Granted, I'm posting this after there are already 300+ posts, so I doubt anyone will ever see this, but I'll mention it anyways.
There's a very simple and easy way to have the natural effect of the karma cap w/o actually having one. Once your karma reaches a certain number (say 25, because that's where the +1 kicks in), make it so that things that would give you a +1 karma only have a percentage chance of giving you that +1, and the percentage drops as your karma goes up. Alternatively if you didn't want randomness but didn't mind fractional karma, just award a fractional amount of karma, with that fraction tending to 0 as karma -> infinity.
What would this do? Suppose we have a poster that tends to average 2 +1s for every -1. Under the uncapped system that poster's karma would go to infinity. Under the cap they'd slam into the cap. Under this proposal, their karma would rise slower and slower until they reach around where they're getting 0.5 for every +1, and would stabilize there. Whatever this number would be, it'd be a lower fixed number than the person who averages 10 +1s for every -1 -- a feature that neither unlimited or capped karma has.
This does nothing to eliminate karma whoring, but it does make it less and less worthwhile over time. The only way to do anything like that is to change the +1/-1 system itself, and I don't have an answer for that one.
This reminds me of the fun I have with Radio Shack asking for my name/address/etc. When I lived in Indiana I would regularly tell them I'm "Richard Fromage, 1060 W. Addison, Chicago IL, 60613"-- if that address doesn't ring a bell, go rent the (original) Blues Brothers movie.
Makes me wonder how much junk mail the Chicago Cubs get and routinely dispose of...
It's a fairly simple and straightforward sysadmin exercise to make a script (well, one line, really) to delete all files in/tmp that haven't been accessed in say 3 days or so. (It involves reading the manpage on 'find'.)
Heck, at cc.purdue.edu they put quotas on/tmp at one point, just to ensure that students wouldn't fill up/tmp and bring down the machine.
It seems to me that a large part of the reason for bogus scenarios like the Gabriel Knight 3 one listed is: lack of playtesting.
There are numerous stories back in the Zork days of players playing Zork and thinking of something that the designers never thought of, and the designers going back and fixing things to make them better. For example, instead of grues, Zork originally had bottomless pits. If you wandered into a dark room without a light, you had a good chance of falling into a bottomless pit. Players soon pointed out, however, that if there was a bottomless pit on the second floor of the white house, you'd probably be able to see it from the first floor. Hence the invention of the grue (notionally, the 'dungeon master' went around the whole dungeon filling up the bottomless pits, but this displaced the grues, who now wander the darkness looking for food.)
Another such example was the troll. As everyone knows, the troll had an axe that he would attack the player with. Players complained that when they killed the troll, the troll's axe would disappear too, and finally the designers said 'ok, ok, you can get the troll's axe now.'-- of course what they did was just code up the axe as another weapon. However, then one of the players said 'great, I can get the axe now, I'm going to go to the forest and chop down the trees.'
The point of this all being: designing good adventure games is Hard; you're Not Going To Think Of Everything. Best to let independent eyes and brains work on it for you some; you'll find things you never would have thought of.
Oh fuck off. There is no huge corporate conspiracy to keep whatever shit home-town folk band you like off the airwaves. The reason that the music you like is not popular is that it is shit. The record companies have no secret mind control formula which is not accessible to anyone else; what they do have is a skill in producing a product (the "marketing" is part of the product; people like to have "stars" rather than bearded recorder playing granolas). Your music is shit. Deal with it.
So, by this rationale, the only good music is the music that sells the most? Ergo, Britney Spears, the Backstreet Boys and N'Sync are the best music of the 90s?
What I'd like to see is sales figures for albums that ignores all sales in the first 5 or 10 years or so since the album was released. I.e. how well does the album sell once the marketing hype wears off, and how well does it last? Just for the sake of example, Michael Jackson's _Thriller_ may have outsold Dark Side of the Moon, but I'll bet DSotM outsells Thriller at least 5 to 1 nowadays.
What we really need is approval voting. What this means: you get a list of all the candidates for office, and you vote yes or no on each one. The candidate with the highest number of yes votes wins the election.
This lets people express their vote much better than standard methods. Do you like Nader but don't think he has a chance to win, and would rather not see Bush win? Vote yes to Gore and Nader and no to everyone else. Hate the major parties? Vote no to Bush AND Gore, and yes to everyone else. Fed up with the whole process? Vote no to everyone. This last option actually has influence under approval voting-- imagine how it would look if the winning candidate was approved by less than 30 percent of the voters. A side benefit is that it encourages candidates to not engage in negative campaigning, since it's easier to increase their own approval than decrease everyone else's.
The best part is that implementing this doesn't require changing the Constitution, or anything so drastic. Local voting laws are all that need to be changed. Personally I believe that all we'd need is one state to implement state-wide approval voting, and the natural publicity from that would take it from there.
Kinda for the same reason I can't hold a party in which is sponsors where I make money and play a bunch of music foreveryone, even if I check to see if they bought the cd's themselves.
I don't know what world you live in, but my people call them bars or dance clubs.
Personally, in this case I don't mind CmdrTaco's comments with the story. Just think of it this way: it kept 100 or more people from posting them instead. The argument the corporate droid gave is obviously bogus, so best to knock it down straightaway and then we can argue how bogus it is instead of whether it is bogus.:-)
Er, the use of 'lapped' in the article title is kind of misleading. In auto racing, if you've lapped someone, it means you have a lead on them equal to the entire racecourse, and you've passed them again. It's important on oval-type tracks because when a yellow light comes on (from an accident or whatever), all the cars on the same lap bunch back up together again-- but the lapped cars are still a lap behind. A full lap is a Significant Lead and is not easy to overcome. To say that Linux has lapped Apple is thus an entirely different statement than saying that Linux has passed Apple.
Now, arguably, you could say that Microsoft had lapped the entire field in terms of the desktop market, and that perhaps now Linux is threatening to get on the lead lap with Microsoft.
As the AC stated, there are patterns in the prime numbers, but what it seems you're asking is, is there a pattern that all the prime numbers fall into. Or, to put it another way, given all the prime numbers from 2 to n, find k=f(n) such that n+1 thru k-1 are composite but k is prime. Personally I doubt this one will ever be solved.
Heck, we haven't even been able to solve relatively minor prime-related problems, like the twin primes conjecture: Are there infinitely many pairs of primes of the form n and n+2? Or a followup, which I thought of but have no idea if anyone's done any research on, which I call 'quad primes': pairs of twin primes of the form n, n+2, n+6, n+8. It's easy to show that except for 3 and 5, all twin primes are of the form 6k-1 and 6k+1, and similarly that except for 5-7-11-13 all quad primes are of the form 30k+(11,13,17,19). But past that, who knows?
Perhaps it's time to revive the old VMS idea, which is supposed to be in the GNU kernel, of file versioning on every file. How many times have you gone into a directory and seen something like blah.cfg.old, blah.cfg.OLD, blah.cfg.goodcopy, etc.? Arguably something more sophisticated than sticking a number on the end is needed, but it would be a start at least.
One detail that is generally overlooked is how NeXTStep ran on multiple hardware platforms-- NeXT hardware, Intel, Sun, and HP. And the coolest feature of all was that under NeXTStep 3, binaries could be compiled 'fat' and they would run on any of those platforms. That's right, the same binary running on vastly different hardware. Personally, I think it deserves a nomination to the 'Greatest Hacks Of All Time' list-- well, an honorable mention at least.
GUI tools are like any other tool: they can be good or evil, depending on how they work.
If the GUI tool provides an easy way to do things, provides all the options there in a convienent package, etc., and doesn't break when the underlying config files change because you edited them by hand when the system is in single-user mode, then it is a Good Tool. Example: veritas's vmsa, which among other things lets you create a vxfs filesystem with a few clicks, in exactly the way you want it (a process that takes many more steps when done from the command line).
If the GUI tool makes itself The Only Way This Can Be Done, and prevents editing the underlying config files by hand, or breaks the system, or other Unwanted Things, then it is a Bad Tool. Example: AIX's smit, which is good for many things but takes over/etc/inittab so that if you change it by hand, it quietly undoes your changes and you're left scratching your head wondering what happened and why your changes disappeared on you.
From what I've seen so far, the Helix tools sound like Good Tools, and should be praised appropriately.
I've played enough pinball simulators to know-- until force feedback tech gets a lot better than it is now, pinball emulators will always be a sad shadow of the real thing. Any good player can feel the kickers and the bumpers when they fire, can nudge the machine when they fire to put some extra kick into it, and so forth. I've seen some decently-designed tables in emulators (Timeshock comes to mind), but until force-feedback gets better than just the controller shaking in your hand (ala the pinball machine Earthshaker from 1989), we'll never have good pinball emulation.
Ask a race driver about the force feedback from the video games Hard Drivin' and Race Drivin'; they'll tell you it's a bit better than say Pole Position, but still far short. Same concept here.
The advantage that Europe and Japan have over the US is that they have greater population density, making technologies that can take advantage of it, such as wireless and public rail transportation, much more cost-effective. It's a simple matter of geography, nothing more.
I just went out and bought a bunch of the funky controllers that are available for the PSX. I have 2 big heavy Namco controllers with real joysticks (essential for Robotron), a mouse, a trackball, and an analog-joystick controller. This works pretty well for most games, even in some ways you wouldn't expect (the trackball actually makes a really good substitute for Tempest's paddle), but not really for all of them-- nothing will ever replace Stargate's wacked 6-button-plus-joystick setup, for example. If anyone's interested in hearing more, send me email (no, it's not clickable, look it up thru my user info) and I'll let you know what works and doesn't work for me.
It happened a couple of years ago for the PlayStation, and has been going strong ever since. There are about a dozen or so 'collection' games of old arcade classics, and there are a bunch of 'new' versions of old games-- just off the top of my head, I own the new version of: Xevious, Pong (really!), Q*Bert, Frogger, Centipede, Missile Command, Tempest, Robotron, and Space Invaders. Several of these include the original version as one of the play options. Heck, I originally bought a Playstation so I could play the original Bosconian off Namco 1-- ok, I'll admit it was for Bust-A-Move as well, but I would have bought it just for Bosconian.:-)
Everyone has heard this before: "You should send all your email encrypted, not just the important stuff, so that (insert many good reasons here)".
In a perfect world, this would in fact be the optimal solution. But, realistically, since this isn't happening any time soon, how about some other options that at least get us part of the way there?
1) Add encryption into sendmail's transmission of mail. The goal here is to have the actual email traffic sent over the open wire encrypted, like ssh traffic is. OpenSSL could be used for this. Of course this does nothing to protect you if your computer is cracked or stolen (including by the gov't), hence:
2) Store mail (in/var/mail) in an encrypted form. Again, this involves adding something onto sendmail, this time when the mail is written into the mail spool. This would be a bit more tricky than 1), since user mail programs would either have to be able to do the decryption or call some agent to do it for them.
Despite the complications in 2), these still both seem very doable to me. 1) would require extending the SMTP protocol slightly, so that mailers can talk to each other and send things between each other in an encrypted form only if both are encrypted-capable. Has anyone else thought of doing something like this before? Is it as feasible as I think it is (or am I just low on sleep and not thinking clearly)?
Tom Scholz, the guy behind the band Boston, has a master's in EE from MIT. Tom Lehrer, one of the mainstays on Dr. Demento, has a bachelor's and master's in math from Harvard. I can't recall his name (Dexter?), but the lead singer from The Offspring (yes, the same band that praised Napster and then proceeded to show Napster's hypocracy by selling Napster shirts and getting told by Napster to stop) has a biochem degree from (I think it is) Emory.
Anyone have other examples? It's not nearly as uncommon as you'd think.
What with all the news lately about Napster vs. RIAA and 2600 vs. MPAA etc, it's good to actually see some genuine good news for a change. Of course, this isn't the end-all and be-all, but I never expect that. As a wise man once said, "The price of freedom is eternal vigilance."
Any computer system of any size needs a sysadmin for one obvious reason, at least: it frees up the users to use the computer, and not have to worry about keeping the thing running. If there are 10 users on a system and each is spending 10 percent of their time keeping the system running, then they are already devoting a full person's time keeping the system going. Better to let one person do that and let the others spend their full time using the computer.
It's common in any large group: people specialize in what they're best at. It's been going on since the beginning of time: towns didn't have everyone dabbling as a blacksmith, or a preacher, or a mortician, even though anyone could do those things. Generally even very small towns usually had one of each. The same applies with sysadmins.
A good friend of mine just yesterday went to see what version FreeBSD was up to, went "holy smokes they're on 4.0 already" and sent off his order to buy 4.0 on cd-rom.
Slashdot Mirror
Granted, I'm posting this after there are already 300+ posts, so I doubt anyone will ever see this, but I'll mention it anyways.
There's a very simple and easy way to have the natural effect of the karma cap w/o actually having one. Once your karma reaches a certain number (say 25, because that's where the +1 kicks in), make it so that things that would give you a +1 karma only have a percentage chance of giving you that +1, and the percentage drops as your karma goes up. Alternatively if you didn't want randomness but didn't mind fractional karma, just award a fractional amount of karma, with that fraction tending to 0 as karma -> infinity.
What would this do? Suppose we have a poster that tends to average 2 +1s for every -1. Under the uncapped system that poster's karma would go to infinity. Under the cap they'd slam into the cap. Under this proposal, their karma would rise slower and slower until they reach around where they're getting 0.5 for every +1, and would stabilize there. Whatever this number would be, it'd be a lower fixed number than the person who averages 10 +1s for every -1 -- a feature that neither unlimited or capped karma has.
This does nothing to eliminate karma whoring, but it does make it less and less worthwhile over time. The only way to do anything like that is to change the +1/-1 system itself, and I don't have an answer for that one.
---
This reminds me of the fun I have with Radio Shack asking for my name/address/etc. When I lived in Indiana I would regularly tell them I'm "Richard Fromage, 1060 W. Addison, Chicago IL, 60613"-- if that address doesn't ring a bell, go rent the (original) Blues Brothers movie.
Makes me wonder how much junk mail the Chicago Cubs get and routinely dispose of...
---
It's a fairly simple and straightforward sysadmin exercise to make a script (well, one line, really) to delete all files in /tmp that haven't been accessed in say 3 days or so. (It involves reading the manpage on 'find'.)
/tmp at one point, just to ensure that students wouldn't fill up /tmp and bring down the machine.
Heck, at cc.purdue.edu they put quotas on
---
It seems to me that a large part of the reason for bogus scenarios like the Gabriel Knight 3 one listed is: lack of playtesting.
There are numerous stories back in the Zork days of players playing Zork and thinking of something that the designers never thought of, and the designers going back and fixing things to make them better. For example, instead of grues, Zork originally had bottomless pits. If you wandered into a dark room without a light, you had a good chance of falling into a bottomless pit. Players soon pointed out, however, that if there was a bottomless pit on the second floor of the white house, you'd probably be able to see it from the first floor. Hence the invention of the grue (notionally, the 'dungeon master' went around the whole dungeon filling up the bottomless pits, but this displaced the grues, who now wander the darkness looking for food.)
Another such example was the troll. As everyone knows, the troll had an axe that he would attack the player with. Players complained that when they killed the troll, the troll's axe would disappear too, and finally the designers said 'ok, ok, you can get the troll's axe now.'-- of course what they did was just code up the axe as another weapon. However, then one of the players said 'great, I can get the axe now, I'm going to go to the forest and chop down the trees.'
The point of this all being: designing good adventure games is Hard; you're Not Going To Think Of Everything. Best to let independent eyes and brains work on it for you some; you'll find things you never would have thought of.
---
Oh fuck off. There is no huge corporate conspiracy to keep whatever shit home-town folk band you like off the airwaves. The reason that the music you like is not popular is that it is shit. The record companies have no secret mind control formula which is not accessible to anyone else; what they do have is a skill in producing a product (the "marketing" is part of the product; people like to have "stars" rather than bearded recorder playing granolas). Your music is shit. Deal with it.
So, by this rationale, the only good music is the music that sells the most? Ergo, Britney Spears, the Backstreet Boys and N'Sync are the best music of the 90s?
What I'd like to see is sales figures for albums that ignores all sales in the first 5 or 10 years or so since the album was released. I.e. how well does the album sell once the marketing hype wears off, and how well does it last? Just for the sake of example, Michael Jackson's _Thriller_ may have outsold Dark Side of the Moon, but I'll bet DSotM outsells Thriller at least 5 to 1 nowadays.
---
What we really need is approval voting. What this means: you get a list of all the candidates for office, and you vote yes or no on each one. The candidate with the highest number of yes votes wins the election.
This lets people express their vote much better than standard methods. Do you like Nader but don't think he has a chance to win, and would rather not see Bush win? Vote yes to Gore and Nader and no to everyone else. Hate the major parties? Vote no to Bush AND Gore, and yes to everyone else. Fed up with the whole process? Vote no to everyone. This last option actually has influence under approval voting-- imagine how it would look if the winning candidate was approved by less than 30 percent of the voters. A side benefit is that it encourages candidates to not engage in negative campaigning, since it's easier to increase their own approval than decrease everyone else's.
The best part is that implementing this doesn't require changing the Constitution, or anything so drastic. Local voting laws are all that need to be changed. Personally I believe that all we'd need is one state to implement state-wide approval voting, and the natural publicity from that would take it from there.
---
Kinda for the same reason I can't hold a party in which is sponsors where I make money and play a bunch of music foreveryone, even if I check to see if they bought the cd's themselves.
I don't know what world you live in, but my people call them bars or dance clubs.
---
Personally, in this case I don't mind CmdrTaco's comments with the story. Just think of it this way: it kept 100 or more people from posting them instead. The argument the corporate droid gave is obviously bogus, so best to knock it down straightaway and then we can argue how bogus it is instead of whether it is bogus. :-)
---
The actual story is here -- the url given is a page that's "today's" news, which is now yesterday's news.
---
Er, the use of 'lapped' in the article title is kind of misleading. In auto racing, if you've lapped someone, it means you have a lead on them equal to the entire racecourse, and you've passed them again. It's important on oval-type tracks because when a yellow light comes on (from an accident or whatever), all the cars on the same lap bunch back up together again-- but the lapped cars are still a lap behind. A full lap is a Significant Lead and is not easy to overcome. To say that Linux has lapped Apple is thus an entirely different statement than saying that Linux has passed Apple.
Now, arguably, you could say that Microsoft had lapped the entire field in terms of the desktop market, and that perhaps now Linux is threatening to get on the lead lap with Microsoft.
---
I wonder, when Mac OS X is released and added to that table, what it will look like?
:-)
Probably entry after entry of 'click on this'
---
As the AC stated, there are patterns in the prime numbers, but what it seems you're asking is, is there a pattern that all the prime numbers fall into. Or, to put it another way, given all the prime numbers from 2 to n, find k=f(n) such that n+1 thru k-1 are composite but k is prime. Personally I doubt this one will ever be solved.
Heck, we haven't even been able to solve relatively minor prime-related problems, like the twin primes conjecture: Are there infinitely many pairs of primes of the form n and n+2? Or a followup, which I thought of but have no idea if anyone's done any research on, which I call 'quad primes': pairs of twin primes of the form n, n+2, n+6, n+8. It's easy to show that except for 3 and 5, all twin primes are of the form 6k-1 and 6k+1, and similarly that except for 5-7-11-13 all quad primes are of the form 30k+(11,13,17,19). But past that, who knows?
---
Perhaps it's time to revive the old VMS idea, which is supposed to be in the GNU kernel, of file versioning on every file. How many times have you gone into a directory and seen something like blah.cfg.old, blah.cfg.OLD, blah.cfg.goodcopy, etc.? Arguably something more sophisticated than sticking a number on the end is needed, but it would be a start at least.
---
One detail that is generally overlooked is how NeXTStep ran on multiple hardware platforms-- NeXT hardware, Intel, Sun, and HP. And the coolest feature of all was that under NeXTStep 3, binaries could be compiled 'fat' and they would run on any of those platforms. That's right, the same binary running on vastly different hardware. Personally, I think it deserves a nomination to the 'Greatest Hacks Of All Time' list-- well, an honorable mention at least.
---
GUI tools are like any other tool: they can be good or evil, depending on how they work.
/etc/inittab so that if you change it by hand, it quietly undoes your changes and you're left scratching your head wondering what happened and why your changes disappeared on you.
If the GUI tool provides an easy way to do things, provides all the options there in a convienent package, etc., and doesn't break when the underlying config files change because you edited them by hand when the system is in single-user mode, then it is a Good Tool. Example: veritas's vmsa, which among other things lets you create a vxfs filesystem with a few clicks, in exactly the way you want it (a process that takes many more steps when done from the command line).
If the GUI tool makes itself The Only Way This Can Be Done, and prevents editing the underlying config files by hand, or breaks the system, or other Unwanted Things, then it is a Bad Tool. Example: AIX's smit, which is good for many things but takes over
From what I've seen so far, the Helix tools sound like Good Tools, and should be praised appropriately.
---
I've played enough pinball simulators to know-- until force feedback tech gets a lot better than it is now, pinball emulators will always be a sad shadow of the real thing. Any good player can feel the kickers and the bumpers when they fire, can nudge the machine when they fire to put some extra kick into it, and so forth. I've seen some decently-designed tables in emulators (Timeshock comes to mind), but until force-feedback gets better than just the controller shaking in your hand (ala the pinball machine Earthshaker from 1989), we'll never have good pinball emulation.
Ask a race driver about the force feedback from the video games Hard Drivin' and Race Drivin'; they'll tell you it's a bit better than say Pole Position, but still far short. Same concept here.
---
The advantage that Europe and Japan have over the US is that they have greater population density, making technologies that can take advantage of it, such as wireless and public rail transportation, much more cost-effective. It's a simple matter of geography, nothing more.
---
I just went out and bought a bunch of the funky controllers that are available for the PSX. I have 2 big heavy Namco controllers with real joysticks (essential for Robotron), a mouse, a trackball, and an analog-joystick controller. This works pretty well for most games, even in some ways you wouldn't expect (the trackball actually makes a really good substitute for Tempest's paddle), but not really for all of them-- nothing will ever replace Stargate's wacked 6-button-plus-joystick setup, for example. If anyone's interested in hearing more, send me email (no, it's not clickable, look it up thru my user info) and I'll let you know what works and doesn't work for me.
---
Winston Churchhill once said "Democracy would be the worst form of government, if it wasn't for all the others."
Seems to me that Mr. Moody's statement is incomplete-- Linux would be the worst OS, if it wasn't for all the others.
---
It happened a couple of years ago for the PlayStation, and has been going strong ever since. There are about a dozen or so 'collection' games of old arcade classics, and there are a bunch of 'new' versions of old games-- just off the top of my head, I own the new version of: Xevious, Pong (really!), Q*Bert, Frogger, Centipede, Missile Command, Tempest, Robotron, and Space Invaders. Several of these include the original version as one of the play options. Heck, I originally bought a Playstation so I could play the original Bosconian off Namco 1-- ok, I'll admit it was for Bust-A-Move as well, but I would have bought it just for Bosconian. :-)
---
Everyone has heard this before: "You should send all your email encrypted, not just the important stuff, so that (insert many good reasons here)".
/var/mail) in an encrypted form. Again, this involves adding something onto sendmail, this time when the mail is written into the mail spool. This would be a bit more tricky than 1), since user mail programs would either have to be able to do the decryption or call some agent to do it for them.
In a perfect world, this would in fact be the optimal solution. But, realistically, since this isn't happening any time soon, how about some other options that at least get us part of the way there?
1) Add encryption into sendmail's transmission of mail. The goal here is to have the actual email traffic sent over the open wire encrypted, like ssh traffic is. OpenSSL could be used for this. Of course this does nothing to protect you if your computer is cracked or stolen (including by the gov't), hence:
2) Store mail (in
Despite the complications in 2), these still both seem very doable to me. 1) would require extending the SMTP protocol slightly, so that mailers can talk to each other and send things between each other in an encrypted form only if both are encrypted-capable. Has anyone else thought of doing something like this before? Is it as feasible as I think it is (or am I just low on sleep and not thinking clearly)?
---
Tom Scholz, the guy behind the band Boston, has a master's in EE from MIT. Tom Lehrer, one of the mainstays on Dr. Demento, has a bachelor's and master's in math from Harvard. I can't recall his name (Dexter?), but the lead singer from The Offspring (yes, the same band that praised Napster and then proceeded to show Napster's hypocracy by selling Napster shirts and getting told by Napster to stop) has a biochem degree from (I think it is) Emory.
Anyone have other examples? It's not nearly as uncommon as you'd think.
---
What with all the news lately about Napster vs. RIAA and 2600 vs. MPAA etc, it's good to actually see some genuine good news for a change. Of course, this isn't the end-all and be-all, but I never expect that. As a wise man once said, "The price of freedom is eternal vigilance."
---
Any computer system of any size needs a sysadmin for one obvious reason, at least: it frees up the users to use the computer, and not have to worry about keeping the thing running. If there are 10 users on a system and each is spending 10 percent of their time keeping the system running, then they are already devoting a full person's time keeping the system going. Better to let one person do that and let the others spend their full time using the computer.
It's common in any large group: people specialize in what they're best at. It's been going on since the beginning of time: towns didn't have everyone dabbling as a blacksmith, or a preacher, or a mortician, even though anyone could do those things. Generally even very small towns usually had one of each. The same applies with sysadmins.
---
A good friend of mine just yesterday went to see what version FreeBSD was up to, went "holy smokes they're on 4.0 already" and sent off his order to buy 4.0 on cd-rom.
Some days, you can't win for losing.
---