Exactly. When I was in school a few years ago I had to deal with a residential internet provider that thought it would be amusing to suddenly block outgoing TCP port 22 for no particularly clear reason, and I ended up having to trek down to the labs to do all my assignments rather than just SSH from home through most of my senior year.
Now I live in the middle of Seattle, I could get DSL or whatever easily enough, and I chose a full T1 from Speakeasy for $460/mo instead, and I get no blocked ports, no bandwidth caps, and my own/27.
CPU cache happens on time scales around six orders of magnitude faster than anything the disk does; of course it has dedicated hardware. For the same reason it's pretty inflexible about replacement policy.
The whole point of either of my suggestions is to keep the same POSIX-compatible atime semantics without the overhead of all those disk writes by distinguishing between an inode being written because of an atime update or for some other reason. In the latter case it's probably really important to get it out ahead of other writes so the on-disk filesystem is consistent, but in the former case (which is probably much more common) it isn't a problem if some get lost when a filesystem doesn't get cleanly unmounted, so it can be written out at lower priority than just about everything else.
The problem of determining priorities for disk writes is inherently an issue for the operating system; it depends on the specific content of the write (inode writes should probably go ahead of data writes, but atime updates should be near the lowesst priority), and the disk controller can't see that and shouldn't try to implement such a system specific feature in hardware anyway. You'd want a completely different scheme if you were going to provide user processes guaranteed bandwidth to the disk, like IRIX does, for example.
Anyway, the semantic question of determining the relative priorities of disk operations is a separate question from the implementation of that disk scheduling once you know what the priorities are, which probably should be done in hardware (and for all I know probably is).
It probably isn't really reliable to assume the inode is close to the file, but it probably is in cache >95% of the time because it had to be read during the open(). Anyway, even if the inode is in cache when you read the file you still have to seek to it to write out the atime update the same as if it weren't.
atime updates happen when you read the file's content, but not on stat(), so either the head has to move anyway to read the data, and then back to the inode to update it, or your reads have to block until after it gets updated. I really hope they went with the former option.:)
Anyway, in my proposal you'd be able to merge atime updates for lots of different files together, but you could also keep it in the inode and just have a separate state to say that just the atime is dirty and it's less important to write it out. Now that I'm reading the thread linked to in the article, Arjan van de Ven proposes exactly that and Theodore Tso points out that it would still be POSIX-compliant to lose atime updates if a filesystem gets unmounted uncleanly.
Why should atime updates have to be written out to disk immediately? It probably isn't the end of the world if a few get lost if a filesystem doesn't get unmounted cleanly, and it probably updates a *lot* more frequently than anything else in the inode, so why not just have the filesystem keep the atimes separately from the rest of the metadata somewhere?
It would only take a little bit of space to hold all the atimes on most filesystems (4 bytes per atime times say 250,000 files plus 5% for indexing overhead (you'd have to map inode numbers to indices into the array of atimes) is just a little over a megabyte), so if you just set that aside somewhere, cached a copy in memory somewhere, and wrote out updates whenever there was some free bandwidth to the disk, you'd be able to merge updates for many different files together instead of having to write out an entire block for every atime update, that you have to write out immediately because it counts as an inode update and it'd be bad to let those fall out of sync.
Wow, my girlfriend and I aren't the only ones.:)
Don't do it. If you did, it would make the front page instantly and bring the world to a grinding halt as straight male geeks everywhere wasted days on end looking at the pictures.:)
It's a little harder than just having some surgery. How does spending an afternoon hiding in your office because you forgot to take your hormones that morning and now you can't stop crying sound?
Grrr. Bloody Slashdot made my comment illegible. Re-posted with non-broken formatting:
I know I'm going to regret debating gender with an anonymous coward on Slashdot, but referring to trans women as "transgenderred [sic] males" is just a little too offensive to let pass. There's good evidence to suggest that transsexualism is caused by congenital differences in certain parts of the brain involved with gender identity and sexuality [1] [2]. Essentially, we really are born with female brains. There's also evidence to suggest that a number of anatomical features of our brains shift to opposite-sex proportions during the first few months on hormones [3].
As for your ideas on explaining the sex ratio in technology jobs by men having a wider IQ distribution than women, I just don't think the numbers hold up. While there seems to be some support for men's IQ distribution having a larger standard deviation than women's, it just isn't enough to support your claim that "the chance of a person having an IQ of 125 is eight times more likely for males than females." The idea certainly isn't an implaausible one at extremely high IQ scores; all it would take would be an X-linked trait that sufficiently influences IQ for people with XY chromosomes (mostly masculine-gendered) to have a wider distribution than people with XX chromosomes (mostly feminine-gendered), and XY people *do* have a wider range of variation on a lot of different traits, probably because of a mechanism like this.
The problem is that the effect would be just too small to make that big a difference; some quick googling turned up at most support for a one or two point difference in standard deviation, and that certainly wouldn't lead to a factor of eight difference at scores as low as 125. A one point difference in male versus female standard deviation of IQ in a population with an overall standard deviation of 16 points leads to men being twice as likely as women to have an IQ over 150, and only one person in a thousand of either sex scores that high. No matter what inflated opinions Slashdotters might have of themselves, I can say very assuredly that the great majority of IT workers do not have 150+ IQs. This sort of theory is probably only useful for explaining sex ratios in elite groups like Nobel prize winners, and I'd be rather skeptical even then of the assumption that the distribution of extremely high IQ scores can be obtained just by extrapolating the distribution for the middle-scoring bulk of the population.
I know I'm going to regret debating gender with an anonymous coward on Slashdot, but referring to trans women as "transgenderred [sic] males" is just a little too offensive to let pass. There's good evidence to suggest that transsexualism is caused by congenital differences in certain parts of the brain involved with gender identity and sexuality [1] [2]. Essentially, we really are born with female brains. There's also evidence to suggest that a number of anatomical features of our brains shift to opposite-sex proportions during the first few months on hormones [3].
As for your ideas on explaining the sex ratio in technology jobs by men having a wider IQ distribution than women, I just don't think the numbers hold up. While there seems to be some support for men's IQ distribution having a larger standard deviation than women's, it just isn't enough to support your claim that "the chance of a person having an IQ of 125 is eight times more likely for males than females." The idea certainly isn't an implaausible one at extremely high IQ scores; all it would take would be an X-linked trait that sufficiently influences IQ for people with XY chromosomes (mostly masculine-gendered) to have a wider distribution than people with XX chromosomes (mostly feminine-gendered), and XY people *do* have a wider range of variation on a lot of different traits, probably because of a mechanism like this.
The problem is that the effect would be just too small to make that big a difference; some quick googling turned up at most support for a one or two point difference in standard deviation, and that certainly wouldn't lead to a factor of eight difference at scores as low as 125. A one point difference in male versus female standard deviation of IQ in a population with an overall standard deviation of 16 points leads to men being twice as likely as women to have an IQ over 150, and only one person in a thousand of either sex scores that high. No matter what inflated opinions Slashdotters might have of themselves, I can say very assuredly that the great majority of IT workers do not have 150+ IQs. This sort of theory is probably only useful for explaining sex ratios in elite groups like Nobel prize winners, and I'd be rather skeptical even then of the assumption that the distribution of extremely high IQ scores can be obtained just by extrapolating the distribution for the middle-scoring bulk of the population.
[1] A Sex Difference in the Human Brain and its Relation to Transsexuality
[2] Male-to-Female Transsexuals Have Female Neuron Numbers in a Limbic Nucleus.
[3] Changing your sex changes your brain: influences of testosterone and estrogen on adult human brain structure
As a polyamorous lesbian transsexual, I'd be the perfect candidate.:)
Except that I can say from personal experience that I still have plenty of sexual desire left after chemically supressing pretty much all of my testosterone production.
On the other hand, I'm physiologically incapable of either becoming pregnant or getting anyone else pregnant.
Light pressure is just incident power divided by the speed of light. With the half rotating toward the sun shielded as you described, Venus displays an area of 5.75 * 10^13 m^2, and is located about 1.08 * 10^11 m from the sun. The sun has a luminosity of about 3.8 * 10^26 W, so at Venus' orbit the irradiance from the sun is about 2592 W/m^2, so the unshaed half of Venus receives an incident power of 1.49 * 10^17 W, translating to a force of 4.97 * 10^8 N. If we make the approximation that this is all concentrated on a lever arm of half the radius of Venus, this gives a torque of about 1.5 * 10^15 N*m. Venus has a moment of inertia of 7.13 * 10^37 kg*m^2, so this torque would cause an angular acceleration of 2.11 * 10^-23 radians/s^2. The current angular velocity of Venus is 2.99 * 10^-7 radians/s, and the desired angular velocity is 7.27 * 10^-5 radians/s, so it would take 3.43 * 10^18 seconds so spin up Venus this way, or about 109 billion years. That's neglecting a factor of two depending on reflection (this is correct if Venus absorbs all the incident radiation), and the fact that the sun's power output will not remain constant over that time frame, nor, very probably, will the existence of Venus.
Terraforming Mars is probably a lot more practical than Venus. The big problem with Mars is the low gravity, so light gasses in the atmosphere leak away to space. I know Mars can't retain hydrogen, for example, but I'm not sure whether it could hold an oxygen atmosphere. It seems to manage to hold carbon dioxide just fine. Anyway, if you gave it an Earth-like atmosphere, it would probably be able to hold onto it, if not indefinitely, for millions of years, not just a few thousand, so even if it won't be permanent with continuing maintenance it still might be worth it.
Venus only rotates once every 243 days. You'd have to speed it up quite a bit or you'd just have huge temperature differences between the day and night sides and giant wind storms in between. Approximating Venus as a sphere of uniform density, it has a moment of inertia of and a rotational kinetic energy of 3.19 * 10^24 J. To accelerate it to the same rotation rate as Earth, you would have to raise that to 1.88 * 10^29 J. On the other hand a typical comet has a mass of up to 10^15 kg. They do move very fast, though, due to their highly eccentric orbits. That pages quotes Halley's comet as having an orbital velocity near its perihelion of 68 km/s. Using that velocity and the 10^15 kg mass figure, one comet would have a kinetic energy of about 2 * 10^24 J. In other words, even if you could couple the comits' energy into Venus' rotation with 100% efficiency (yeah, right), you would still need to hit it with about 150,000 comets to get it up to Earth-like rotation rates. This does not strike me as a particularly practical plan.
You want to cure "compliance, order, and irrational submission to authority" by having an authority tell people what to believe? I think you need to think that one through a little more.
He said the kid was using a machine at a university in Sweden, not that he attended said university. Swedes can have insecure machines too, you know.
Anyway, I was in university at 16, and was done with high school at 15, so it isn't that much of a stretch for this kid to be in university at 15. I could have been if I hadn't spent that year sitting around goofing off and reading Misner, Thorne and Wheeler instead.
Slashdot Mirror
Exactly. When I was in school a few years ago I had to deal with a residential internet provider that thought it would be amusing to suddenly block outgoing TCP port 22 for no particularly clear reason, and I ended up having to trek down to the labs to do all my assignments rather than just SSH from home through most of my senior year. Now I live in the middle of Seattle, I could get DSL or whatever easily enough, and I chose a full T1 from Speakeasy for $460/mo instead, and I get no blocked ports, no bandwidth caps, and my own /27.
CPU cache happens on time scales around six orders of magnitude faster than anything the disk does; of course it has dedicated hardware. For the same reason it's pretty inflexible about replacement policy.
The whole point of either of my suggestions is to keep the same POSIX-compatible atime semantics without the overhead of all those disk writes by distinguishing between an inode being written because of an atime update or for some other reason. In the latter case it's probably really important to get it out ahead of other writes so the on-disk filesystem is consistent, but in the former case (which is probably much more common) it isn't a problem if some get lost when a filesystem doesn't get cleanly unmounted, so it can be written out at lower priority than just about everything else.
The problem of determining priorities for disk writes is inherently an issue for the operating system; it depends on the specific content of the write (inode writes should probably go ahead of data writes, but atime updates should be near the lowesst priority), and the disk controller can't see that and shouldn't try to implement such a system specific feature in hardware anyway. You'd want a completely different scheme if you were going to provide user processes guaranteed bandwidth to the disk, like IRIX does, for example.
Anyway, the semantic question of determining the relative priorities of disk operations is a separate question from the implementation of that disk scheduling once you know what the priorities are, which probably should be done in hardware (and for all I know probably is).
Um, okay. How is the HD supposed to know about OS-level data structures? An inode looks just like any other block at that level.
I meant without changing the semantics. atime is actually used for a few things.
It probably isn't really reliable to assume the inode is close to the file, but it probably is in cache >95% of the time because it had to be read during the open(). Anyway, even if the inode is in cache when you read the file you still have to seek to it to write out the atime update the same as if it weren't.
atime updates happen when you read the file's content, but not on stat(), so either the head has to move anyway to read the data, and then back to the inode to update it, or your reads have to block until after it gets updated. I really hope they went with the former option. :)
Anyway, in my proposal you'd be able to merge atime updates for lots of different files together, but you could also keep it in the inode and just have a separate state to say that just the atime is dirty and it's less important to write it out. Now that I'm reading the thread linked to in the article, Arjan van de Ven proposes exactly that and Theodore Tso points out that it would still be POSIX-compliant to lose atime updates if a filesystem gets unmounted uncleanly.
Why should atime updates have to be written out to disk immediately? It probably isn't the end of the world if a few get lost if a filesystem doesn't get unmounted cleanly, and it probably updates a *lot* more frequently than anything else in the inode, so why not just have the filesystem keep the atimes separately from the rest of the metadata somewhere? It would only take a little bit of space to hold all the atimes on most filesystems (4 bytes per atime times say 250,000 files plus 5% for indexing overhead (you'd have to map inode numbers to indices into the array of atimes) is just a little over a megabyte), so if you just set that aside somewhere, cached a copy in memory somewhere, and wrote out updates whenever there was some free bandwidth to the disk, you'd be able to merge updates for many different files together instead of having to write out an entire block for every atime update, that you have to write out immediately because it counts as an inode update and it'd be bad to let those fall out of sync.
You have a point there. :)
Wow, my girlfriend and I aren't the only ones. :)
Don't do it. If you did, it would make the front page instantly and bring the world to a grinding halt as straight male geeks everywhere wasted days on end looking at the pictures. :)
It's a little harder than just having some surgery. How does spending an afternoon hiding in your office because you forgot to take your hormones that morning and now you can't stop crying sound?
Grrr. Bloody Slashdot made my comment illegible. Re-posted with non-broken formatting:
I know I'm going to regret debating gender with an anonymous coward on Slashdot, but referring to trans women as "transgenderred [sic] males" is just a little too offensive to let pass. There's good evidence to suggest that transsexualism is caused by congenital differences in certain parts of the brain involved with gender identity and sexuality [1] [2]. Essentially, we really are born with female brains. There's also evidence to suggest that a number of anatomical features of our brains shift to opposite-sex proportions during the first few months on hormones [3].
As for your ideas on explaining the sex ratio in technology jobs by men having a wider IQ distribution than women, I just don't think the numbers hold up. While there seems to be some support for men's IQ distribution having a larger standard deviation than women's, it just isn't enough to support your claim that "the chance of a person having an IQ of 125 is eight times more likely for males than females." The idea certainly isn't an implaausible one at extremely high IQ scores; all it would take would be an X-linked trait that sufficiently influences IQ for people with XY chromosomes (mostly masculine-gendered) to have a wider distribution than people with XX chromosomes (mostly feminine-gendered), and XY people *do* have a wider range of variation on a lot of different traits, probably because of a mechanism like this.
The problem is that the effect would be just too small to make that big a difference; some quick googling turned up at most support for a one or two point difference in standard deviation, and that certainly wouldn't lead to a factor of eight difference at scores as low as 125. A one point difference in male versus female standard deviation of IQ in a population with an overall standard deviation of 16 points leads to men being twice as likely as women to have an IQ over 150, and only one person in a thousand of either sex scores that high. No matter what inflated opinions Slashdotters might have of themselves, I can say very assuredly that the great majority of IT workers do not have 150+ IQs. This sort of theory is probably only useful for explaining sex ratios in elite groups like Nobel prize winners, and I'd be rather skeptical even then of the assumption that the distribution of extremely high IQ scores can be obtained just by extrapolating the distribution for the middle-scoring bulk of the population.
[1] A Sex Difference in the Human Brain and its Relation to Transsexuality
[2] Male-to-Female Transsexuals Have Female Neuron Numbers in a Limbic Nucleus
[3] Changing your sex changes your brain: influences of testosterone and estrogen on adult human brain structure
I know I'm going to regret debating gender with an anonymous coward on Slashdot, but referring to trans women as "transgenderred [sic] males" is just a little too offensive to let pass. There's good evidence to suggest that transsexualism is caused by congenital differences in certain parts of the brain involved with gender identity and sexuality [1] [2]. Essentially, we really are born with female brains. There's also evidence to suggest that a number of anatomical features of our brains shift to opposite-sex proportions during the first few months on hormones [3]. As for your ideas on explaining the sex ratio in technology jobs by men having a wider IQ distribution than women, I just don't think the numbers hold up. While there seems to be some support for men's IQ distribution having a larger standard deviation than women's, it just isn't enough to support your claim that "the chance of a person having an IQ of 125 is eight times more likely for males than females." The idea certainly isn't an implaausible one at extremely high IQ scores; all it would take would be an X-linked trait that sufficiently influences IQ for people with XY chromosomes (mostly masculine-gendered) to have a wider distribution than people with XX chromosomes (mostly feminine-gendered), and XY people *do* have a wider range of variation on a lot of different traits, probably because of a mechanism like this. The problem is that the effect would be just too small to make that big a difference; some quick googling turned up at most support for a one or two point difference in standard deviation, and that certainly wouldn't lead to a factor of eight difference at scores as low as 125. A one point difference in male versus female standard deviation of IQ in a population with an overall standard deviation of 16 points leads to men being twice as likely as women to have an IQ over 150, and only one person in a thousand of either sex scores that high. No matter what inflated opinions Slashdotters might have of themselves, I can say very assuredly that the great majority of IT workers do not have 150+ IQs. This sort of theory is probably only useful for explaining sex ratios in elite groups like Nobel prize winners, and I'd be rather skeptical even then of the assumption that the distribution of extremely high IQ scores can be obtained just by extrapolating the distribution for the middle-scoring bulk of the population. [1] A Sex Difference in the Human Brain and its Relation to Transsexuality [2] Male-to-Female Transsexuals Have Female Neuron Numbers in a Limbic Nucleus. [3] Changing your sex changes your brain: influences of testosterone and estrogen on adult human brain structure
As a polyamorous lesbian transsexual, I'd be the perfect candidate. :)
Except that I can say from personal experience that I still have plenty of sexual desire left after chemically supressing pretty much all of my testosterone production.
On the other hand, I'm physiologically incapable of either becoming pregnant or getting anyone else pregnant.
Yay! Mars, here I come!
Rubber Alligator. They can't seem to decide whether it's a crocodile or an alligator, but apparently it's "irresitable".
Obviously, so he can have his zombie lawyer eat the prosecutor's brain.
We've already got it.
I notice you don't mention him actually doing anything that would have harmed the company.
Light pressure is just incident power divided by the speed of light. With the half rotating toward the sun shielded as you described, Venus displays an area of 5.75 * 10^13 m^2, and is located about 1.08 * 10^11 m from the sun. The sun has a luminosity of about 3.8 * 10^26 W, so at Venus' orbit the irradiance from the sun is about 2592 W/m^2, so the unshaed half of Venus receives an incident power of 1.49 * 10^17 W, translating to a force of 4.97 * 10^8 N. If we make the approximation that this is all concentrated on a lever arm of half the radius of Venus, this gives a torque of about 1.5 * 10^15 N*m. Venus has a moment of inertia of 7.13 * 10^37 kg*m^2, so this torque would cause an angular acceleration of 2.11 * 10^-23 radians/s^2. The current angular velocity of Venus is 2.99 * 10^-7 radians/s, and the desired angular velocity is 7.27 * 10^-5 radians/s, so it would take 3.43 * 10^18 seconds so spin up Venus this way, or about 109 billion years. That's neglecting a factor of two depending on reflection (this is correct if Venus absorbs all the incident radiation), and the fact that the sun's power output will not remain constant over that time frame, nor, very probably, will the existence of Venus.
Terraforming Mars is probably a lot more practical than Venus. The big problem with Mars is the low gravity, so light gasses in the atmosphere leak away to space. I know Mars can't retain hydrogen, for example, but I'm not sure whether it could hold an oxygen atmosphere. It seems to manage to hold carbon dioxide just fine. Anyway, if you gave it an Earth-like atmosphere, it would probably be able to hold onto it, if not indefinitely, for millions of years, not just a few thousand, so even if it won't be permanent with continuing maintenance it still might be worth it.
Venus only rotates once every 243 days. You'd have to speed it up quite a bit or you'd just have huge temperature differences between the day and night sides and giant wind storms in between. Approximating Venus as a sphere of uniform density, it has a moment of inertia of and a rotational kinetic energy of 3.19 * 10^24 J. To accelerate it to the same rotation rate as Earth, you would have to raise that to 1.88 * 10^29 J. On the other hand a typical comet has a mass of up to 10^15 kg. They do move very fast, though, due to their highly eccentric orbits. That pages quotes Halley's comet as having an orbital velocity near its perihelion of 68 km/s. Using that velocity and the 10^15 kg mass figure, one comet would have a kinetic energy of about 2 * 10^24 J. In other words, even if you could couple the comits' energy into Venus' rotation with 100% efficiency (yeah, right), you would still need to hit it with about 150,000 comets to get it up to Earth-like rotation rates. This does not strike me as a particularly practical plan.
It's called a geosynchronous orbit.
You want to cure "compliance, order, and irrational submission to authority" by having an authority tell people what to believe? I think you need to think that one through a little more.
He said the kid was using a machine at a university in Sweden, not that he attended said university. Swedes can have insecure machines too, you know. Anyway, I was in university at 16, and was done with high school at 15, so it isn't that much of a stretch for this kid to be in university at 15. I could have been if I hadn't spent that year sitting around goofing off and reading Misner, Thorne and Wheeler instead.
Try a neutron bomb.
Do you really expect anyone on Slashdot to know what 'littoral' means?