According to Star Trek timelines, the U.S.S. Enterprise (aircraft carrier) was destroyed by a nuclear bomb during the Eugenics wars in 1996. (Notable because it DIDN'T happen:). Bonus points if you remember the name of the "nuclear wessel" in ST:IV (hint: don't think too hard!).
And there was a space station called "Enterprise" somewhere between the space shuttle and the NX-1. I think it's one of the pictures in somebody's ready room, can't remember which. Yeah, there's a list of this somewhere... maybe the techical manual?
The ratio between each pair of terms is a progressively better approximation of the golden mean. Really.
If you have way too much time on your hands (as my Linear Algebra teacher once did), you can construct a non-iterative formula for the sequence out of this. Requires diagonalization, messy enough you actually have to do it by hand... but it actually works.
So, the golden ratio is pretty easy to approximate. Nature's done it for millions of years. You just haven't tried hard enough...:)
That could be why the footprint is so big... I'm running apps written in Swing - although even then the GUI is nice and responsive. I do stand corrected... the Java-based IDE (either JBuilder or Eclipse) is 40Meg, the program itself is only 20Meg (still way too big IMHO - I haven't even loaded any data!). Oops - now another correction - after a minute of running idle, my program's memory drops to under 10Meg! Spooky... but insanely wasteful on startup.
It seriously takes 3-4 seconds for an "open file" dialog box to come up (Solaris runtime). The second time it's down to maybe 1/3 of a second (thank you hotspot!), and I'd bet I could take that time out by caching the dialog instead of making a new one each time. It's that first pass being way too slow though...
Hotspot in of itself is pretty good. A year and a half ago, I was cursing Java for slow everything - now it's just the startup time and memory footprint - the rest of the language has shaped up nicely. Another year of improvements and I just might switch...:-)
Never gotten to play with one of those native compilers:-(. But I'd bet a good one would take out half my complaints right now! However, I just don't feel the memory footprint is likely to go down.
Hotspot server is better? Very interesting... Thanks for the info!
Just realized what I just said. Java is better the longer the program runs? Very, very strange.../shudder/. Meaning, Java is exactly the WRONG thing to use for an applet and the RIGHT thing to use for a heavyweight editor!
/Retreats to a Java editor to think about this.../
1) Hotspot: definitely interprets TODAY. Designed to interpret. Just intelligently compiles the most-commonly used code. And yes, it's a major speed boost to do things this way. But tight C/C++ would be faster. Roughly, all Java ~ mediocre C++, which means it's easier to write good Java, but Java won't solve all problems.
2) Trying to compile everything (bytecode->native) beforehand is what makes the JVM so slow. A programming language that loads 40Megs of memory for a GUI and takes several seconds to load (on a fast machine!) is bad. And then you close the program, and lose all the gains running it and compiling efficiently earned you...
I'd think a big problem with this idea would be keeping the data stable. With most OSes, you can just reboot and that'll reset all the in-memory data structures. If one of those very-important structures gets corrupted, the entire OS becomes unusable. Can't flush things, can't restore stability = Kernel Panic, or the like - the OS kills itself to save the data on disk (because corrupted kernel info can cause corrupted writes...). Since it's (comparatively) easy to reboot the OS and re-create all the data structures, it's a good trade-off.
A LOT of work goes into keeping the disk in a stable state. Just look at the variety of filesystems out there right now - anything interesting in the past 30 years has been concerned with writing consistent data and crash-proofing disks, so that they can survive a system crash / hardware failure. Adding that sort of synchronization overhead to every in-memory data structure that _might_ have to be echoed to disk would absolutely kill performance - and assumes the OS itself is flawless.
So yeah, an all-RAM OS would work. You just couldn't "reboot". Ever. Just suspend - better hope there are NO memory leaks, NO bugs, 100% reliable hardware - nothing that could crash the OS. Not even BSD claims to be this stable.
I just thought that someone could put out a patch for Linux 3.1, something to "accelerate" graphics by moving many GUI parts from the X server in user-mode to something (a driver?) in kernel-mode...
The truly scary part is, I can't rule it out! It actually could happen... (shudder)
Yeah, it would probably be funny as hell to the guy who puts it out... but he'd be lucky to survive the backlash. Something like that just ain't cool.
Maybe we could just skip over the 3 series and avoid all the MS/Windows references? Please?
The GUI that can successfully allow both to co-exist will be the killer GUI that everyone uses 50 years from now. Power user? Tap one key and you get the command-line, uber-file-manager system. Out of your league? Tap another key and grandma can use it. Or even a whole spectrum - tap the "easier" key until it's easy enough to use! Tap the "harder" key enough times, and you get asked for a root password...
The catch is, when will that GUI be invented? I think it's possible, but I haven't seen any GUI even in the ballpark yet. (The closest examples? Win+MS-DOS box, or better yet Gnome/KDE + terminal)
Reminds me of an article I ran across a little while ago...
The Air Force Academy (I don't know about the others) apparently keeps records of where all the students using their internet connections surf. As in, since they decided that porn was unbecoming of a student, they keep a list of the students "suspected of visiting known porn sites". This made the news because someone accidentally forwarded the list to the entire student body. Oops! If my memory serves (someone can search the Denver Post archives if they're interested), ~150 names were on that list... out of a student body of 4000.
The implications: the fact that they are keeping a list and NOT summarily kicking students out means that the AFA considers porn a problem - but a correctable problem. They would probably treat copyright issues similarly. Having computers seized would be pretty serious... or an over-reaction.
With code spread (properly) across a bunch of files, the limiting factor is usually disk speed. Load the compiler, load the files... that's sloshing a LOT of stuff around, and likely involves some swapping. Especially when you throw a fancy graphical debugger on top, and a graphical IDE... And if there's a web browser doing disk caching in the background as well, you take another performance hit.
Best bang for the buck? Tweak the hard drive configuration (keep disk accesses on different drives) and have LOADS of RAM. The processor ain't the bottleneck.
50 years from now when historians digs through 2002 e-mail logs, they'll probably think the most heavily consumed product in the country was (insert random spam product here).
Ah, the legacies we'll leave... based on YOUR e-mail, what will YOU be remembered as?
I wonder if the serial number / MAC address go only to the MS server. If they let this info get down to other clients as some sort of unique ID, they've just created a HUGE hole... (and now I'm wondering if this would be DMCA violation material). If so, this is a purely idiotic design - and call this an academic criticism.
Anyway, who's idea was it to make GUIDs have the MAC address inside? MS already has a love affair with GUIDs (just look at the mess called COM), but putting identifying information into a supposedly unique number strikes me as a really bad decision. I.e. one of the worst in history for privacy concerns. Something like this is a disaster waiting to happen.
Privacy and security are related. Because the first attacks against a secure system are to take multiple pieces of "private" information and cross-index it. Ignore privacy, and you've just weakened security. Think, peoples. Think.
Yeah, been bitten by that one. But that brings up a question I hope someone can answer for me:
Unicode. Since a unicode character is two bytes, is unicode big-endian or little-endian? I'd think XML would eventually move to fully unicode, but the advantage of a character string is that it avoids endian issues...
Nah, study physics and engineering for a year or so in college and this takes all of 15 minutes. (well, add another minute if I have to have my roommate double-check my facts...) Believe me, there is nothing here a decent physics student wouldn't know after a year or two. Besides, the fun part of physics is breaking theories - a la sci-fi.
That and I happened to read a book called "The Physics of Star Trek" (author was a prof at Berkeley), and have a good memory.
You don't seriously believe I patrol Slashdot just to talk Star Trek do you?:)
Proton torpedoes are Star Wars. Photon torpedoes are Star Trek. A proton torpedo doesn't even make sense - you can't do very many reactions with a proton. A photon torpedo is basically an antimatter charge. Antimatter + matter = lots of energy, in the form of photons!
Phase pistols. Made up - but only requires a particle that interferes with the strong nuclear force. Not too unreasonable.
Warp. It was actually Earth to Neptune in 6 minutes (as my roommate is remembering). First: Star Trek has never been consistent with the Warp scales, the original series was simply exponentially scaling while the later series were converging to Warp 10, and a lot of the "future" timelines seemed to switch back to the old scale. And Enterprise seems to use its own scale also. But: even traveling at warp, you have to add the time to drop in and out of warp, accelerate and deccelrate (within the limits of inertial damping - which doesn't exist!).
The maximum sustainable warp for Voyager is pretty low, as sustainable means engines running 24/7/365 AND no fuel concerns. Voyager has many problems with low fuel (how often do they look for antimatter?), and presumably takes the warp core offline periodically for maintenence (it would be a boring episode, so we don't see that. But there was a Next Generation episode that referenced maintenence).
But, Star Wars and Star Trek warp/hyperspace theories are difference (and neither are provable - though neither has been disproven!). Star Wars is alternate-universe-hyperspace, and you use energy to move into hyperspace and another chunk of energy to move out. Star Trek is a bubble of space where physics are different (i.e. the speed of light is much higher) and the bubble is maintained by a constant inflow of energy. And yes, Star Wars hyperdrive is faster than Star Trek warp drive - Star Wars hyperdrive would be closer to Star Trek Borg hyperdrive (which the Federation never could duplicate). But, Star Trek warp drive scales better - the theoretical top speed is much higher (you could get arbitrarily close to infinity).
And I don't know about that Wookie... I think the Klingon would give him a head-but the Wookie would never forget. And Klingon fighting is much more of a martial art - it might work fine against bigger adversaries. My opinion is whoever knocks the other guy down first would win...
Signed,
A Trekkie who rooms with a Star Wars fan / physicist
Latency, or the time to find the data on the drive, would be much faster - no mechanical drive parts involved.
Access time is latency + transfer speed. And transfer speed is going to be somewhat slower. (Ram = always at max speed (~132MB/s?). HD = up to max speed (~133MB/s w/ U-ATA 133), if from HD cache or contiguous sectors only - but for random access, it's typically ~2-5MB/s).
And the latencies of hard drives are on the orders of milliseconds, while RAM is on the order of tens of nanoseconds.
For those who bothered to follow the link clear back to the SEC filing, I found this section very interesting:
Intellectual Property Rights.....
From time to time Microsoft receives notices from others claiming Microsoft infringes their intellectual property rights. The number of these claims is expected to grow. Responding to these claims may require Microsoft to enter into royalty and licensing agreements on unfavorable terms,
require Microsoft to stop selling or to redesign affected products, or to pay damages or to satisfy indemnification commitments with the Company's customers.
Now, I may be overly cynical here, but it appears to me that MS has a business strategy: rip someone off, then when (err... if) that someone sues them, pay a pitance back, or discontinue the product line to avoid paying serious damages. In other words: cheat, then fix it when someone call you on the cheat. Unfortunately, it'd work - very few people can afford to challenge MS on much of anything and still force a settlement worth more than beans.
Optimization - a 10% slowdown will be magnified exponentially, and thus will be easier to find (and replicate)
Reliability - remember how everyone (er... competent admins) load-tests servers so that they could handle a Slashdot-effect load? Theoretically, someone could have an interest in running many, many concurrent applications under Wine - what better way to flush out the bugs in the system than to give it an extreme load?
Extensibility - a well-designed system will degrade gracefully under extreme loads. If it doesn't (i.e. it degrades exponentially), then the code probably needs to be reworked to be more efficient.
But things like this typically follow a scale:
1) does it work period? (i.e. can cygwin run under wine - 1 nesting)
2) does it work in the small-number case (i.e. 2-5 nestings, or thereabout)
3) does it work in the extreme case? (i.e. 10^(2-5) nestings) - which means that most inefficiency bugs are flushed out and the design scales well
Just about every system can fit into one of these categories - but only the most robust fit into #3. Example: Linux threading. Right now it passes 1 (you can multithread), passes 2 (having a number of threads/process under ~100 doesn't really change performance), but fails 3 (the 2.5 kernel developers are working on that one right now - but ~10,000 kernel threads will bring the system to its knees).
Re:Have we grown complacent?
on
Due Diligence?
·
· Score: 1
I couldn't agree more... When I first set up my new server (which I had been playing with for serveral months behind NAT) on the Internet, it got hit with Slapper w/in 24 hours. I had taken an update the night before, but didn't issue/sbin/service httpd restart - (foolish me - so used to Windows informing me I have to restart that I assumed Linux didn't) - woke up the next moring and hit ps, found 1hr of CPU time for a ".cinik" proces... and spent the better part of the morning cleaning up that. And the IRC server and backdoor that came over wget while I was deleting the files. Slapper gone w/in a few hours... (it took a while before I was sure there wasn't a root problem)
Then, a week later, I got a (very polite:) e-mail from a sysadmin in Slovakia telling me I had portscanned his machines, and I probably had a Slapper infection.
I keep wondering, if I hadn't hit "ps" that morning just to verify my new server was working right, would I have found the worm at all? Or would I have ended up with a rootkit?
The real reason I think Linux is so dangerous, and why I can't advocate anyone using it, it that Linux is so very easy to set up - yet it's a very powerful OS that requires a certain level of skill to secure. Honestly, it's like giving my mother a free copy of WinNT 4 and telling her to upgrade from Win3.1. Yes it's an upgrade, she gets something very useful for free, but I'm probably not doing the world a favor - I don't go around giving bazookas to kids either.
That's just it - every time I've tried to install, the X setup has either "just worked" or has required more time than I'm prepared to give. Win2K, on the other hand, "just works" at low resolution, which lets me find the drivers on the web I need to get high resolution to work.
Windows at least fails gracefully and I have a fighting chance of fixing it. But configuring X? The only resources I have there are man pages, info pages, and comments in config files - and if anyone thinks those are helpful to someone who's never configured X before, that someone needs to get out of their ivory tower. Chances are there's a simple issue I could solve by going out and finding some web page - but I am not (yet) capable of doing that from Linux's command line.
Why is it that every install of Linux recommends the most generic, mainstream hardware? Everyone claims its because that hardware has the most stable drivers. The side effect is, while Windows is willing to provide less-than-perfect support for specialized hardware, Linux basically demands that you either download a driver and compile it yourself (non-trivial, esp. if it's the network card) or hack something else to work (which most people don't have the skill to do - myself included).
In short: Windows drivers are pretty much at a 80% effective level, but fail gracefully. Linux drivers can reach 95%, but you have a (ballpark) 1 in 2 odds of having unsupported hardware and being at 5%. Bluntly, with a new system I'm more likely to get it working with Windows, and once it's working I'm less inclined to blow out the OS and try something new.
Don't get me wrong - I like Linux and use it a lot. But only in controlled circumstances - as a server (where I don't need X), or under VMWare (where the hardware is specified, and when it crashes I can move out to Windows to search for help).
But, as I'm procrastinating something else, I'll bite. Good trolls are hard to come by, and I've been itching to write a response like this...
Basically, if the US government had faked a moon landing, the Soviets would have had a field day proclaiming it loudly and to the entire world - rubbing in a blunder like that would be better for them than even winning the space race. And, even assuming I have no faith in the US government, I am very sure the 1960s spy network of the Soviets was efficient, and would have found out about a hoax, and would have called the US government on it. No, the Soviets were convinced, and if the Soviets were convinced without any good reason to trust the US anyway, that's good enough for me.
Unless you would care to maintain that the entire Cold War was a hoax?:)
Slashdot Mirror
And there was a space station called "Enterprise" somewhere between the space shuttle and the NX-1. I think it's one of the pictures in somebody's ready room, can't remember which. Yeah, there's a list of this somewhere... maybe the techical manual?
1 1 2 3 5 8 13 21 34 etc.
The ratio between each pair of terms is a progressively better approximation of the golden mean. Really.
If you have way too much time on your hands (as my Linear Algebra teacher once did), you can construct a non-iterative formula for the sequence out of this. Requires diagonalization, messy enough you actually have to do it by hand... but it actually works.
So, the golden ratio is pretty easy to approximate. Nature's done it for millions of years. You just haven't tried hard enough... :)
(#45)
http://www.google.com/search?q=http&hl=en&lr=lang_ en&ie=UTF-8&oe=UTF-8&start=40&sa=N
This is High Treason! Mod him down! ;-)
It seriously takes 3-4 seconds for an "open file" dialog box to come up (Solaris runtime). The second time it's down to maybe 1/3 of a second (thank you hotspot!), and I'd bet I could take that time out by caching the dialog instead of making a new one each time. It's that first pass being way too slow though...
Hotspot in of itself is pretty good. A year and a half ago, I was cursing Java for slow everything - now it's just the startup time and memory footprint - the rest of the language has shaped up nicely. Another year of improvements and I just might switch... :-)
Never gotten to play with one of those native compilers :-(. But I'd bet a good one would take out half my complaints right now! However, I just don't feel the memory footprint is likely to go down.
Hotspot server is better? Very interesting... Thanks for the info!
Just realized what I just said. Java is better the longer the program runs? Very, very strange... /shudder/. Meaning, Java is exactly the WRONG thing to use for an applet and the RIGHT thing to use for a heavyweight editor!
1) Hotspot: definitely interprets TODAY. Designed to interpret. Just intelligently compiles the most-commonly used code. And yes, it's a major speed boost to do things this way. But tight C/C++ would be faster. Roughly, all Java ~ mediocre C++, which means it's easier to write good Java, but Java won't solve all problems.
2) Trying to compile everything (bytecode->native) beforehand is what makes the JVM so slow. A programming language that loads 40Megs of memory for a GUI and takes several seconds to load (on a fast machine!) is bad. And then you close the program, and lose all the gains running it and compiling efficiently earned you...
I'd think a big problem with this idea would be keeping the data stable. With most OSes, you can just reboot and that'll reset all the in-memory data structures. If one of those very-important structures gets corrupted, the entire OS becomes unusable. Can't flush things, can't restore stability = Kernel Panic, or the like - the OS kills itself to save the data on disk (because corrupted kernel info can cause corrupted writes...). Since it's (comparatively) easy to reboot the OS and re-create all the data structures, it's a good trade-off.
A LOT of work goes into keeping the disk in a stable state. Just look at the variety of filesystems out there right now - anything interesting in the past 30 years has been concerned with writing consistent data and crash-proofing disks, so that they can survive a system crash / hardware failure. Adding that sort of synchronization overhead to every in-memory data structure that _might_ have to be echoed to disk would absolutely kill performance - and assumes the OS itself is flawless.
So yeah, an all-RAM OS would work. You just couldn't "reboot". Ever. Just suspend - better hope there are NO memory leaks, NO bugs, 100% reliable hardware - nothing that could crash the OS. Not even BSD claims to be this stable.
I just thought that someone could put out a patch for Linux 3.1, something to "accelerate" graphics by moving many GUI parts from the X server in user-mode to something (a driver?) in kernel-mode...
The truly scary part is, I can't rule it out! It actually could happen... (shudder)
Yeah, it would probably be funny as hell to the guy who puts it out... but he'd be lucky to survive the backlash. Something like that just ain't cool.
Maybe we could just skip over the 3 series and avoid all the MS/Windows references? Please?
The GUI that can successfully allow both to co-exist will be the killer GUI that everyone uses 50 years from now. Power user? Tap one key and you get the command-line, uber-file-manager system. Out of your league? Tap another key and grandma can use it. Or even a whole spectrum - tap the "easier" key until it's easy enough to use! Tap the "harder" key enough times, and you get asked for a root password...
The catch is, when will that GUI be invented? I think it's possible, but I haven't seen any GUI even in the ballpark yet. (The closest examples? Win+MS-DOS box, or better yet Gnome/KDE + terminal)
The Air Force Academy (I don't know about the others) apparently keeps records of where all the students using their internet connections surf. As in, since they decided that porn was unbecoming of a student, they keep a list of the students "suspected of visiting known porn sites". This made the news because someone accidentally forwarded the list to the entire student body. Oops! If my memory serves (someone can search the Denver Post archives if they're interested), ~150 names were on that list... out of a student body of 4000.
The implications: the fact that they are keeping a list and NOT summarily kicking students out means that the AFA considers porn a problem - but a correctable problem. They would probably treat copyright issues similarly. Having computers seized would be pretty serious... or an over-reaction.
Best bang for the buck? Tweak the hard drive configuration (keep disk accesses on different drives) and have LOADS of RAM. The processor ain't the bottleneck.
Whenever I tried to conquer the world, I always switched to fundamentalism...
Oh.
THIS is why we need export restrictions on computer games. :)
If I lose my keys, do I need to pay for a new car?
Ah, the legacies we'll leave... based on YOUR e-mail, what will YOU be remembered as?
Anyway, who's idea was it to make GUIDs have the MAC address inside? MS already has a love affair with GUIDs (just look at the mess called COM), but putting identifying information into a supposedly unique number strikes me as a really bad decision. I.e. one of the worst in history for privacy concerns. Something like this is a disaster waiting to happen.
Privacy and security are related. Because the first attacks against a secure system are to take multiple pieces of "private" information and cross-index it. Ignore privacy, and you've just weakened security. Think, peoples. Think.
Unicode. Since a unicode character is two bytes, is unicode big-endian or little-endian? I'd think XML would eventually move to fully unicode, but the advantage of a character string is that it avoids endian issues...
That and I happened to read a book called "The Physics of Star Trek" (author was a prof at Berkeley), and have a good memory.
You don't seriously believe I patrol Slashdot just to talk Star Trek do you? :)
Proton torpedoes are Star Wars. Photon torpedoes are Star Trek. A proton torpedo doesn't even make sense - you can't do very many reactions with a proton. A photon torpedo is basically an antimatter charge. Antimatter + matter = lots of energy, in the form of photons!
Phase pistols. Made up - but only requires a particle that interferes with the strong nuclear force. Not too unreasonable.
Warp. It was actually Earth to Neptune in 6 minutes (as my roommate is remembering). First: Star Trek has never been consistent with the Warp scales, the original series was simply exponentially scaling while the later series were converging to Warp 10, and a lot of the "future" timelines seemed to switch back to the old scale. And Enterprise seems to use its own scale also. But: even traveling at warp, you have to add the time to drop in and out of warp, accelerate and deccelrate (within the limits of inertial damping - which doesn't exist!).
The maximum sustainable warp for Voyager is pretty low, as sustainable means engines running 24/7/365 AND no fuel concerns. Voyager has many problems with low fuel (how often do they look for antimatter?), and presumably takes the warp core offline periodically for maintenence (it would be a boring episode, so we don't see that. But there was a Next Generation episode that referenced maintenence).
But, Star Wars and Star Trek warp/hyperspace theories are difference (and neither are provable - though neither has been disproven!). Star Wars is alternate-universe-hyperspace, and you use energy to move into hyperspace and another chunk of energy to move out. Star Trek is a bubble of space where physics are different (i.e. the speed of light is much higher) and the bubble is maintained by a constant inflow of energy. And yes, Star Wars hyperdrive is faster than Star Trek warp drive - Star Wars hyperdrive would be closer to Star Trek Borg hyperdrive (which the Federation never could duplicate). But, Star Trek warp drive scales better - the theoretical top speed is much higher (you could get arbitrarily close to infinity).
And I don't know about that Wookie... I think the Klingon would give him a head-but the Wookie would never forget. And Klingon fighting is much more of a martial art - it might work fine against bigger adversaries. My opinion is whoever knocks the other guy down first would win...
Signed,
A Trekkie who rooms with a Star Wars fan / physicist
Latency, or the time to find the data on the drive, would be much faster - no mechanical drive parts involved.
Access time is latency + transfer speed. And transfer speed is going to be somewhat slower. (Ram = always at max speed (~132MB/s?). HD = up to max speed (~133MB/s w/ U-ATA 133), if from HD cache or contiguous sectors only - but for random access, it's typically ~2-5MB/s).
And the latencies of hard drives are on the orders of milliseconds, while RAM is on the order of tens of nanoseconds.
But things like this typically follow a scale:
1) does it work period? (i.e. can cygwin run under wine - 1 nesting)
2) does it work in the small-number case (i.e. 2-5 nestings, or thereabout)
3) does it work in the extreme case? (i.e. 10^(2-5) nestings) - which means that most inefficiency bugs are flushed out and the design scales well
Just about every system can fit into one of these categories - but only the most robust fit into #3. Example: Linux threading. Right now it passes 1 (you can multithread), passes 2 (having a number of threads/process under ~100 doesn't really change performance), but fails 3 (the 2.5 kernel developers are working on that one right now - but ~10,000 kernel threads will bring the system to its knees).
Then, a week later, I got a (very polite :) e-mail from a sysadmin in Slovakia telling me I had portscanned his machines, and I probably had a Slapper infection.
I keep wondering, if I hadn't hit "ps" that morning just to verify my new server was working right, would I have found the worm at all? Or would I have ended up with a rootkit?
The real reason I think Linux is so dangerous, and why I can't advocate anyone using it, it that Linux is so very easy to set up - yet it's a very powerful OS that requires a certain level of skill to secure. Honestly, it's like giving my mother a free copy of WinNT 4 and telling her to upgrade from Win3.1. Yes it's an upgrade, she gets something very useful for free, but I'm probably not doing the world a favor - I don't go around giving bazookas to kids either.
Windows at least fails gracefully and I have a fighting chance of fixing it. But configuring X? The only resources I have there are man pages, info pages, and comments in config files - and if anyone thinks those are helpful to someone who's never configured X before, that someone needs to get out of their ivory tower. Chances are there's a simple issue I could solve by going out and finding some web page - but I am not (yet) capable of doing that from Linux's command line.
Why is it that every install of Linux recommends the most generic, mainstream hardware? Everyone claims its because that hardware has the most stable drivers. The side effect is, while Windows is willing to provide less-than-perfect support for specialized hardware, Linux basically demands that you either download a driver and compile it yourself (non-trivial, esp. if it's the network card) or hack something else to work (which most people don't have the skill to do - myself included).
In short: Windows drivers are pretty much at a 80% effective level, but fail gracefully. Linux drivers can reach 95%, but you have a (ballpark) 1 in 2 odds of having unsupported hardware and being at 5%. Bluntly, with a new system I'm more likely to get it working with Windows, and once it's working I'm less inclined to blow out the OS and try something new.
Don't get me wrong - I like Linux and use it a lot. But only in controlled circumstances - as a server (where I don't need X), or under VMWare (where the hardware is specified, and when it crashes I can move out to Windows to search for help).
Really. I wanted to see an innovative use of duct tape on a Mac.
Lousy Slashdot editors. Convincing me to read a story when there's actually no duct tape involved.
But, as I'm procrastinating something else, I'll bite. Good trolls are hard to come by, and I've been itching to write a response like this...
Basically, if the US government had faked a moon landing, the Soviets would have had a field day proclaiming it loudly and to the entire world - rubbing in a blunder like that would be better for them than even winning the space race. And, even assuming I have no faith in the US government, I am very sure the 1960s spy network of the Soviets was efficient, and would have found out about a hoax, and would have called the US government on it. No, the Soviets were convinced, and if the Soviets were convinced without any good reason to trust the US anyway, that's good enough for me.
Unless you would care to maintain that the entire Cold War was a hoax? :)