We don't live in a democracy, we live in a republic.
One of the primary principles is the Rule of Law. That means that an informed person usually knows what side of the law they are on before trial. That is only true if judges strictly interpret the law rather than do what they personally think is right.
The Rule of Law is part of the structure, but I see it as an important aspect of our society. The alternative is the Rule of Man, and, historically, that leads to worse outcomes. A law is something anyone can see, or can at least ask a lawyer about in advance. Society is more stable and predictable that way.
Or maybe even "Law supports consumers, judge stands up for the law"?
The last thing this country needs is judges who stand up for consumers or any other group. Judges are not elected, and appointed for life. Don't give me this "benevolent dictator" stuff. Judges should strictly interpret the law. If it's unclear, rule on the side of caution and wait 'til Congress or the state clarifies it.
I'm tired of coming across something that won't compile correctly or encountering a bug. Then when you research it you come across 100 posts of people asking the same question and getting the same response of fix it yourself or do a search the answer is already here.
It sounds like you're talking about some small project with a small user base. Exactly the kind of product that probably wouldn't even exist in the commercial world. So, someone makes a solution for a fairly obscure problem, you find it, and its not up to your standards of quality. But at least it's there, and the fact that you're looking at it probably means it doesn't exist anywhere else.
When you compare apples to apples, i.e. a substantial proprietary product with a substantial free product, you get much different results. The free products often have fewer bugs, fewer "gotchas", and very good documentation. When you have a question, usually a quick search of the mailing list archives solves your problem, or failing that a post to the mailing list will often get helpful replies.
My experience with commercial software is much worse. When a feature doesn't work a certain way, sorry, too bad. If there is a bad interaction with other software, or it is somehow incompatible with other software configurations on your computer, again, too bad. The only really "supported" configuration is installing onto a fresh default install of windows XP.
This may be what is currently happening in the US as well.
Maybe so. It depends how you look at it because usually when freedoms in one area go away you get a smaller improvement in freedom somewhere else. Of course that's a political trick. It certainly seems to me like the U.S. is becoming less free.
My point was not that the U.S. is perfect, but that it's the most free country around right now. You could make a good argument for other countries like Canada or Australia or Japan, but the U.S. is certainly ahead of places like India. Bribery in India is the normal course of events.
Sure, the U.S. Congress is corrupt. But compared to India, it's a big improvement. Sometimes it seems like there are lots of things I'd like to change about America, but I also like to keep track of the things that are good about America that we don't want to change. We don't want to become like India.
I would really like to see ODF become more popular.
Could this mean that Safari, which is based on Konqueror, might be able to at least view ODF files?
I think ODF could take off if Macs could effectively use them. I don't see any disadvantage to Apple at all to include KOffice or OpenOffice so that Apple users have something to use even if they don't buy (or more likely, pirate) MS Office.
Also, Google/GMail should support this format! Why not? They allow viewing of MS Word documents as HTML, why not an.odt? That would help a lot.
However, keep in mind that people who live in the U.S. actually still have a lot of freedom in comparison to places like India.
It's worrisome because the U.S. is losing freedom while India is gaining freedom. But India is still far, far away from allowing people the freedom we have here. There is rampant corruption (corruption in the U.S. Congress is nothing compared to India) and overwhelming government control of business. Property rights (and other civil liberties) in India are not protected as much as they are in the U.S.
I see this as a situation where the governments in places like India let the people have a little freedom for a while to get the economy going, and then fall back into government control before the people ever have the level of freedom that exists in the U.S. today.
That's actually true. It's not the dpkg itself, or even apt that makes debian great. It's all the contributors who care about the packages and do a good job making them. I have had great experience with debian for the most part. I have kept my desktop system up to date on "unstable" for about 6-7 years without ever reinstalling. I've switched hard drives, and done all kinds of other changes (including innumerable kernel upgrades), but it's never required much more than "apt-get dist-upgrade". I even messed up my root filesystem once and somehow managed to recover, and getting the packages properly in place was the least of my worries.
Maybe RPMs are better now, but back when I started using them (whatever came before RH 6.0, I think it was 5.2), that would have been about impossible. I don't even notice upgrading dpkg, but upgrading rpm with rpm was hell. I seem to recall some problems upgrading libc also.
Kernel modules have unlimited access to all kernel memory. That means that a bug can corrupt any other part of the kernel, including the buffer cache.
Userspace drivers run in their own address space. Most likely they will share memory with other processes (including other processes that perform traditional kernel tasks), however this access is much more clearly defined and limited.
Crashes with userspace drivers can clearly identify the offending driver, and often can restart the driver or somehow recover. Crashes with kernel modules, well, anything can happen. And you have no idea which driver is the culprit.
In all honesty, it's mostly an accountability thing. Recovery from driver malfunction is tricky. However, if you get a new driver with your new hardware and it keeps crashing on you and informing you of the offending driver, you're going to return the hardware. That means the hardware manufacturers will write better code to avoid that problem. That's the real benefit.
trying to build Windows to be extremely tolerant of crap software and bizarre library calls, and to keep running as long as possible
Actually, at the first hint of a kernelspace problem, windows blue screens. Linux is less picky about those types of problems, and only crashes when there's not really a way to keep running.
I doubt it. All he did was make a meaningless comparison of the number of deaths from car accidents vs the number of deaths from terrorism. Then he attacked the motives of people who want to monitor phone calls.
That's a waste of time. Attacking motives is very hard to argue about, and really adds nothing. For most policies, there are both good and bad people both for and against the policy. It doesn't matter.
You made the point that there is a cost-benefit analysis to be done. That is a very good point, and one that I really agree with. Some measures are worth it, and some are not. We don't just pretend that dangerous people don't exist, but there are limits (based on cost-benefit analysis) as to how far we go to stop them.
I'll make you a deal. It seems like you share some political philosophies with the poster I was responding to, but you also have some grasp of a coherent argument. How about if, every time he feels like posting, he tells you, and you post instead. I think the world would be a better place.
I would like to think that most of the containers comes from countries we trust, like UK, Australia, etc. Obviously we don't get many containers from Mexico or Canada.
There's always a weakness. But the fewer options we allow the terrorists, the longer it will take for them to be successful.
It's a cost/benefit analysis. Significantly improving the security of the border is cheap. We make a wall, hire a few thousand people, maybe use the national guard. We put up cameras and have some people ready to chase down people who run over the wall. I don't know exactly how much, but compared to our GDP it's nothing.
The cost of monitoring all shipping containers is prohibitive. Maybe we can just monitor the ones from questionable countries, I don't really know how many containers we accept from those countries. After all, there are already many rules and standards for accepting or rejecting ships from a harbor before they are let in.
Some ideas are worth the cost and some are not. I think it's a bad idea to just throw up our hands and give up. "Oh well, anyone with $20M and a bad mood can blow up Los Angeles". I don't think so. We should try to make it expensive for the terrorists if we can, which will at least reduce the problem.
And there's also a possibility that Scientology is right.
Why does everyone consider the nuclear terrorist such an implausibility? To me it seems likely that it will happen to some country, most likely a country with lax border enforcement. I'm not suggesting that Bush's anti-terrorist stuff is actually preventing this outcome. But I think that a controlled southern border would be a good US policy.
And certainly there are more than a few politicians using the terror threat to their advantage. But that doesn't make it less of a threat. We should still implement reasonable policies (like controlling the southern border) designed to prevent serious terrorism.
Way to attack a straw man. Where did he say that terrorism should be legal?
The statement he made had no logic or analysis to it. How did he make the logical jump from "more people die in car accidents than terrorist attacks" to "ban all cars"? He made no attempt to connect the two or make a meaningful comparison.
people should worry more about traffic safety than about terrorism
That doesn't suggest any policy at all, so that's hard to argue with. It also leaves a lot of questions. How much do people currently worry about terrorism versus traffic safety? Is it out of balance? How do those people act on that worry? Are the actions reasonable?
That statement really adds nothing to the argument at all. It's completely vague and empty.
Don't dismiss this issue, that would be oversimplifying.
You didn't provide a policy, so I have nothing to argue against. I can't argue against "action", because it's not defined.
The "insightful" post I responded to did nothing other than attack the motives of the people behind a policy, which is unproductive. The only idea he put forth was devoid of logic (banning cars because more people die in car accidents than from terrorism).
Even though you have a higher chance of dying from car accidents (why don't we ban all cars?), people are scared shitless of terrorists.
How was this insightful at all? It's meaningless and doesn't stand up to the simplest analysis. Fewer people die from murder than car accidents (in the U.S.), are you saying that murder should be legal?
Not only that, there are very legitimate concerns about terrorism. There is a possibility that terrorists could get a nuclear warhead and run across the border with it and blow up an entire city. I'm not saying that Bush is helping to prevent that at all, I'm just saying there is a concern.
Attack policies and ideas all you want. Dismissing the issues is oversimplifying. Attacking the motives of people is unproductive and condescending. Generally there are many groups for and against any given policy, and they all have different motives.
Good idea, and that's close to what actually happens. Microkernels use extensive sharing among the pieces which are partof a traditional kernel (i.e. drivers, fs, etc.). As I understand it, the problem that Linus is referring to is that there are so many interactions that it requires so much sharing that it actually becomes more complex.
I think a good analogy is if you are running a business. You can contract out to accomplish any business task you need to. Sometimes it's efficient: you can contract with a machine shop to machine a piece that's important for your product. But sometimes it's inefficient, also. For example, it could be a bad idea to outsource your secretary. You interact with your secretary so much that if she's at another office, she's not effective at all.
So, there's a balance between sharing too much and not sharing enough. Linus thinks that balance is close to what we have right now in the most successful general-purpose operating systems (Linux, Windows, FreeBSD,...). Microkernel developers are more academic, which is no surprise. Academics generally take things to their logical conclusion, and since there is no fundamental reason why a device driver needs to be in the kernel, the logical thing is to remove it. In practice, there are so many complex interactions that it becomes more complicated when the processes are more independent.
Keep in mind that monolithic kernels also use sharing. Generally, when you fork() a process (turn one process into 2 processes), those processes logically each have their own address space. But the kernel does not copy all of the memory for that process, because that's wasteful. Instead, it usually copies only a small amount, and shares the rest. When one process writes, the data is no longer the same and can no longer be shared, so a copy is made at that point. That's called copy-on-write. There's also more direct sharing through explicitly shared memory, where you have to be careful to synchronize the access.
There is a lot of sharing going on in either a micro or monolithic kernel.
When it comes to OS design, there are many ways where you can go very, very wrong.
If we did not have people spending years researching these microkernels, the performance would be well outside of reason, even on modern hardware for basic desktop use.
How is kernel compartmentalization going to protect against users installing spyware and doing things they're already authorized to do?
Through isolation, it allows you to recover a lot more easily. Sure, you can install bad stuff, but if everything is more isolated you can mitigate the damage and recover 100%.
I think that the main thing you're looking for is paravirtualization, either through something like Xen or a microkernel. That's a great form of isolation.
I'm just not going to schedule that rmmod ans insmod.
First off, the disk scheduler is not the same as the process scheduler. However, I suppose it does need to read those programs from disk if they are not already in memory.
You make an interesting point, but I think you miss the point of a microkernel. User processes are not priviledged by the hardware (i.e., they do not run in supervisor mode). But many user level processes are still logically priviledged. Consider that there are many programs that must be run as root. They are still user processes, but they are being run as the user "root", which can do more than the user "joe". So, the process scheduler or the I/O scheduler may be user processes, but you still must place some trust in them to do their job (just as you trust the root-owned user mode processes to do their job).
The benefit is that things like drivers aren't as likely to interfere with other important kernel processes.
But I'd even more prefer to see the driver written correctly to start with!
Microkernels actually may help with that as well. If it is very obvious to the OS -- and to the user -- which drivers are crashing, that will provide incentive for the hardware vendors to write drivers correctly. Right now there is no accountability, so as long as the whole system works most of the time, users will buy it. But with microkernels, if new hardware comes out and you have review sites saying "That hardware driver is crashing left and right", users won't buy it. Nobody can point fingers anymore.
In particular, nobody will point fingers at MS Windows when the real problem is crappy 3rd party drivers.
My point was that when you are measuring the performance of a high level language like perl, python, or ruby, the normal use case does not really test the performance of the language. Mostly those languages quickly jump into optimized C routines, which are hard to beat (admittedly, ruby and python aren't as heavily optimized as perl).
However, the normal use case for java can test the speed of the language itself. Generally, a higher proportion of the work is done in the JVM versus the perl interpreter, so the speed of the JVM matters much more. In perl what matters is the speed of the C routines that are called.
In ruby, if you have a sophisticated, performance critical algorithm, the correct thing to do is write it in C as a module. Then just call it with a couple lines of ruby code. You can do that in Java, also, but it's not quite the normal process.
Slashdot Mirror
We don't live in a democracy, we live in a republic.
One of the primary principles is the Rule of Law. That means that an informed person usually knows what side of the law they are on before trial. That is only true if judges strictly interpret the law rather than do what they personally think is right.
The Rule of Law is part of the structure, but I see it as an important aspect of our society. The alternative is the Rule of Man, and, historically, that leads to worse outcomes. A law is something anyone can see, or can at least ask a lawyer about in advance. Society is more stable and predictable that way.
How about "Judge stands up for the law"?
Or maybe even "Law supports consumers, judge stands up for the law"?
The last thing this country needs is judges who stand up for consumers or any other group. Judges are not elected, and appointed for life. Don't give me this "benevolent dictator" stuff. Judges should strictly interpret the law. If it's unclear, rule on the side of caution and wait 'til Congress or the state clarifies it.
I'm tired of coming across something that won't compile correctly or encountering a bug. Then when you research it you come across 100 posts of people asking the same question and getting the same response of fix it yourself or do a search the answer is already here.
It sounds like you're talking about some small project with a small user base. Exactly the kind of product that probably wouldn't even exist in the commercial world. So, someone makes a solution for a fairly obscure problem, you find it, and its not up to your standards of quality. But at least it's there, and the fact that you're looking at it probably means it doesn't exist anywhere else.
When you compare apples to apples, i.e. a substantial proprietary product with a substantial free product, you get much different results. The free products often have fewer bugs, fewer "gotchas", and very good documentation. When you have a question, usually a quick search of the mailing list archives solves your problem, or failing that a post to the mailing list will often get helpful replies.
My experience with commercial software is much worse. When a feature doesn't work a certain way, sorry, too bad. If there is a bad interaction with other software, or it is somehow incompatible with other software configurations on your computer, again, too bad. The only really "supported" configuration is installing onto a fresh default install of windows XP.
This may be what is currently happening in the US as well.
Maybe so. It depends how you look at it because usually when freedoms in one area go away you get a smaller improvement in freedom somewhere else. Of course that's a political trick. It certainly seems to me like the U.S. is becoming less free.
My point was not that the U.S. is perfect, but that it's the most free country around right now. You could make a good argument for other countries like Canada or Australia or Japan, but the U.S. is certainly ahead of places like India. Bribery in India is the normal course of events.
Sure, the U.S. Congress is corrupt. But compared to India, it's a big improvement. Sometimes it seems like there are lots of things I'd like to change about America, but I also like to keep track of the things that are good about America that we don't want to change. We don't want to become like India.
I would really like to see ODF become more popular.
.odt? That would help a lot.
Could this mean that Safari, which is based on Konqueror, might be able to at least view ODF files?
I think ODF could take off if Macs could effectively use them. I don't see any disadvantage to Apple at all to include KOffice or OpenOffice so that Apple users have something to use even if they don't buy (or more likely, pirate) MS Office.
Also, Google/GMail should support this format! Why not? They allow viewing of MS Word documents as HTML, why not an
From http://m-w.com/dictionary/enhertiently
The word you've entered isn't in the dictionary. Click on a spelling suggestion below or try again using the search box to the right.
Suggestions for enhertiently:
1. interdentally
2. intermittently
3. interdental
4. intolerantly
5. inheritances
6. intertidally
7. antirealist
8. antihunter
9. antinarrative
10. antirealists
Huh? At first I thought it was a real word because I couldn't think of a word that made sense in that position.
Yes, you're right.
However, keep in mind that people who live in the U.S. actually still have a lot of freedom in comparison to places like India.
It's worrisome because the U.S. is losing freedom while India is gaining freedom. But India is still far, far away from allowing people the freedom we have here. There is rampant corruption (corruption in the U.S. Congress is nothing compared to India) and overwhelming government control of business. Property rights (and other civil liberties) in India are not protected as much as they are in the U.S.
I see this as a situation where the governments in places like India let the people have a little freedom for a while to get the economy going, and then fall back into government control before the people ever have the level of freedom that exists in the U.S. today.
That's actually true. It's not the dpkg itself, or even apt that makes debian great. It's all the contributors who care about the packages and do a good job making them. I have had great experience with debian for the most part. I have kept my desktop system up to date on "unstable" for about 6-7 years without ever reinstalling. I've switched hard drives, and done all kinds of other changes (including innumerable kernel upgrades), but it's never required much more than "apt-get dist-upgrade". I even messed up my root filesystem once and somehow managed to recover, and getting the packages properly in place was the least of my worries.
Maybe RPMs are better now, but back when I started using them (whatever came before RH 6.0, I think it was 5.2), that would have been about impossible. I don't even notice upgrading dpkg, but upgrading rpm with rpm was hell. I seem to recall some problems upgrading libc also.
The market value is, almost by definition, the "cost of infringement".
The market value of the patent is the value of the lawsuit that you can file.
the company that owns this so-called patent only has it for the purposes of suing other people.
Why else would you own a patent?
Kernel modules have unlimited access to all kernel memory. That means that a bug can corrupt any other part of the kernel, including the buffer cache.
Userspace drivers run in their own address space. Most likely they will share memory with other processes (including other processes that perform traditional kernel tasks), however this access is much more clearly defined and limited.
Crashes with userspace drivers can clearly identify the offending driver, and often can restart the driver or somehow recover. Crashes with kernel modules, well, anything can happen. And you have no idea which driver is the culprit.
In all honesty, it's mostly an accountability thing. Recovery from driver malfunction is tricky. However, if you get a new driver with your new hardware and it keeps crashing on you and informing you of the offending driver, you're going to return the hardware. That means the hardware manufacturers will write better code to avoid that problem. That's the real benefit.
trying to build Windows to be extremely tolerant of crap software and bizarre library calls, and to keep running as long as possible
Actually, at the first hint of a kernelspace problem, windows blue screens. Linux is less picky about those types of problems, and only crashes when there's not really a way to keep running.
What he was trying to get at was this
I doubt it. All he did was make a meaningless comparison of the number of deaths from car accidents vs the number of deaths from terrorism. Then he attacked the motives of people who want to monitor phone calls.
That's a waste of time. Attacking motives is very hard to argue about, and really adds nothing. For most policies, there are both good and bad people both for and against the policy. It doesn't matter.
You made the point that there is a cost-benefit analysis to be done. That is a very good point, and one that I really agree with. Some measures are worth it, and some are not. We don't just pretend that dangerous people don't exist, but there are limits (based on cost-benefit analysis) as to how far we go to stop them.
I'll make you a deal. It seems like you share some political philosophies with the poster I was responding to, but you also have some grasp of a coherent argument. How about if, every time he feels like posting, he tells you, and you post instead. I think the world would be a better place.
You make a good point.
I would like to think that most of the containers comes from countries we trust, like UK, Australia, etc. Obviously we don't get many containers from Mexico or Canada.
There's always a weakness. But the fewer options we allow the terrorists, the longer it will take for them to be successful.
It's a cost/benefit analysis. Significantly improving the security of the border is cheap. We make a wall, hire a few thousand people, maybe use the national guard. We put up cameras and have some people ready to chase down people who run over the wall. I don't know exactly how much, but compared to our GDP it's nothing.
The cost of monitoring all shipping containers is prohibitive. Maybe we can just monitor the ones from questionable countries, I don't really know how many containers we accept from those countries. After all, there are already many rules and standards for accepting or rejecting ships from a harbor before they are let in.
Some ideas are worth the cost and some are not. I think it's a bad idea to just throw up our hands and give up. "Oh well, anyone with $20M and a bad mood can blow up Los Angeles". I don't think so. We should try to make it expensive for the terrorists if we can, which will at least reduce the problem.
And there's also a possibility that Scientology is right.
Why does everyone consider the nuclear terrorist such an implausibility? To me it seems likely that it will happen to some country, most likely a country with lax border enforcement. I'm not suggesting that Bush's anti-terrorist stuff is actually preventing this outcome. But I think that a controlled southern border would be a good US policy.
And certainly there are more than a few politicians using the terror threat to their advantage. But that doesn't make it less of a threat. We should still implement reasonable policies (like controlling the southern border) designed to prevent serious terrorism.
Way to attack a straw man. Where did he say that terrorism should be legal?
The statement he made had no logic or analysis to it. How did he make the logical jump from "more people die in car accidents than terrorist attacks" to "ban all cars"? He made no attempt to connect the two or make a meaningful comparison.
people should worry more about traffic safety than about terrorism
That doesn't suggest any policy at all, so that's hard to argue with. It also leaves a lot of questions. How much do people currently worry about terrorism versus traffic safety? Is it out of balance? How do those people act on that worry? Are the actions reasonable?
That statement really adds nothing to the argument at all. It's completely vague and empty.
Don't dismiss this issue, that would be oversimplifying.
You didn't provide a policy, so I have nothing to argue against. I can't argue against "action", because it's not defined.
The "insightful" post I responded to did nothing other than attack the motives of the people behind a policy, which is unproductive. The only idea he put forth was devoid of logic (banning cars because more people die in car accidents than from terrorism).
Even though you have a higher chance of dying from car accidents (why don't we ban all cars?), people are scared shitless of terrorists.
How was this insightful at all? It's meaningless and doesn't stand up to the simplest analysis. Fewer people die from murder than car accidents (in the U.S.), are you saying that murder should be legal?
Not only that, there are very legitimate concerns about terrorism. There is a possibility that terrorists could get a nuclear warhead and run across the border with it and blow up an entire city. I'm not saying that Bush is helping to prevent that at all, I'm just saying there is a concern.
Attack policies and ideas all you want. Dismissing the issues is oversimplifying. Attacking the motives of people is unproductive and condescending. Generally there are many groups for and against any given policy, and they all have different motives.
that compasses and other navigational devices currently compensate for the difference
How would that work? How would the compass know which way to compensate and by how much?
jobs for everyone and no homeless...damned commie scum. Real patriots live on the streets.
Too bad it doesn't provide food for everyone. I suppose you think communism is great except for the tens of millions of people who starved to death.
Good idea, and that's close to what actually happens. Microkernels use extensive sharing among the pieces which are partof a traditional kernel (i.e. drivers, fs, etc.). As I understand it, the problem that Linus is referring to is that there are so many interactions that it requires so much sharing that it actually becomes more complex.
...). Microkernel developers are more academic, which is no surprise. Academics generally take things to their logical conclusion, and since there is no fundamental reason why a device driver needs to be in the kernel, the logical thing is to remove it. In practice, there are so many complex interactions that it becomes more complicated when the processes are more independent.
I think a good analogy is if you are running a business. You can contract out to accomplish any business task you need to. Sometimes it's efficient: you can contract with a machine shop to machine a piece that's important for your product. But sometimes it's inefficient, also. For example, it could be a bad idea to outsource your secretary. You interact with your secretary so much that if she's at another office, she's not effective at all.
So, there's a balance between sharing too much and not sharing enough. Linus thinks that balance is close to what we have right now in the most successful general-purpose operating systems (Linux, Windows, FreeBSD,
Keep in mind that monolithic kernels also use sharing. Generally, when you fork() a process (turn one process into 2 processes), those processes logically each have their own address space. But the kernel does not copy all of the memory for that process, because that's wasteful. Instead, it usually copies only a small amount, and shares the rest. When one process writes, the data is no longer the same and can no longer be shared, so a copy is made at that point. That's called copy-on-write. There's also more direct sharing through explicitly shared memory, where you have to be careful to synchronize the access.
There is a lot of sharing going on in either a micro or monolithic kernel.
When it comes to OS design, there are many ways where you can go very, very wrong.
If we did not have people spending years researching these microkernels, the performance would be well outside of reason, even on modern hardware for basic desktop use.
How is kernel compartmentalization going to protect against users installing spyware and doing things they're already authorized to do?
Through isolation, it allows you to recover a lot more easily. Sure, you can install bad stuff, but if everything is more isolated you can mitigate the damage and recover 100%.
I think that the main thing you're looking for is paravirtualization, either through something like Xen or a microkernel. That's a great form of isolation.
I'm just not going to schedule that rmmod ans insmod.
First off, the disk scheduler is not the same as the process scheduler. However, I suppose it does need to read those programs from disk if they are not already in memory.
You make an interesting point, but I think you miss the point of a microkernel. User processes are not priviledged by the hardware (i.e., they do not run in supervisor mode). But many user level processes are still logically priviledged. Consider that there are many programs that must be run as root. They are still user processes, but they are being run as the user "root", which can do more than the user "joe". So, the process scheduler or the I/O scheduler may be user processes, but you still must place some trust in them to do their job (just as you trust the root-owned user mode processes to do their job).
The benefit is that things like drivers aren't as likely to interfere with other important kernel processes.
But I'd even more prefer to see the driver written correctly to start with!
Microkernels actually may help with that as well. If it is very obvious to the OS -- and to the user -- which drivers are crashing, that will provide incentive for the hardware vendors to write drivers correctly. Right now there is no accountability, so as long as the whole system works most of the time, users will buy it. But with microkernels, if new hardware comes out and you have review sites saying "That hardware driver is crashing left and right", users won't buy it. Nobody can point fingers anymore.
In particular, nobody will point fingers at MS Windows when the real problem is crappy 3rd party drivers.
My point was that when you are measuring the performance of a high level language like perl, python, or ruby, the normal use case does not really test the performance of the language. Mostly those languages quickly jump into optimized C routines, which are hard to beat (admittedly, ruby and python aren't as heavily optimized as perl).
However, the normal use case for java can test the speed of the language itself. Generally, a higher proportion of the work is done in the JVM versus the perl interpreter, so the speed of the JVM matters much more. In perl what matters is the speed of the C routines that are called.
In ruby, if you have a sophisticated, performance critical algorithm, the correct thing to do is write it in C as a module. Then just call it with a couple lines of ruby code. You can do that in Java, also, but it's not quite the normal process.