I believe the article explains the basic function of a scheduler before going into the extremely technical description that maybe most people here didn't read. A good summary taken from the article:
Whether your computer is executing 1000 tasks simultaneously or 10, the scheduler will always take the same amount of time to pick the next task.
There are things happening all the time that affect task priority, changing which task should run next, etc. The hard drive is ready with data, same w/ the network card or even your mouse and keyboard input. The point is that the items that you are sorting in effect change their values all the time, so in fact you do not have a maintained sorted list at the beginning of every time slice.
The best in-place sorting algorithm running on unbounded data on an infinite precision computer is O(log(n)). This is basically a component of a binary search type sort, algorithmically no worse than the best generalized sort, O(n log(n))
I'm assuming you've taken an algorithms course, possibly in C.
Let's say you redefine the problem so that you must sort a stream of integers between 1 and 10. You simply make a 10 element array (which, in C, is indexable in O(1) time) of linked lists. When a number comes in, you add it to the appropriate list in O(1) time, either at the head or tail (if you want a FIFO type system).
In an algorithms course, you learn that the best sort is O(n log(n)). If it was a good one, you learn that if you restrict the data being sorted as I have done above, it is possible to attain O(n).
To have an O(1) scheduler, you have to maintain constant execution time in the face of an arbitrary quantity of your processes changing their sorting values between executions. That's not hard, like I did above you can just simplify the problem until you get to O(1), although you'd have to simplify it much further than I did.
What is hard is doing it in such a way that the system doesn't lose data, doesn't force the user or other processes to wait unnecessarily (latency), and utilizes the processors in the system at near 100% efficiency (throughput).
Even to the average geek with a basic grasp of algorithms, it seems hard to beat O(n) when you factor in that external events can change process priorities.
That said, it doesn't seem impossible either. I wonder if it is truly better in every way, even on a 386 (i.e., is it O(1) at the cost of taking more time to run when n is small), or if it is only better on today's machines when the scheduler is using only a small fraction of the CPU power or may even be stored in an on-chip cache marked permanent or something.
I can't tell if you're an idiot or what, but the complete function of the scheduler is to determine what process runs next. The point of the article is that despite the seeming complexity of adding more processes to a running machine, the scheduler always takes the same amount of time to determine which process should run next regardless of how many processes there are to select from.
A sorting algorithm isn't finished until the data is sorted. A scheduler isn't finished until the next process is selected. You wouldn't measure a sorting algorithm by the time it takes to sort all the data anyone in the world would ever use it to sort, would you?
So, she breaks into other people's machines to send spam about money scams, and you think peddling Viagra is beneath her? You think she does business, i.e. trading e-mail addresses with only non-Viagra spammers?
I've got some swamp land in Florida, if you're interested.
The points would have been better if the specific vulnerability discussed could possibly have been used in a large scale virus attack. The difficulty for the author to come up with a useful example says something about OSX, I'd say.
With the latest major OSX "Security Hole," you have to be on the same LAN as the target, and you have to wait until that target reboots. That simply doesn't meet large scale virus propagation requirements.
Your argument, although technically correct, is foolish. People buy SuSE for the CD with the software on it and support of said software, not for its value as a plastic disc.
shit... I can't either, and I run a medium sized ISP... tried connecting directly from my mail server too. For those less technically savvy, the parent listed the MX records for att.com.
Either that, or the recent worm/viruses make the article somewhat questionable.
Re:'windows' mentioned in article.
on
MRAM in 2004?
·
· Score: 1
You're assuming everything closes correctly when do do that. Plus, Office XP (only SP2?) won't let you shut down if you have any of it's programs running.
Governments should tax things ONLY to raise revenue.
Where did this idea that Governments should use taxes as tools of social engineering ever get started?
I hope the original poster's intent was to point out that the government should think about the ripple effects of taxes, and the resulting net gain in revenue over time. It is not true that raising taxes always increases revenue. In fact, given current taxing I dare say raising taxes will always decrease revenue in the long run as people and businesses leave or generally become less successful.
It seems to me that the very idea of paying someone to write free software is the very antithesis of what free software is all about.
Paying someone to write free software is apparently a bad thing in your opinion.
Far better would be something like the Ford Foundation giving grants to folks after they have a track record.
The implication here is that the grants would be for free software: "folks" being free software developers, and their "track records" being their history of producing free software. This requires paying someone to write free software.
Or, are you denying that interpretation? If so, we're forced to conclude this:
Original Slashdot story: Gee, don't you think governments should fund free software development?
Your answer: Paying someone to write free software is bad. A better idea is for an organization like the Ford Foundation to give a grant to someone with a track record in something unrelated to free software develop something also unrelated to free software.
The way I see it, you have a choice between contradicting yourself, having no point, or clarifying what the hell you were trying to say in the first place.
Slashdot Mirror
And they'd end up with a solution that wouldn't be as good as OS X.
I believe the article explains the basic function of a scheduler before going into the extremely technical description that maybe most people here didn't read. A good summary taken from the article:
Whether your computer is executing 1000 tasks simultaneously or 10, the scheduler will always take the same amount of time to pick the next task.
There are things happening all the time that affect task priority, changing which task should run next, etc. The hard drive is ready with data, same w/ the network card or even your mouse and keyboard input. The point is that the items that you are sorting in effect change their values all the time, so in fact you do not have a maintained sorted list at the beginning of every time slice.
The best in-place sorting algorithm running on unbounded data on an infinite precision computer is O(log(n)). This is basically a component of a binary search type sort, algorithmically no worse than the best generalized sort, O(n log(n))
I'm assuming you've taken an algorithms course, possibly in C.
Let's say you redefine the problem so that you must sort a stream of integers between 1 and 10. You simply make a 10 element array (which, in C, is indexable in O(1) time) of linked lists. When a number comes in, you add it to the appropriate list in O(1) time, either at the head or tail (if you want a FIFO type system).
In an algorithms course, you learn that the best sort is O(n log(n)). If it was a good one, you learn that if you restrict the data being sorted as I have done above, it is possible to attain O(n).
To have an O(1) scheduler, you have to maintain constant execution time in the face of an arbitrary quantity of your processes changing their sorting values between executions. That's not hard, like I did above you can just simplify the problem until you get to O(1), although you'd have to simplify it much further than I did.
What is hard is doing it in such a way that the system doesn't lose data, doesn't force the user or other processes to wait unnecessarily (latency), and utilizes the processors in the system at near 100% efficiency (throughput).
Even to the average geek with a basic grasp of algorithms, it seems hard to beat O(n) when you factor in that external events can change process priorities.
That said, it doesn't seem impossible either. I wonder if it is truly better in every way, even on a 386 (i.e., is it O(1) at the cost of taking more time to run when n is small), or if it is only better on today's machines when the scheduler is using only a small fraction of the CPU power or may even be stored in an on-chip cache marked permanent or something.
You're taking advice from someone who you don't value enough to help out occasionally?
If your parents moved, would you help them? Is your family that worthless?
Uneventfully? Sun and SGI didn't have millions if idiots running their systems either. Who would have noticed if there were problems and why?
I can't tell if you're an idiot or what, but the complete function of the scheduler is to determine what process runs next. The point of the article is that despite the seeming complexity of adding more processes to a running machine, the scheduler always takes the same amount of time to determine which process should run next regardless of how many processes there are to select from.
A sorting algorithm isn't finished until the data is sorted. A scheduler isn't finished until the next process is selected. You wouldn't measure a sorting algorithm by the time it takes to sort all the data anyone in the world would ever use it to sort, would you?
How about the 8th commandment: "Thou Shalt Not Steal."
The only spammers able to deliver any content are those that break into other people's machines.
Not to mention the theft of services of normal e-mail that the recipient didn't want...
So, she breaks into other people's machines to send spam about money scams, and you think peddling Viagra is beneath her? You think she does business, i.e. trading e-mail addresses with only non-Viagra spammers?
I've got some swamp land in Florida, if you're interested.
The points would have been better if the specific vulnerability discussed could possibly have been used in a large scale virus attack. The difficulty for the author to come up with a useful example says something about OSX, I'd say.
With the latest major OSX "Security Hole," you have to be on the same LAN as the target, and you have to wait until that target reboots. That simply doesn't meet large scale virus propagation requirements.
True, but you do have to enter your password when using sudo.
I'm ok with saying that Microsoft's software is worse than Apple's software, for example.
Your argument, although technically correct, is foolish. People buy SuSE for the CD with the software on it and support of said software, not for its value as a plastic disc.
Saying, for example, that FreeBSD's memory management is great has no meaning unless compared to Linux's memory management.
"Greatness stands on its own" means that you're describing something that is so good it is obviously better than other things that are comparable.
shit... I can't either, and I run a medium sized ISP... tried connecting directly from my mail server too. For those less technically savvy, the parent listed the MX records for att.com.
Mod this guy up.
You may want to read up on Palladium.
If a new mail protocol was used, that new protocol could prevent the spam without a tax.
Either that, or the recent worm/viruses make the article somewhat questionable.
You're assuming everything closes correctly when do do that. Plus, Office XP (only SP2?) won't let you shut down if you have any of it's programs running.
Governments should tax things ONLY to raise revenue.
Where did this idea that Governments should use taxes as tools of social engineering ever get started?
I hope the original poster's intent was to point out that the government should think about the ripple effects of taxes, and the resulting net gain in revenue over time. It is not true that raising taxes always increases revenue. In fact, given current taxing I dare say raising taxes will always decrease revenue in the long run as people and businesses leave or generally become less successful.
Uhh... so by your logic, the gov. must want to me to stop working and earning money.
Uhh... so by your logic, the gov. always does what it should do?
The TTL values of the blacklists are low to allow for timely removal from the list.
Yes, it is. Everyone makes mistakes, including Microsoft.
So you've signed the NDA and seen "SCO's code?" If not, then I'm unclear as to what claim you are referring.
>> Microsoft isn't allowed to use guns to get its way.
Microsoft has done many things to get its way, not all of which are legal, and many of which are more powerful than guns.
Let me illustrate a little more clearly.
It seems to me that the very idea of paying someone to write free software is the very antithesis of what free software is all about.
Paying someone to write free software is apparently a bad thing in your opinion.
Far better would be something like the Ford Foundation giving grants to folks after they have a track record.
The implication here is that the grants would be for free software: "folks" being free software developers, and their "track records" being their history of producing free software. This requires paying someone to write free software.
Or, are you denying that interpretation? If so, we're forced to conclude this:
Original Slashdot story:
Gee, don't you think governments should fund free software development?
Your answer:
Paying someone to write free software is bad. A better idea is for an organization like the Ford Foundation to give a grant to someone with a track record in something unrelated to free software develop something also unrelated to free software.
The way I see it, you have a choice between contradicting yourself, having no point, or clarifying what the hell you were trying to say in the first place.