There are no stong competitors and I truely believe that the would could run on Windows XP for another 5 years without problems. There is no real incentive to get a new operating system out.
Most of the world is already planning on running WinXP for another 5-10 years.
Vista's marketshare will be driven by new systems with a small bump from people who upgrade (10% of all WinXP owners might be generous).
Combine that with the fact that computers built in the last few years will probably still be in use come 2010 and 2015. Things start looking a bit thin for how fast Vista will take over the market.
Some of the best registration code that I've seen tells you to copy the entire e-mail into the clipboard and then the program pulls what it needs. Unfortunately, I don't recall which program that was... probably UltraEdit, SecureCRT, FTP Voyager or maybe something else.
I don't know about the states but I have a feeling Linux is still just as a) unknown b) scaring c) looked at as a server OS to the general population as here in Sweden.
Even though I'm technically savvy, we're still running Windows on the desktops. But we're rapidly moving servers to Linux.
Now, I'll probably play guinea-pig next year and get a new laptop with Ubuntu on it running Windows XP in a VMWare Workstation VM. Depending on how that goes will determine whether we start switching away from Windows on the desktop in 2010 (our next upgrade window).
(Getting all the servers onto unix/linux is the first step and is going to take a few years. After that, linux on the desktop becomes much easier because we're more familiar with it as a server environment. Why so long? Because I only have two hands and one brain. If I don't understand it, I can't teach the others in the group.)
There won't be any further migration, infact the Firefox 2.0 release might give Microsoft an opportunity given that it won't anything exciting to the average Firefox user.
Hmmm, one of Firefox's strong points is extensions. For that reason alone, I'm staying with Firefox vs switching to Opera (or back to IE7).
Some of the extensions I use don't fit the corporate ideals. Mozilla has no evil reasons to break such extensions, but what about extensions (or whatever they are called) written for IE7?
I get my motherboard bundles from MWave, along with most of the other parts. I got tired of spending hours trying to match up memory, CPUs and motherboards and paying the $9 to MWave to do the assembly is worth it.
$0099 BA30107 Antec Sonata II w/ SmartPower 2.0 450W PSU
$0045 small hard drive (usually 80GB)
$0025 DVD-ROM
$0009 Floppy drive
=====
$0178
Software:
$0131 AA15070 WindowsXP Pro OEM
$0299 AA24200 Microsoft Office Pro 2003 OEM
Shipping runs about $75-100 for all that, maybe a little less if we ordered in a single order. Since I live on the east coast, Fry's is not an option. Usually we have to spend another $50-$200 on various software, but that varies on the user and what they need to do their jobs.
Remote users get a 2nd $80 hard drive for Second Copy backups of local e-mail files and someplace to store Knoppix/NTFSClone images of the primary drive. Office machines backup to a central server and I simply use a hidden partition on the main drive to store the image files. (Image files are also stored elsewhere on a central server as a backup.)
Alternately for the case, we're considering use of the Antec Minuet 300 cases. But I still need to test one of them. They have a few disadvantages over the Antec Sonata II cases: low-profile cards required, only 1 HD bay, custom-shaped PSU that may not be easy to replace in a hurry. But for a dual-core linux firewall, we figure we can put 2 notebook drives inside and get good results. We'll simply keep a spare MT300 PSU on-hand in case the primary PSU fails. (Since we can't just run down to the local CompUSA to replace it with a standard PSU.)
The Minuet300 case would also shave a few dollars off the cost, but if we switched to laptop drives, costs would go back up (due to the need for the $15 IDE to laptop drive converters).
All that and I still come in at 2/3 to 3/4 the cost of a Dell. Mostly I'm just giving up Dell warranty coverage. Since the users call me *anyway* when the Dells act up, Dells service isn't gaining me anything (in terms of making my job less stressful). But by DIY, I get 100% commodity parts that can be easily swapped out. The CPUs are 3-year warranty and I use 5-year warranty hard drives. PSUs are another common failure point, but they're easy enough to replace after the 1-year warranty expires.
One other thing I do is to run Prime95's torture test and a disk benchmark on the system for a few days to a full week as a burn-in check. That hopefully turns up any issues with the system before we deliver it to the end-user.
All that and it's fun to play with new hardware. These users are still using Win98 machines from the 1997-2000 era. Even the 1GB Athlon 3000+ that I dropped onto a user's desk the other week seems "fast" to them (and it will be one of the slowest systems in the facility after the upgrade cycle).
That's the real kicker. Can the industry field enough multithread applications, that can scale well, to make using a four core processor worth the money? Shaving a few minutes off a DVD encode does not justify the large price tag these four core processors will cost.
OTOH, it might drive the dual-core prices down again!
So, it's quite possible that today's MB will be good for at least a few years. (I don't see applications or OSes for the desktop exceeding the current 8GB max supported RAM for a little while, you can even fit most DVDs into RAM with 8GB.)
If so, that would be the first time in quite a while that it's true. Other then say the start of the Socket 939 era. I do agree with your sentiment that we're at the starting point of a whole bunch of new technologies with a bit of running room.
What I typically find is that older motherboards don't always support the newer CPUs. And if memory technology has gotten faster (PC2100 vs PC3200 for instance), you're going to be buying new RAM anyway.
At which point... spending another $85 on a new and better motherboard starts to make sense. Maybe they worked out some of the older bugs, or changed some of the layout to make certain things faster. I guess if you bought an AM2 Sempron for $45 and later upgraded to a quad-core Athlon64 it would make sense.
(I prefer to upgrade CPU/RAM/MB as a single unit. With only the addition of more memory at a later date to an existing bundle.)
Um... RAID6 uses (2) disks for parity instead of just one. So that's actually 7 data disks, 2 parity disks and 1 hot-spare. The advantage of RAID6 is that if you lose one disk, you degrade to a RAID5 array (I think). That gives you time to get that hot spare worth of parity back up and running before another two more disks fail.
(Versus racing the clock in degraded RAID5 where one more disk going will take out everything.)
Sometimes you need to use tools other then MemTest86+ to diagnose flaky / transient issues. MemTest86+ doesn't put enough of a load on the CPU to catch things like flaky memory timings. My preference is the Prime95 client, which folks have used for close to 10 years as a torture test. In fact, the Prime95 folks finally added a "torture test" option many years ago. QuickPar is also sensitive to flaky systems and will often catch errors (but isn't as thourough of a test as Prime95 is).
System shutdowns are typically either:
- Thermal initiated (so use SpeedFan and monitor your temps)
- Power related, the PSU can't keep up with the load, drops a voltage and the motherboard / CPU panics.
If your system can run Prime95 for 48 hours straight with zero errors, you've got a good stable system. We typically run Prime95 as our burn-in test while exercising the disks with another program for a few days prior to releasing the machine to the end-user.
All of the new systems that I'm building for the main office are now X2 3800+ AM2s (on an Asus M2NPV-VM microATX motherboard). If I remember my numbers correctly, it's about $550 for hardware parts and $450 for WinXP and MSOffice Pro. The additional cost for the X2 3800+ vs the 3000+ is low enough that we switched once the price cuts took effect.
(Yes, I haven't quite gotten us off the Microsoft wagon yet on the desktop. Although we're slowly converting some folks over to OS X.)
My estimate is that these systems will have lifespans of 8-10 years. I'm trying to get everyone upgraded before WinXP gets pulled from the market. We'd like to avoid Vista for at least the first year or two if possible. The users are general office users who won't even put a 3000+ to the grindstone. The X2 decision was more for longevity (figuring that it will add a few years of useful life to the machine).
Gee, I notice the latency on a single-core machine *all* the time.
Drives me nutz. I have to really work hard at making my dual-CPU desktop feel slow. But on the single-core machine, it's as easy as running one or two CPU intensive task(s).
This was much more evident (and noticeable) back in the days of the dual and quad CPU P3s.
I think one reason we haven't seen too many consumer-grade 10000 rpm drives is because, as far as I have experienced, they're really damn loud. I can't imagine a non-geek being happy with a drive whose spinning noise drowns out the people in the room talking to each other.
The newer FDB gets rid of a lot of the whine of the older drives. But you still have the head chatter which seems to be a bit sharper (noticable) on the WD Raptors.
The bigger issue, I'd imagine, with 10k drives is heat. Which requires either heatsinks, better case design, or a dedicated fan to move air across the drive.
Granted, access times probably haven't declined like transfer rates.
Access time is closely coupled with the rotational velocity of the platter (how long before the sector comes under the read head) along with how fast the head can move from track to track (seek time). Not exactly sure where the latency term fits in (whether it's coupled with the rotational velocity of the disk or a combined value of seek time + rotational latency).
The 10k Raptors have a pretty decent seek time (I think they use the 2.5" form factor platters instead of 3.5" form factor platters) and a lower rotation latency.
Not sure why head seek times haven't improved. Maybe they're already fast enough to be in position before the sector comes around the bend.
I've bought the Family Pack several times, to update my laptop, my G5, my daughter's laptop, etc. Yes, I could save $70 by purchasing the single license and copying it on all the machines, but I feel that the extra $70 to be legal is more than fair. Apple has done a good job of adding value to each release, and they need to get something for their work.
And the Family Packs offer us a way to purchase more licenses without feeling like we're paying through the nose for all of those upgrades. So we get to stay legal, Apple gets a little additional money, and everyone is happy all the way around.
When the controller/circuit board is mated to the platter assembly it is programmed for that specific set of platters and any anomolies on those platters. This is why you cannot take a circuitboard/controller from one drive and put it on another one.
You didn't know that data recovery services do take platters from a dead drive and mount them in an identical make/model that works in order to recover the data? While you may not be able to arbitrarily take platters from designs other then the same model line and place them in a different system, platters are not specifically tied to the controller/circuit board that they were originally mated with.
(This is why data recovery companies tout their access to a clean room, so that they can do these transplants without introducing dust onto the surface of the platters.)
The primary mode of failure that I've seen is heat death. Which probably means that the controller chips have baked themselves due to inadequate airflow. (All it takes is a minor amount of forced air movement across the surface of a drive to keep it cool.) I'm pretty sure that the platters are not affected by heat or are at least resilient enough to handle temperatures higher then what would kill the controller silicon.
Do newer FDB drives suffer the same issues as the older ball bearing designs? (Not that I've ever seen a drive die because the bearings on the motor gave out.)
USB 2.0 is faster than FireWire 400 (which is generally used for DV content), so I'd say that it is fast enough.
That's arguable (and a bit of a holy war). Looking only at the 480Mbits of USB 2 vs Firewire's 400Mbits glosses over the differences in the two protocols. In reality, both are usually constrained by the speed of the disks, which are identical for both implementations.
But if you're a determined fence sitter, go with a dual-interface external enclosure such as the BYTECC ME-835U2F enclosures. They offer both USB 2 and Firewire 400 and include a fan inside that moves air directly over the hard drive (which will extend the life of the hard drive). This enclosure also has an internal PSU so there's no special wall warts to lose or misplace.
(We've got 8 of these enclosures in service and we haven't heard any complaints in the past few months.)
Um, 500GB drives have been available on the market for about 10+ months now (maybe a bit longer). They represent the upper end of what longitudinal recording was capable of packing into a 3.5" form factor.
The advent of perpendicular recording from multiple vendors (Hitachi has been dragging their heels on a 3.5" PR drive) will hopefully drive prices down on the 500GB and 750GB drives. Or at least accellerate the price drops.
Will there ever be an upper limit to hard drives? I know we just started using perpendicular technology, but there must be some kind of physical limit to the platters. Another question is why is it hard to find SCSI drives in these high capacities? Or at least in newer SAS drives.
From what I've read over the past year, perpendicular recording supposedly will offer densities somewhere between 2x and 5x over existing longitudinal recording methods. That puts 3.5" SATA/IDE drive somewhere in the range of 1TB to 2.5TB before they hit the wall again. For 2.5" SCSI, 600MB up to 1.5TB. I suspect that things will top out around 3x-4x densities over existing drives. (GMR longitudinal recording was supposed to bring us greater gains then it did. You can look back at the original announcements of bit densities and then look at what finally made it to market at the high end.)
SCSI drives are a different form factor. They use smaller platters inside to allow for higher rotational velocities (10k/15k RPM) and faster seek times. That limits their capacity per platter.
(I did all the math about 3 months ago for another article, looking at existing bit densities vs what perpendicular recording bit densities were estimated to be at the upper end.)
The college that I went to had a 2 semester system with a short 3-week semester wedged into January. The 3-week J-term was typically dedicated to a single 3-credit class or a pair of 1-credit classes. So we had to learn Fortran77 and Numerical Analysis within 3 weeks. Fortunately, Fortran77 is a pretty simple language for this purpose or things would've been a bit rough.
It was actually one of my favorite classes in college as it appealed to the geek within (learning about floating point representations) as well as being practical (estimating and bounding error values). Our textbook was called "Elementary Numerical Analysis" and had a bright green cover.
I might be off-base remembering it as being required. Half the engineering majors may have taken another course during that J-term. Other schools might lump it in with Physics 101 (along with significant digits) or one of the math classes (Calculus? Linear Algebra?).
We also studied Numerical Analysis. Except it was only a required course for the Engineering majors (EE, ME, CE). It was a 3-week course taught in a mini January term using Fortran77.
Kind of odd that it wasn't taught to the CS folks.
I don't think I have ever met a household IBM. Nobody ever got fired for buying IBM, but householders rarely need to worry about getting fired. Personally I bought a Compaq transportable and a Tandy Desktop.
At least one classmate back in 85 or 86 owned an IBM at home. Don't recall if it was a PC or XT at the time.
For those of us who aren't programming geniuses- what would you use to store a monetary amount, besides a floating-point format?
In databases? Currency formats. Which are specifically designed not to lead to rounding errors. Some of them even allow you to specify the number of places after the decimal.
In code? A numeric type designed for currency work (typically an add-in library). In a pinch you can use a 32bit integer and use the last 2 decimal digits as "cents", but you'll run the risk of overflow.
Slashdot Mirror
There are no stong competitors and I truely believe that the would could run on Windows XP for another 5 years without problems. There is no real incentive to get a new operating system out.
Most of the world is already planning on running WinXP for another 5-10 years.
Vista's marketshare will be driven by new systems with a small bump from people who upgrade (10% of all WinXP owners might be generous).
Combine that with the fact that computers built in the last few years will probably still be in use come 2010 and 2015. Things start looking a bit thin for how fast Vista will take over the market.
Some of the best registration code that I've seen tells you to copy the entire e-mail into the clipboard and then the program pulls what it needs. Unfortunately, I don't recall which program that was... probably UltraEdit, SecureCRT, FTP Voyager or maybe something else.
Definitely about as simple as it gets.
I don't know about the states but I have a feeling Linux is still just as a) unknown b) scaring c) looked at as a server OS to the general population as here in Sweden.
Even though I'm technically savvy, we're still running Windows on the desktops. But we're rapidly moving servers to Linux.
Now, I'll probably play guinea-pig next year and get a new laptop with Ubuntu on it running Windows XP in a VMWare Workstation VM. Depending on how that goes will determine whether we start switching away from Windows on the desktop in 2010 (our next upgrade window).
(Getting all the servers onto unix/linux is the first step and is going to take a few years. After that, linux on the desktop becomes much easier because we're more familiar with it as a server environment. Why so long? Because I only have two hands and one brain. If I don't understand it, I can't teach the others in the group.)
(polite applause for all)
Definitely one of the better off-topic threads I've seen in a few days.
There won't be any further migration, infact the Firefox 2.0 release might give Microsoft an opportunity given that it won't anything exciting to the average Firefox user.
Hmmm, one of Firefox's strong points is extensions. For that reason alone, I'm staying with Firefox vs switching to Opera (or back to IE7).
Some of the extensions I use don't fit the corporate ideals. Mozilla has no evil reasons to break such extensions, but what about extensions (or whatever they are called) written for IE7?
I get my motherboard bundles from MWave, along with most of the other parts. I got tired of spending hours trying to match up memory, CPUs and motherboards and paying the $9 to MWave to do the assembly is worth it.
$0152 MB-BA22656 AMD ATHLON 64 X2 3800+ AM2
$0085 Asus M2NPV-VM
$0086 Kingston 2x512MB DDR2 533
$0009 Test Bundle
=====
$0332 1GB RAM
$0383 2GB RAM (+$51)
$0099 BA30107 Antec Sonata II w/ SmartPower 2.0 450W PSU
$0045 small hard drive (usually 80GB)
$0025 DVD-ROM
$0009 Floppy drive
=====
$0178
Software:
$0131 AA15070 WindowsXP Pro OEM
$0299 AA24200 Microsoft Office Pro 2003 OEM
Shipping runs about $75-100 for all that, maybe a little less if we ordered in a single order. Since I live on the east coast, Fry's is not an option. Usually we have to spend another $50-$200 on various software, but that varies on the user and what they need to do their jobs.
Remote users get a 2nd $80 hard drive for Second Copy backups of local e-mail files and someplace to store Knoppix/NTFSClone images of the primary drive. Office machines backup to a central server and I simply use a hidden partition on the main drive to store the image files. (Image files are also stored elsewhere on a central server as a backup.)
Alternately for the case, we're considering use of the Antec Minuet 300 cases. But I still need to test one of them. They have a few disadvantages over the Antec Sonata II cases: low-profile cards required, only 1 HD bay, custom-shaped PSU that may not be easy to replace in a hurry. But for a dual-core linux firewall, we figure we can put 2 notebook drives inside and get good results. We'll simply keep a spare MT300 PSU on-hand in case the primary PSU fails. (Since we can't just run down to the local CompUSA to replace it with a standard PSU.)
The Minuet300 case would also shave a few dollars off the cost, but if we switched to laptop drives, costs would go back up (due to the need for the $15 IDE to laptop drive converters).
All that and I still come in at 2/3 to 3/4 the cost of a Dell. Mostly I'm just giving up Dell warranty coverage. Since the users call me *anyway* when the Dells act up, Dells service isn't gaining me anything (in terms of making my job less stressful). But by DIY, I get 100% commodity parts that can be easily swapped out. The CPUs are 3-year warranty and I use 5-year warranty hard drives. PSUs are another common failure point, but they're easy enough to replace after the 1-year warranty expires.
One other thing I do is to run Prime95's torture test and a disk benchmark on the system for a few days to a full week as a burn-in check. That hopefully turns up any issues with the system before we deliver it to the end-user.
All that and it's fun to play with new hardware. These users are still using Win98 machines from the 1997-2000 era. Even the 1GB Athlon 3000+ that I dropped onto a user's desk the other week seems "fast" to them (and it will be one of the slowest systems in the facility after the upgrade cycle).
That's the real kicker. Can the industry field enough multithread applications, that can scale well, to make using a four core processor worth the money? Shaving a few minutes off a DVD encode does not justify the large price tag these four core processors will cost.
OTOH, it might drive the dual-core prices down again!
(Going for the eternal optimist stance here...)
So, it's quite possible that today's MB will be good for at least a few years. (I don't see applications or OSes for the desktop exceeding the current 8GB max supported RAM for a little while, you can even fit most DVDs into RAM with 8GB.)
If so, that would be the first time in quite a while that it's true. Other then say the start of the Socket 939 era. I do agree with your sentiment that we're at the starting point of a whole bunch of new technologies with a bit of running room.
What I typically find is that older motherboards don't always support the newer CPUs. And if memory technology has gotten faster (PC2100 vs PC3200 for instance), you're going to be buying new RAM anyway.
At which point... spending another $85 on a new and better motherboard starts to make sense. Maybe they worked out some of the older bugs, or changed some of the layout to make certain things faster. I guess if you bought an AM2 Sempron for $45 and later upgraded to a quad-core Athlon64 it would make sense.
(I prefer to upgrade CPU/RAM/MB as a single unit. With only the addition of more memory at a later date to an existing bundle.)
Don't think there are AM2 SMP boards yet. The dual-core Opteron 2xx chips are still all 940 pin.
Um... RAID6 uses (2) disks for parity instead of just one. So that's actually 7 data disks, 2 parity disks and 1 hot-spare. The advantage of RAID6 is that if you lose one disk, you degrade to a RAID5 array (I think). That gives you time to get that hot spare worth of parity back up and running before another two more disks fail.
(Versus racing the clock in degraded RAID5 where one more disk going will take out everything.)
Sometimes you need to use tools other then MemTest86+ to diagnose flaky / transient issues. MemTest86+ doesn't put enough of a load on the CPU to catch things like flaky memory timings. My preference is the Prime95 client, which folks have used for close to 10 years as a torture test. In fact, the Prime95 folks finally added a "torture test" option many years ago. QuickPar is also sensitive to flaky systems and will often catch errors (but isn't as thourough of a test as Prime95 is).
System shutdowns are typically either:
- Thermal initiated (so use SpeedFan and monitor your temps)
- Power related, the PSU can't keep up with the load, drops a voltage and the motherboard / CPU panics.
If your system can run Prime95 for 48 hours straight with zero errors, you've got a good stable system. We typically run Prime95 as our burn-in test while exercising the disks with another program for a few days prior to releasing the machine to the end-user.
All of the new systems that I'm building for the main office are now X2 3800+ AM2s (on an Asus M2NPV-VM microATX motherboard). If I remember my numbers correctly, it's about $550 for hardware parts and $450 for WinXP and MSOffice Pro. The additional cost for the X2 3800+ vs the 3000+ is low enough that we switched once the price cuts took effect.
(Yes, I haven't quite gotten us off the Microsoft wagon yet on the desktop. Although we're slowly converting some folks over to OS X.)
My estimate is that these systems will have lifespans of 8-10 years. I'm trying to get everyone upgraded before WinXP gets pulled from the market. We'd like to avoid Vista for at least the first year or two if possible. The users are general office users who won't even put a 3000+ to the grindstone. The X2 decision was more for longevity (figuring that it will add a few years of useful life to the machine).
Gee, I notice the latency on a single-core machine *all* the time.
Drives me nutz. I have to really work hard at making my dual-CPU desktop feel slow. But on the single-core machine, it's as easy as running one or two CPU intensive task(s).
This was much more evident (and noticeable) back in the days of the dual and quad CPU P3s.
I think one reason we haven't seen too many consumer-grade 10000 rpm drives is because, as far as I have experienced, they're really damn loud. I can't imagine a non-geek being happy with a drive whose spinning noise drowns out the people in the room talking to each other.
The newer FDB gets rid of a lot of the whine of the older drives. But you still have the head chatter which seems to be a bit sharper (noticable) on the WD Raptors.
The bigger issue, I'd imagine, with 10k drives is heat. Which requires either heatsinks, better case design, or a dedicated fan to move air across the drive.
Granted, access times probably haven't declined like transfer rates.
Access time is closely coupled with the rotational velocity of the platter (how long before the sector comes under the read head) along with how fast the head can move from track to track (seek time). Not exactly sure where the latency term fits in (whether it's coupled with the rotational velocity of the disk or a combined value of seek time + rotational latency).
The 10k Raptors have a pretty decent seek time (I think they use the 2.5" form factor platters instead of 3.5" form factor platters) and a lower rotation latency.
Not sure why head seek times haven't improved. Maybe they're already fast enough to be in position before the sector comes around the bend.
I've bought the Family Pack several times, to update my laptop, my G5, my daughter's laptop, etc. Yes, I could save $70 by purchasing the single license and copying it on all the machines, but I feel that the extra $70 to be legal is more than fair. Apple has done a good job of adding value to each release, and they need to get something for their work.
And the Family Packs offer us a way to purchase more licenses without feeling like we're paying through the nose for all of those upgrades. So we get to stay legal, Apple gets a little additional money, and everyone is happy all the way around.
Smart business move on Apple's part.
When the controller/circuit board is mated to the platter assembly it is programmed for that specific set of platters and any anomolies on those platters. This is why you cannot take a circuitboard/controller from one drive and put it on another one.
You didn't know that data recovery services do take platters from a dead drive and mount them in an identical make/model that works in order to recover the data? While you may not be able to arbitrarily take platters from designs other then the same model line and place them in a different system, platters are not specifically tied to the controller/circuit board that they were originally mated with.
(This is why data recovery companies tout their access to a clean room, so that they can do these transplants without introducing dust onto the surface of the platters.)
The primary mode of failure that I've seen is heat death. Which probably means that the controller chips have baked themselves due to inadequate airflow. (All it takes is a minor amount of forced air movement across the surface of a drive to keep it cool.) I'm pretty sure that the platters are not affected by heat or are at least resilient enough to handle temperatures higher then what would kill the controller silicon.
Do newer FDB drives suffer the same issues as the older ball bearing designs? (Not that I've ever seen a drive die because the bearings on the motor gave out.)
USB 2.0 is faster than FireWire 400 (which is generally used for DV content), so I'd say that it is fast enough.
That's arguable (and a bit of a holy war). Looking only at the 480Mbits of USB 2 vs Firewire's 400Mbits glosses over the differences in the two protocols. In reality, both are usually constrained by the speed of the disks, which are identical for both implementations.
But if you're a determined fence sitter, go with a dual-interface external enclosure such as the BYTECC ME-835U2F enclosures. They offer both USB 2 and Firewire 400 and include a fan inside that moves air directly over the hard drive (which will extend the life of the hard drive). This enclosure also has an internal PSU so there's no special wall warts to lose or misplace.
(We've got 8 of these enclosures in service and we haven't heard any complaints in the past few months.)
Um, 500GB drives have been available on the market for about 10+ months now (maybe a bit longer). They represent the upper end of what longitudinal recording was capable of packing into a 3.5" form factor.
The advent of perpendicular recording from multiple vendors (Hitachi has been dragging their heels on a 3.5" PR drive) will hopefully drive prices down on the 500GB and 750GB drives. Or at least accellerate the price drops.
Will there ever be an upper limit to hard drives? I know we just started using perpendicular technology, but there must be some kind of physical limit to the platters. Another question is why is it hard to find SCSI drives in these high capacities? Or at least in newer SAS drives.
From what I've read over the past year, perpendicular recording supposedly will offer densities somewhere between 2x and 5x over existing longitudinal recording methods. That puts 3.5" SATA/IDE drive somewhere in the range of 1TB to 2.5TB before they hit the wall again. For 2.5" SCSI, 600MB up to 1.5TB. I suspect that things will top out around 3x-4x densities over existing drives. (GMR longitudinal recording was supposed to bring us greater gains then it did. You can look back at the original announcements of bit densities and then look at what finally made it to market at the high end.)
SCSI drives are a different form factor. They use smaller platters inside to allow for higher rotational velocities (10k/15k RPM) and faster seek times. That limits their capacity per platter.
(I did all the math about 3 months ago for another article, looking at existing bit densities vs what perpendicular recording bit densities were estimated to be at the upper end.)
The college that I went to had a 2 semester system with a short 3-week semester wedged into January. The 3-week J-term was typically dedicated to a single 3-credit class or a pair of 1-credit classes. So we had to learn Fortran77 and Numerical Analysis within 3 weeks. Fortunately, Fortran77 is a pretty simple language for this purpose or things would've been a bit rough.
It was actually one of my favorite classes in college as it appealed to the geek within (learning about floating point representations) as well as being practical (estimating and bounding error values). Our textbook was called "Elementary Numerical Analysis" and had a bright green cover.
I might be off-base remembering it as being required. Half the engineering majors may have taken another course during that J-term. Other schools might lump it in with Physics 101 (along with significant digits) or one of the math classes (Calculus? Linear Algebra?).
We also studied Numerical Analysis. Except it was only a required course for the Engineering majors (EE, ME, CE). It was a 3-week course taught in a mini January term using Fortran77.
Kind of odd that it wasn't taught to the CS folks.
I don't think I have ever met a household IBM. Nobody ever got fired for buying IBM, but householders rarely need to worry about getting fired. Personally I bought a Compaq transportable and a Tandy Desktop.
At least one classmate back in 85 or 86 owned an IBM at home. Don't recall if it was a PC or XT at the time.
For those of us who aren't programming geniuses- what would you use to store a monetary amount, besides a floating-point format?
In databases? Currency formats. Which are specifically designed not to lead to rounding errors. Some of them even allow you to specify the number of places after the decimal.
In code? A numeric type designed for currency work (typically an add-in library). In a pinch you can use a 32bit integer and use the last 2 decimal digits as "cents", but you'll run the risk of overflow.