I did discuss it with the Linux kernel maintainers at the time (over a year ago -- I don't recall exactly when), and they didn't seem to care much. But no, I didn't file a bug report.
I did, however, take a look at the relevant kernel code. It didn't look like an easy fix (else I would have attempted it myself). It looked like it would take at least a partial rewrite of the VM subsystem.
If it were fixed, would I come back to Linux? Probably not on that machine. Oh, I should have mentioned another nice thing Solaris did for me -- the new Fault Manager pinpointed a hardware problem (a marginal DIMM) that had been causing occasional crashes for months.
On the down side, getting Solaris running on that machine initially was a major pain. To make a long story short, the motherboard's onboard SCSI controller caused some conflict that kept Solaris from booting (though Linux had no trouble with it). I had to turn it off in the BIOS and buy a PCI-X card to get SCSI. But we all know that motherboard compatibility and driver support are Solaris' weak point.
This is not a problem that can be tuned away. I'll tell you exactly what's going on. The kernel has, for each process, a table which contains one entry for each contiguous region of address space with the same page protections. Since the Lisp implementation I'm using makes use of page protections to implement its GC write barrier -- a very useful technique for an SMP garbage collector -- it creates lots of small regions, so that this table gets quite large. And, there are algorithms in the kernel that are quadratic, or worse, in the size of the table.
The result is that as the Lisp heap grows past a couple of GB, one of the CPUs (I'm doing this, BTW, on a quad Opteron with 16GB of DRAM) comes to spend 100% of its time in the kernel, doing whatever this quadratic algorithm is doing, and the machine becomes pretty much unresponsive.
Solaris has no trace of this behavior. Clearly, it comes from a culture where the OS is expected to scale in many dimensions, and quadratic algorithms are strictly forbidden.
I'm using Solaris because the data mining application I'm building (in Lisp) brings the Linux kernel absolutely to its knees. Solaris runs it just fine on the same hardware. (We're talking 30+ GB of heap -- Linux is dead meat after 3 to 4 GB.)
A friend of mine says this is because the Linux kernel hackers optimize for the common case, not for extreme cases. I suspect this is correct. To put it another way, they are more into cycle shaving than analyzing the time and space complexity of their algorithms -- just as one might expect from smart hackers with a relatively weak computer science background.
The result is a kernel that does great on normal workloads, but just falls over when subjected to unusual stresses. Unless and until this is corrected, there will be a need for Solaris.
A lot of us have given up on ATI because their Linux driver support is so poor.
I'll wager that if you were willing to swap in an Nvidia card, you could install almost any distro with little difficulty. You might be able to find a compatible card on eBay for $200 or so.
That said, I found a thread for you. I found this by searching Google on "linux dell e1505 x1400". This is generally how I've worked through past Linux hardware issues. (Maybe you already did that.)
I can't promise that will work, though. All I can tell you for certain is that the last time I went through this, I tossed the ATI card and got an Nvidia. I know, it sucks, but it's really ATI's fault, not Linux's.
Until a few years ago, a lot of SCSI drives were formattable with arbitrary sector sizes (I don't know if they could take absolutely any size in a given range, but it didn't have to be a power of 2). I have an old machine (from the mid-1980s) whose drive is formatted, IIRC, in 2560 byte sectors.
16k is too big because it's larger than the page size used by most OSes. 4k is a good choice. The additional improvement in space efficiency from going to 16k vs. 4k would be small.
Q W D F J Y U K L P A S E T G H N I O R Z X C V ; B M
(Please imagine the second and third rows slid to the right appropriately; I don't see how to get Slashdot to indent them.)
My design goals were very similar to those of the Asset keyboard, and as you can see I came up with a very similar layout. I call it YAKL, for -- you guessed it -- "Yet Another Keyboard Layout".
I would never go back to QWERTY; YAKL keeps me on the home row much more, and is accordingly more comfortable and (I'm reasonably sure) faster. But I don't know that I would recommend YAKL, or Asset, to anyone else. There are two reasons. First, as I discovered, learning a new layout isn't just about learning the location of each key individually; it's about learning digraphs and trigraphs. Since the whole point of YAKL was to get more frequently used keys onto the home row, as it turns out, almost every common digraph and trigraph uses at least one letter that has been moved. Bottom line: it still took me a long time to learn the new layout. I don't think it would have taken much longer to learn Dvorak.
And the advantage of Dvorak over YAKL (or any other homebrew layout) is that it's standardized, and you can select it from the layout menu on all the major OSes; you don't have to go through the contortions of creating custom layouts.
You don't sound like someone who has experienced what a properly set up MS solution can offer. Maybe that's the main source of discontent here.
Perhaps. But I wasn't arguing that Macs are necessarily superior (though obviously I have my personal preference). I was simply claiming that they are just as good for any uses I or my colleagues have (except for developing Windows software, for which Windows is obviously necessary).
Actually I don't work in a situation where the desktops are all controlled by IT -- it's a small company, and we're all fairly technical, so we manage our own. That said, if a company of the kind you describe wanted to use Macs, the technique of running a standard image on every machine could be applied perfectly well.
The point I really want to make here is that you were making a very strong statement about Windows being "simply the best" and that those who prefer Macs have irrational "loyalty" when you don't seem to have much experience with or knowledge about Macs to base that on. Sounds like you have some loyalty too.
The way I see it, people who adamantly choose Apple or some 'innovator' solution are the ones who have loyalty involved with their decisions instead of sheer usability/efficiency etc. People who choose Microsoft do so because it is simply the best solution in today's workplace for productivity.
I see -- "simply the best", but you're not a fanboy, eh?
Your information is out of date. It is perfectly possible to be just as productive on a Mac as on Windows -- probably more so, when you consider how much less maintenance and fiddling the OS itself requires. Yes, I use a Mac in a mostly-Windows workplace. Yes, I also run Microsoft Office; but I don't even run Windows in a virtual machine, because I just have no reason to.
For me it's really simple. If I can run the PDA apps I care about on the iPhone, I will want one; and if I can't, I won't. Remains to be seen which is the case.
How good is WPA these days, anyway? If I put a strong password on the router, use WPA-PSK with a strong key, turn off SSID broadcast, and allow connections only from specified MAC addresses, how safe am I? (This is a home network in an apartment complex; I can see 10 or 12 other SSIDs from here. So I'm inclined to think that turning off SSID broadcast will keep me under anyone's radar, since there are other, less well secured networks in the vicinity. Still it would be nice not to be completely counting on that.)
I would suggest the term "plutocracy". "Fascism", as some other posters have suggested, is different -- while fascism may be highly compatible with plutocracy, they don't necessarily go together (would you call Cuba a plutocracy?) -- and "fascism" is obviously a much more loaded term.
"Corporate welfare" is not a bad description either.
From TFA: The Prius costs an average of $3.25 per mile driven over a lifetime of 100,000 miles - the expected lifespan of the Hybrid. The Hummer, on the other hand, costs a more fiscal $1.95 per mile to put on the road over an expected lifetime of 300,000 miles.
300,000 miles out of a Hummer? Give me a break! I'll bet the average life is about half that. This more reasonable assumption puts the Hummer's cost per mile at a whopping $3.90.
That's so cool that people still remember Hunt the Wumpus. The author of that game is an old friend of mine -- I'll mention this the next time I see him.
Check out JML which has ESC/Java2 to provide static contract checking for Java, Spec# (C# with contracts) which uses the Spec# verifier for static checking of contracts, and Eiffel with ESpec-Verify for static checking of Eiffel contracts.
While these systems can statically verify certain relatively simple kinds of contracts, nobody knows how to build a theorem prover powerful enough to verify most interesting contracts without human guidance.
Specifically, we can build fully automatic provers for first-order logic; but most interesting contracts have inductive proofs, and induction is inherently higher-order, and nobody knows how to build a fully (or even mostly) automated prover for higher-order logic.
Personally, I think it's one of the most interesting problems in all of AI, and I'm surprised that not many people seem to be working on it.
Have you read Cory Doctorow's essay on "Intellectual Property in the Web 2.0 age"? It's hard to find now, because he complained about people stealing it and putting in their websites so they took it down
ROFL!!! No, I haven't seen it. You really posted that without seeing the irony???
Why does everyone here want this not to work? Seems to me this could be the alternative to DRM. It doesn't interfere with fair use at all; it only detects when copyrighted works are made widely available.
If we want to dissuade the entertainment industry from using DRM, it seems incumbent upon us, as technologists, to propose alternatives that at least partially answer copyright owners' legitimate concerns. Seems to me this could be one of them.
Definitely possible. I have an AMD64 box running OpenSolaris that would crash frequently when I ran a certain application, and would crash much more rarely otherwise. I suspected OpenSolaris, but the problem turned out to be a bad DIMM.
(In fact, recent OpenSolaris builds have a feature called the Fault Manager which, once I discovered it, told me exactly which DIMM was bad (based on single-bit ECC correction rates). Very very cool.)
Slashdot Mirror
I did discuss it with the Linux kernel maintainers at the time (over a year ago -- I don't recall exactly when), and they didn't seem to care much. But no, I didn't file a bug report.
I did, however, take a look at the relevant kernel code. It didn't look like an easy fix (else I would have attempted it myself). It looked like it would take at least a partial rewrite of the VM subsystem.
If it were fixed, would I come back to Linux? Probably not on that machine. Oh, I should have mentioned another nice thing Solaris did for me -- the new Fault Manager pinpointed a hardware problem (a marginal DIMM) that had been causing occasional crashes for months.
On the down side, getting Solaris running on that machine initially was a major pain. To make a long story short, the motherboard's onboard SCSI controller caused some conflict that kept Solaris from booting (though Linux had no trouble with it). I had to turn it off in the BIOS and buy a PCI-X card to get SCSI. But we all know that motherboard compatibility and driver support are Solaris' weak point.
+5, Insightful? Oh, come on, mods.
This is not a problem that can be tuned away. I'll tell you exactly what's going on. The kernel has, for each process, a table which contains one entry for each contiguous region of address space with the same page protections. Since the Lisp implementation I'm using makes use of page protections to implement its GC write barrier -- a very useful technique for an SMP garbage collector -- it creates lots of small regions, so that this table gets quite large. And, there are algorithms in the kernel that are quadratic, or worse, in the size of the table.
The result is that as the Lisp heap grows past a couple of GB, one of the CPUs (I'm doing this, BTW, on a quad Opteron with 16GB of DRAM) comes to spend 100% of its time in the kernel, doing whatever this quadratic algorithm is doing, and the machine becomes pretty much unresponsive.
Solaris has no trace of this behavior. Clearly, it comes from a culture where the OS is expected to scale in many dimensions, and quadratic algorithms are strictly forbidden.
I'm using Solaris because the data mining application I'm building (in Lisp) brings the Linux kernel absolutely to its knees. Solaris runs it just fine on the same hardware. (We're talking 30+ GB of heap -- Linux is dead meat after 3 to 4 GB.)
A friend of mine says this is because the Linux kernel hackers optimize for the common case, not for extreme cases. I suspect this is correct. To put it another way, they are more into cycle shaving than analyzing the time and space complexity of their algorithms -- just as one might expect from smart hackers with a relatively weak computer science background.
The result is a kernel that does great on normal workloads, but just falls over when subjected to unusual stresses. Unless and until this is corrected, there will be a need for Solaris.
A lot of us have given up on ATI because their Linux driver support is so poor.
I'll wager that if you were willing to swap in an Nvidia card, you could install almost any distro with little difficulty. You might be able to find a compatible card on eBay for $200 or so.
That said, I found a thread for you. I found this by searching Google on "linux dell e1505 x1400". This is generally how I've worked through past Linux hardware issues. (Maybe you already did that.)
I can't promise that will work, though. All I can tell you for certain is that the last time I went through this, I tossed the ATI card and got an Nvidia. I know, it sucks, but it's really ATI's fault, not Linux's.
Good luck!
Until a few years ago, a lot of SCSI drives were formattable with arbitrary sector sizes (I don't know if they could take absolutely any size in a given range, but it didn't have to be a power of 2). I have an old machine (from the mid-1980s) whose drive is formatted, IIRC, in 2560 byte sectors.
16k is too big because it's larger than the page size used by most OSes. 4k is a good choice. The additional improvement in space efficiency from going to 16k vs. 4k would be small.
The longer block sizes add reliability because the error correcting codes have more to work with at a time (more data bits, but also more ECC bits).
As for wasted space, that's under the filesystem's control, not the drive's.
Windows XP is going to be looking very obsolete and dated within the year.
Depends on what you're using it for. I'm still getting good use out of a couple of Windows 2000 machines.
I use a layout I invented in 1992:
(Please imagine the second and third rows slid to the right appropriately; I don't see how to get Slashdot to indent them.)
My design goals were very similar to those of the Asset keyboard, and as you can see I came up with a very similar layout. I call it YAKL, for -- you guessed it -- "Yet Another Keyboard Layout".
I would never go back to QWERTY; YAKL keeps me on the home row much more, and is accordingly more comfortable and (I'm reasonably sure) faster. But I don't know that I would recommend YAKL, or Asset, to anyone else. There are two reasons. First, as I discovered, learning a new layout isn't just about learning the location of each key individually; it's about learning digraphs and trigraphs. Since the whole point of YAKL was to get more frequently used keys onto the home row, as it turns out, almost every common digraph and trigraph uses at least one letter that has been moved. Bottom line: it still took me a long time to learn the new layout. I don't think it would have taken much longer to learn Dvorak.
And the advantage of Dvorak over YAKL (or any other homebrew layout) is that it's standardized, and you can select it from the layout menu on all the major OSes; you don't have to go through the contortions of creating custom layouts.
You don't sound like someone who has experienced what a properly set up MS solution can offer. Maybe that's the main source of discontent here.
Perhaps. But I wasn't arguing that Macs are necessarily superior (though obviously I have my personal preference). I was simply claiming that they are just as good for any uses I or my colleagues have (except for developing Windows software, for which Windows is obviously necessary).
Actually I don't work in a situation where the desktops are all controlled by IT -- it's a small company, and we're all fairly technical, so we manage our own. That said, if a company of the kind you describe wanted to use Macs, the technique of running a standard image on every machine could be applied perfectly well.
The point I really want to make here is that you were making a very strong statement about Windows being "simply the best" and that those who prefer Macs have irrational "loyalty" when you don't seem to have much experience with or knowledge about Macs to base that on. Sounds like you have some loyalty too.
Yes, "throttling" is an excellent term for what I'd like to do to Comcast execs :)
The way I see it, people who adamantly choose Apple or some 'innovator' solution are the ones who have loyalty involved with their decisions instead of sheer usability/efficiency etc. People who choose Microsoft do so because it is simply the best solution in today's workplace for productivity.
I see -- "simply the best", but you're not a fanboy, eh?
Your information is out of date. It is perfectly possible to be just as productive on a Mac as on Windows -- probably more so, when you consider how much less maintenance and fiddling the OS itself requires. Yes, I use a Mac in a mostly-Windows workplace. Yes, I also run Microsoft Office; but I don't even run Windows in a virtual machine, because I just have no reason to.
For me it's really simple. If I can run the PDA apps I care about on the iPhone, I will want one; and if I can't, I won't. Remains to be seen which is the case.
How did the Nigerian try to defraud you?
How good is WPA these days, anyway? If I put a strong password on the router, use WPA-PSK with a strong key, turn off SSID broadcast, and allow connections only from specified MAC addresses, how safe am I? (This is a home network in an apartment complex; I can see 10 or 12 other SSIDs from here. So I'm inclined to think that turning off SSID broadcast will keep me under anyone's radar, since there are other, less well secured networks in the vicinity. Still it would be nice not to be completely counting on that.)
I would suggest the term "plutocracy". "Fascism", as some other posters have suggested, is different -- while fascism may be highly compatible with plutocracy, they don't necessarily go together (would you call Cuba a plutocracy?) -- and "fascism" is obviously a much more loaded term.
"Corporate welfare" is not a bad description either.
From TFA: The Prius costs an average of $3.25 per mile driven over a lifetime of 100,000 miles - the expected lifespan of the Hybrid. The Hummer, on the other hand, costs a more fiscal $1.95 per mile to put on the road over an expected lifetime of 300,000 miles.
300,000 miles out of a Hummer? Give me a break! I'll bet the average life is about half that. This more reasonable assumption puts the Hummer's cost per mile at a whopping $3.90.
That's so cool that people still remember Hunt the Wumpus. The author of that game is an old friend of mine -- I'll mention this the next time I see him.
Heh -- that bugs me too :) Though it's not as bad as the endless confusion of power with energy, leading to phrases like "megawatts per year".
Seconding this. Definitely check out SICP.
Check out JML which has ESC/Java2 to provide static contract checking for Java, Spec# (C# with contracts) which uses the Spec# verifier for static checking of contracts, and Eiffel with ESpec-Verify for static checking of Eiffel contracts.
While these systems can statically verify certain relatively simple kinds of contracts, nobody knows how to build a theorem prover powerful enough to verify most interesting contracts without human guidance.
Specifically, we can build fully automatic provers for first-order logic; but most interesting contracts have inductive proofs, and induction is inherently higher-order, and nobody knows how to build a fully (or even mostly) automated prover for higher-order logic.
Personally, I think it's one of the most interesting problems in all of AI, and I'm surprised that not many people seem to be working on it.
You assume that there will be no false-positives. There will be many.
I seriously doubt it. What do you base this on?
There are plenty of low-order bits in these files in which to hide a signature -- meaning the signature can be plenty long to avoid false positives.
I'd guess that countermeasures producing false negatives will be the larger problem, but if these people have a way to prevent that, good for them.
Have you read Cory Doctorow's essay on "Intellectual Property in the Web 2.0 age"? It's hard to find now, because he complained about people stealing it and putting in their websites so they took it down
ROFL!!! No, I haven't seen it. You really posted that without seeing the irony???
Why does everyone here want this not to work? Seems to me this could be the alternative to DRM. It doesn't interfere with fair use at all; it only detects when copyrighted works are made widely available.
If we want to dissuade the entertainment industry from using DRM, it seems incumbent upon us, as technologists, to propose alternatives that at least partially answer copyright owners' legitimate concerns. Seems to me this could be one of them.
Definitely possible. I have an AMD64 box running OpenSolaris that would crash frequently when I ran a certain application, and would crash much more rarely otherwise. I suspected OpenSolaris, but the problem turned out to be a bad DIMM.
(In fact, recent OpenSolaris builds have a feature called the Fault Manager which, once I discovered it, told me exactly which DIMM was bad (based on single-bit ECC correction rates). Very very cool.)
That's how I deal with it. Works pretty well as long as I can get a net connection.