oomph from four Mac Minis than you do from an XServe.
http://www.apple.com/macmini/specs.html, yup, you could definitely fit these one-high, and side-by-side into a 1U tray, with room for connectors and venting. I think trays might even be deep enough to hold six of them.
X programs on Ubuntu clients and servers will still have to run in X, but X will be running inside Wayland. Think of it like running Xming inside Windows. I can still 'host' and 'join' X sessions to and from it.
I think this might actually be a good move, X is overkill for 98% of users. A smaller, slimmer, faster-moving display foundation that still allows the 'legacy' X stuff to run in a runtime would be great.
I don't think that we can determine if there actually -is- anything in his/her hand. Take a walk around a city sometime and you'll see plenty of loonies holding the sides of their heads, talking back to the voices they 'hear' inside their mind. I'm guessing schizophrenia manifested itself the same way back then as it does now.
Hell, there could have been a loud noise nearby. I make that move when an ambulance passes by.
It certainly doesn't affect all hardware. I think it's a thing with certain Intel boards/chipsets.
The bug cropped up with VirtualBox 3 and has persisted to this day. Opening the Pulseaudio Volume Meter seems to help, but not entirely fix the issue.
I run VirtualBox on a 3GHz 64-bit Core 2 Linux host. Both Debian and Ubuntu have a frustrating long-term bug that makes sound choppy and unusable for Windows guests. That said, I still run it (since I'm doing 'systems' stuff and no multimedia).
My current system is actually pretty awesome:
1. The host Linux OS has a Samba share that is joined via winbind to...
2. A guest running Windows Server 2008 R2.
Clients are my other machines, and a slew of VMs running XP, Windows 7, and Linux.
I get bare-metal file server performance from Linux and a real Active Directory and WDS deployment server to test Windows stuff for work.
In my state the tax on rolling papers alone is $0.17 each. Drums of tobacco are taxed at 80% of the wholesale cost. It adds up to a very expensive habit, even to roll your own.
In a perfect world, you would call up your state rep/senator and tell them about this crazy idea:
A small group of hip technical folks sit on a committee that sets a 'standard' config once a year. The config just has to hit the bases for 'basic office machine', no advanced 3D, no huge hard drive. This is the machine that 90% of your state's desktop hardware budget will go to, and the schools and municipalities can get on board too. Put the config out to bid, and let vendors fight over the price. You'll end up getting cheaper desktops and cutting a huge amount of IT bullcrap from your state and local government.
I've been to way too many poor public schools that have 'regular office PCs' with $400 3D cards in them to trust the way we do it now. Those vendors might not be convincing when you're buying a PC for yourself, but when you're spending someone else's money, it's easy to get the big HD, expensive CPU, and top-tier video card.
I know folks who thought it was super-smart to roll their own cigarettes. In the end, they pay a huge tax on the tobacco, and a huge tax on papers, and spend a fair amount of their time trying to roll cigarettes in windy new-england weather. In the end, the pre-packaged cigarettes cost about the same or slightly more than the 'rollies', but they're clearly a 'win' when you add the time it takes to hand-roll to the equation.
We have a department at my work that builds their own machines. It's clear that their reasons aren't financial, though. It ends up costing them about the same (since they need staff to maintain hardware that we just RMA), but they have much better -control- of their stuff, which has a lot of value in some corner-cases.
Correct me if I'm wrong, but isn't one head on one side of one low-density platter a lot more reliable than today's highest areal density disks and twice the mechanical pieces?
I always buy 'one generation behind', prices are low and bugs are worked out. I spec machines for our labs, and I always suggest the lowest speed CPU at the current fab process, the smallest hard drive, and maxed-out RAM that matches the FSB. For 'general purpose' office computers, CPU is basically a non-issue, neither are graphics. Hard drive -latency- and insufficient RAM are root of the bottlenecks users (used to) gripe about.