Comparing to cars isn't quite fair. For one thing, Voyager doesn't have to deal with traffic, accelerating, decelerating, etc. Its commute is more like "Hang a left with gravity assist at Jupiter and cruise for the next several years." It doesn't even have to gas up.
It doesn't have to deal with pot holes, school zones, daredevil drivers, stop-n-go traffic, or heck, stopping and going in any sense. It's been slingshotted into space and can make subtle adjustments to its course or pick up the occasional planetary gravity-assist, and that's about it. If it were to totally croak today, it'd still keep flying into the interstellar medium. We just wouldn't be able to hear about it if the telemetry gear croaked.
And before you go on to praise the unprecedented electronics, realize there's plenty of people still developing home brew games for Intellivisions and Ataris, and they're almost as old... All those old consumer electronics gizmos still function, and were produced at a fraction of the cost. The only reason newer electronics haven't run for as long is, well, not enough time has passed for them to have run that long.
You can get anything you want, at Alice's Asteroid
You can get anything you want,
At Alice's As-ter-oid
Float right in, orbit's round the back
Half an AU from the railroad track
You can get anything you want, at Alice's Asteroid.
You could design a single-use code that isn't a random pad, such as assigning meanings to sequences of letters, in essence making a set of "words" for that mission. For example, "CQ" might mean "soldiers", "TQ" might mean tanks, etc. Notice in the ciphertext, the triad JRZ is repeated twice and the first and last 5 characters are the same.
That said, spot checking a few letters, it appears the distribution is pretty flat, suggesting an OTP. If you strike the last 5 letters (assume they're a repeat of the first 5, a sort of framing protocol), you'd expect each letter of the alphabet to get used around 5 times, and that's about what I see.
One Time Pads are indeed impossible to crack, if the pad is indeed used only one time, and is indeed fully random. That's because it makes any message that's the same length as or shorter than the ciphertext equally likely.
If I sent you the ciphertext PDXS, how would you know if it decoded to "EAST", "WEST", or any other four letter word?
I imagine you're also paying electricity, maintenance, serving, mirroring, backups, etc. as opposed to getting an empty, unmaintained, unpowered, unconnected hard drive for that price.
Actually, that's already the reality in the business I'm in. We have ECC on our large L2 and L3 on-chip memories, and have had for 5-6 years. (We added parity checking to the L2 memories around 8-9 years ago, and upgraded to SECDEC a few years later.) We're now adding ECC and parity even at the L1 and cache tag level, as others have.
From what I hear, the fab has informed us that some RAMs at smaller geometries will need active ECC just due to leakage and marginality. The traditional soft error model (ie. an alpha particle knocked charge loose from a cluster of bits) doesn't apply. I may have bits that are persistently unreliable and unrepairable, and so look closer to a hard error than a soft error. ECC can help us cope with these more novel memories. It's becoming more and more a way of life.
Somehow I doubt it means you must always use the absolute latest methods. It just means you need to use an appropriate method taking into account current knowledge. Dirt berms are as old as the hills, so to speak, but it can be entirely appropriate for a civil engineer to specify one as part of a road's right-of-way to control drainage, for example. Likewise, I don't have to write my code in the latest, greatest language de jure if plain old C is appropriate to the task.
Shouldn't a certain number of crashes just be expected? Yes, BUT: Digital Computers allow for no entropy increase, therefore would be immune to the second law of thermodynamics. Except for physical damage, manufacturing tolerance, radiation damage, and operator error. (most notably, programmer error). Generally the first two are caught early, long before the final customer sees the device, and the third only really applies if the computer is going into space or is to be deployed near a strong radiation source.
You do get bit flips in memories (aka. soft errors), and occasionally other problems. There's a failure rate associated with each component known as the Failure In Time (FIT) rate, and bit-flips in memories is a rather common cause. For modern processors with large amounts of memory on-chip (in the form of caches, reorder buffers, etc.), we're seeing increasing need for ECC codes on-chip. Off-chip, servers have had ECC memory for a very long time. Consumer equipment too often lacks ECC, and so many crashes could come from soft errors.
So, I don't think it's correct to say that "deployed near a strong radiation source" is necessary to see flaky behavior from a modern system, even with perfect hardware.
My wife's clamshell iBook (an original 3GB Blueberry) came with MacOS 8.6 as I recall and we upgraded it to MacOS 9. The "System" part of the name was already gone by then. They switched from "System" to "MacOS" with MacOS 7.6.
They have kept the numbers monotonic, but since arriving at OS X, they've only been incrementing the number after the decimal point. Whereas others started only incrementing the leading digit (witness Solaris 2.6 => Solaris 7, for example), Apple's been on 10.something for MacOS X / OS X for 11 years now.
Well, for Ubuntu's version numbers, they go with Year.Month, so not only do you know what year something came out, but also what month. So, 12.04 came out this last April, for example. I couldn't tell you what animal is associated with it and don't even care.
If you think about it, this isn't specific to computers. Look at car names and their goofy letter/number combos, for example. Hell, some consultant got a million bucks a letter selling Infiniti on the letters J and Q.
At least BMW's are usually rooted in functionality, at least for their sedans (series number and engine displacement in deciliters). But quick, which is bigger/faster/better/whatever: a Cadillac STS or Cadillac DTS? If I told you one was the Seville Touring Sedan and the other was the DeVille Touring Sedan, would that even help?
You partially touch on this (eating fish vs. breathing air), but it's worth pointing out an additional component explicitly: What's radioactive, and is it something that your body will absorb and hold onto? For example, your thyroid uses iodine to make hormones. It'll happily absorb up a bunch of radioactive iodine, which is Bad News.
Mod parent up. I came here to post Lockhart's Lament.
And, I happen to disagree with dcollins's reply to your comment. It's perfectly possible to be erudite without knowing what a gerund is, how to diagram participles, etc. It's also perfectly possible to have interesting mathematical insight without the rigor. It's also perfectly possible to make interesting music without a degree in music theory.
The formality does help when it comes time to refine your craft. But, it's a barrier to entry that's wasted on most people. You have to appreciate the art of and wonder mathematics before it makes sense to dive into the full formality.
The cost of hardware also dropped rather dramatically in those 30 years, such that there were actually enough computing resources outside the government and outside large corporations worth hooking up.
It's like saying the government sat on color TV for 20+ years because the FCC didn't allow national color broadcasts until the 50s, despite the first color broadcast being made in 1928. The fact of the matter was that it took that long for the technology to become sufficiently robust, demand great enough, and potential content plentiful enough to justify standardizing and getting it out there.
I'd say the story's similarly true for the Internet. All that explosive growth through the 80s and 90s also paralleled the explosive growth of minicomputers and microcomputers everywhere. It took off more because it was an idea whose time had come than for any other reason.
You're confusing "normal" with "majority" or maybe "common." Something can be perfectly normal without the majority of people doing it or it even being particularly common. For example, it's perfectly normal to have a pet bird, although it's fairly uncommon: only 4% of households have a pet bird.
WTF does "liberal" have to do with it? Now, persecution complexes: Those aren't normal, although they're fairly common amongst a certain sort...
You're confusing "normal" with "majority" or maybe "common."
Something can be perfectly normal without the majority of people doing it or it even being particularly common. For example, it's perfectly normal to have a pet bird, although it's fairly uncommon: only 4% of households have a pet bird.
Calm down. There is a way! What you need to do is buy at least three different varieties of aluminum foil: Reynolds, the store's in-house brand, and one other name brand (any name brand will do, so long as it is NOT the store brand and NOT Reynolds). IMPORANT: Make sure you buy the regular thickness aluminum foil, not the heavy duty! For best results, make sure all roles are at least 12" wide. (Standard Reynolds is about 12.5", which is perfect.)
The goal is to trigger cross interference between the different coherent sources that may be affecting you through a single brand of foil. This is why you need three different brands: Like color filters, if you combine magenta and cyan, you get red; if you combine cyan and yellow, you get green; if you combine magenta and yellow you get blue. But, if you combine all three, you get black -- all light absorbed. Mind you, that's not exactly what we're doing here since mind control lasers don't work exactly like that, but it's a useful metaphor.
The most effective method I know is to pull off strips of foil from each roll. Make each strip approximately 8" long, and then fold it in half 3 times, to make a strip approximately wide. You will need at least 8 such strips from each of the three rolls. IMPORTANT: Do not mix up the strips! Ordering is important!
Now, weave the strips together into a mesh, alternating strips from each of the three rolls -- one strip of Reynolds, one strip of store brand, one strip of other brand. There should be 12 horizontal strips and 12 vertical strips in the mesh altogether.
The resulting mesh will have overlap points for all possible combinations of the three mind control signals. This will fractalize the signals, causing them to decohere, rendering them effectively inoperative.
x32 user-space programs require an x64 kernel. So, the OS would be able to access the full address space. The restriction is only that individual processes are limited to 4GB per process. The x32 mode is really more like a "small model 64-bit mode" than a "fancy 32-bit mode."
So, even if it was "impossible" to build a system around a given chip / chipset with less than 8GB of RAM, an x32-based system would still be able to use all 8 GB from kernel space (since the kernel is x64), despite the fact no single x32 process can directly address[*] more than 4GB at once.
------------------
[*] The important part here is directly address. A single x32 task can indirectly address as much memory as it likes, for example, by mmapping time-varying portions of a large file stored in tmpfs. All of that memory would still be in RAM so long as it all fit.
If I have connection sharing enabled over SSH, I find it'll even autocomplete filenames on remote hosts for me for 'scp' if I already have an ssh connection open to that host. Useful, but creepy at first.
You did, when you asked if Irfanview can uuencode and post to usenet.
That's because most GUIs, even those that allow batch operations, only allow you to do so within the boundaries of that app. This is in contrast to the command line (well, the UNIX command line at least), which derives its power from stringing lots of little tools together rather than relying on a feature being built into a monolithic app.
Why the different measuring stick? Because the default unit of interaction is very different between GUI and CLI, and that's the source of their differing strengths and weaknesses. With GUIs, you most often deal with a large, monolithic app, and if what you want to do isn't built in or available as an app-specific plugin, the app won't do it. With CLIs, you most often deal with many smaller programs that were designed to be strung together in arbitrary ways.
Writing a single command like that is only worth the effort if you plan to reuse it. Otherwise, you're just going to do everything in it one step at a time, which is exactly how you would use GUI tools. No CLI necessary.
If I end up needing to reuse it, I can just copy/paste it into a file and "tada"! A new shell script. (cat > script, [highlight then middle-click], [Ctrl-D], chmod +x script)
But, most of the time when I write such long involved command lines, it's without the intent of reuse, believe it or not. Or only the limited reuse that's facilitated by the command history buffer. Ask my coworkers -- this is how I use UNIX, and I find it exceptionally powerful.
You posit that "it's exactly how you would use GUI tools." No, it isn't. It's not even close.
See, I built everything into that command up-front. The whole pipeline of operations is in one command. I don't have to load program A, do steps 1-2, load program B, do step 3, load program C, do step 4.
With the command line I gave above, I don't have to keep re-selecting files, or anything. I don't have to babysit the process. Once I hit enter and everything looks underway, I can go do something else entirely while that processes. It's zero-touch after hitting enter, for the whole pipeline of operations.
So, yeah, you could complete the task entirely without the CLI. But, you'll be spending much more time handholding the computer through the process.
Slashdot Mirror
Comparing to cars isn't quite fair. For one thing, Voyager doesn't have to deal with traffic, accelerating, decelerating, etc. Its commute is more like "Hang a left with gravity assist at Jupiter and cruise for the next several years." It doesn't even have to gas up.
It doesn't have to deal with pot holes, school zones, daredevil drivers, stop-n-go traffic, or heck, stopping and going in any sense. It's been slingshotted into space and can make subtle adjustments to its course or pick up the occasional planetary gravity-assist, and that's about it. If it were to totally croak today, it'd still keep flying into the interstellar medium. We just wouldn't be able to hear about it if the telemetry gear croaked.
And before you go on to praise the unprecedented electronics, realize there's plenty of people still developing home brew games for Intellivisions and Ataris, and they're almost as old... All those old consumer electronics gizmos still function, and were produced at a fraction of the cost. The only reason newer electronics haven't run for as long is, well, not enough time has passed for them to have run that long.
You can get anything you want, at Alice's Asteroid
You can get anything you want,
At Alice's As-ter-oid
Float right in, orbit's round the back
Half an AU from the railroad track
You can get anything you want, at Alice's Asteroid.
You could design a single-use code that isn't a random pad, such as assigning meanings to sequences of letters, in essence making a set of "words" for that mission. For example, "CQ" might mean "soldiers", "TQ" might mean tanks, etc. Notice in the ciphertext, the triad JRZ is repeated twice and the first and last 5 characters are the same.
That said, spot checking a few letters, it appears the distribution is pretty flat, suggesting an OTP. If you strike the last 5 letters (assume they're a repeat of the first 5, a sort of framing protocol), you'd expect each letter of the alphabet to get used around 5 times, and that's about what I see.
One Time Pads are indeed impossible to crack, if the pad is indeed used only one time, and is indeed fully random. That's because it makes any message that's the same length as or shorter than the ciphertext equally likely.
If I sent you the ciphertext PDXS, how would you know if it decoded to "EAST", "WEST", or any other four letter word?
I imagine you're also paying electricity, maintenance, serving, mirroring, backups, etc. as opposed to getting an empty, unmaintained, unpowered, unconnected hard drive for that price.
Actually, that's already the reality in the business I'm in. We have ECC on our large L2 and L3 on-chip memories, and have had for 5-6 years. (We added parity checking to the L2 memories around 8-9 years ago, and upgraded to SECDEC a few years later.) We're now adding ECC and parity even at the L1 and cache tag level, as others have.
From what I hear, the fab has informed us that some RAMs at smaller geometries will need active ECC just due to leakage and marginality. The traditional soft error model (ie. an alpha particle knocked charge loose from a cluster of bits) doesn't apply. I may have bits that are persistently unreliable and unrepairable, and so look closer to a hard error than a soft error. ECC can help us cope with these more novel memories. It's becoming more and more a way of life.
Somehow I doubt it means you must always use the absolute latest methods. It just means you need to use an appropriate method taking into account current knowledge. Dirt berms are as old as the hills, so to speak, but it can be entirely appropriate for a civil engineer to specify one as part of a road's right-of-way to control drainage, for example. Likewise, I don't have to write my code in the latest, greatest language de jure if plain old C is appropriate to the task.
You do get bit flips in memories (aka. soft errors), and occasionally other problems. There's a failure rate associated with each component known as the Failure In Time (FIT) rate, and bit-flips in memories is a rather common cause. For modern processors with large amounts of memory on-chip (in the form of caches, reorder buffers, etc.), we're seeing increasing need for ECC codes on-chip. Off-chip, servers have had ECC memory for a very long time. Consumer equipment too often lacks ECC, and so many crashes could come from soft errors.
So, I don't think it's correct to say that "deployed near a strong radiation source" is necessary to see flaky behavior from a modern system, even with perfect hardware.
My wife's clamshell iBook (an original 3GB Blueberry) came with MacOS 8.6 as I recall and we upgraded it to MacOS 9. The "System" part of the name was already gone by then. They switched from "System" to "MacOS" with MacOS 7.6.
They have kept the numbers monotonic, but since arriving at OS X, they've only been incrementing the number after the decimal point. Whereas others started only incrementing the leading digit (witness Solaris 2.6 => Solaris 7, for example), Apple's been on 10.something for MacOS X / OS X for 11 years now.
Well, for Ubuntu's version numbers, they go with Year.Month, so not only do you know what year something came out, but also what month. So, 12.04 came out this last April, for example. I couldn't tell you what animal is associated with it and don't even care.
If you think about it, this isn't specific to computers. Look at car names and their goofy letter/number combos, for example. Hell, some consultant got a million bucks a letter selling Infiniti on the letters J and Q.
At least BMW's are usually rooted in functionality, at least for their sedans (series number and engine displacement in deciliters). But quick, which is bigger/faster/better/whatever: a Cadillac STS or Cadillac DTS? If I told you one was the Seville Touring Sedan and the other was the DeVille Touring Sedan, would that even help?
...which to some, may be considered porn. Rule 34, after all.
You partially touch on this (eating fish vs. breathing air), but it's worth pointing out an additional component explicitly: What's radioactive, and is it something that your body will absorb and hold onto? For example, your thyroid uses iodine to make hormones. It'll happily absorb up a bunch of radioactive iodine, which is Bad News.
Mod parent up. I came here to post Lockhart's Lament.
And, I happen to disagree with dcollins's reply to your comment. It's perfectly possible to be erudite without knowing what a gerund is, how to diagram participles, etc. It's also perfectly possible to have interesting mathematical insight without the rigor. It's also perfectly possible to make interesting music without a degree in music theory.
The formality does help when it comes time to refine your craft. But, it's a barrier to entry that's wasted on most people. You have to appreciate the art of and wonder mathematics before it makes sense to dive into the full formality.
So this Dustin Hoffman guy wants to track all the planes. Track all the planes! What is he, some kind of Rain Man?
The cost of hardware also dropped rather dramatically in those 30 years, such that there were actually enough computing resources outside the government and outside large corporations worth hooking up.
It's like saying the government sat on color TV for 20+ years because the FCC didn't allow national color broadcasts until the 50s, despite the first color broadcast being made in 1928. The fact of the matter was that it took that long for the technology to become sufficiently robust, demand great enough, and potential content plentiful enough to justify standardizing and getting it out there.
I'd say the story's similarly true for the Internet. All that explosive growth through the 80s and 90s also paralleled the explosive growth of minicomputers and microcomputers everywhere. It took off more because it was an idea whose time had come than for any other reason.
And it would come without batteries.
You're confusing "normal" with "majority" or maybe "common." Something can be perfectly normal without the majority of people doing it or it even being particularly common. For example, it's perfectly normal to have a pet bird, although it's fairly uncommon: only 4% of households have a pet bird.
WTF does "liberal" have to do with it? Now, persecution complexes: Those aren't normal, although they're fairly common amongst a certain sort...
You're confusing "normal" with "majority" or maybe "common." Something can be perfectly normal without the majority of people doing it or it even being particularly common. For example, it's perfectly normal to have a pet bird, although it's fairly uncommon: only 4% of households have a pet bird.
Calm down. There is a way! What you need to do is buy at least three different varieties of aluminum foil: Reynolds, the store's in-house brand, and one other name brand (any name brand will do, so long as it is NOT the store brand and NOT Reynolds). IMPORANT: Make sure you buy the regular thickness aluminum foil, not the heavy duty! For best results, make sure all roles are at least 12" wide. (Standard Reynolds is about 12.5", which is perfect.)
The goal is to trigger cross interference between the different coherent sources that may be affecting you through a single brand of foil. This is why you need three different brands: Like color filters, if you combine magenta and cyan, you get red; if you combine cyan and yellow, you get green; if you combine magenta and yellow you get blue. But, if you combine all three, you get black -- all light absorbed. Mind you, that's not exactly what we're doing here since mind control lasers don't work exactly like that, but it's a useful metaphor.
The most effective method I know is to pull off strips of foil from each roll. Make each strip approximately 8" long, and then fold it in half 3 times, to make a strip approximately wide. You will need at least 8 such strips from each of the three rolls. IMPORTANT: Do not mix up the strips! Ordering is important!
Now, weave the strips together into a mesh, alternating strips from each of the three rolls -- one strip of Reynolds, one strip of store brand, one strip of other brand. There should be 12 horizontal strips and 12 vertical strips in the mesh altogether.
The resulting mesh will have overlap points for all possible combinations of the three mind control signals. This will fractalize the signals, causing them to decohere, rendering them effectively inoperative.
Good luck, my friend. Be strong!
These days, shouldn't it also try "sudo ./pwn" and/or "sudo -s"?
I came here to say that...
...which reminds me of this hilarious and relevant bit of Fry and Laurie.
x32 user-space programs require an x64 kernel. So, the OS would be able to access the full address space. The restriction is only that individual processes are limited to 4GB per process. The x32 mode is really more like a "small model 64-bit mode" than a "fancy 32-bit mode."
So, even if it was "impossible" to build a system around a given chip / chipset with less than 8GB of RAM, an x32-based system would still be able to use all 8 GB from kernel space (since the kernel is x64), despite the fact no single x32 process can directly address[*] more than 4GB at once.
------------------
[*] The important part here is directly address. A single x32 task can indirectly address as much memory as it likes, for example, by mmapping time-varying portions of a large file stored in tmpfs. All of that memory would still be in RAM so long as it all fit.
If I have connection sharing enabled over SSH, I find it'll even autocomplete filenames on remote hosts for me for 'scp' if I already have an ssh connection open to that host. Useful, but creepy at first.
That's because most GUIs, even those that allow batch operations, only allow you to do so within the boundaries of that app. This is in contrast to the command line (well, the UNIX command line at least), which derives its power from stringing lots of little tools together rather than relying on a feature being built into a monolithic app.
Why the different measuring stick? Because the default unit of interaction is very different between GUI and CLI, and that's the source of their differing strengths and weaknesses. With GUIs, you most often deal with a large, monolithic app, and if what you want to do isn't built in or available as an app-specific plugin, the app won't do it. With CLIs, you most often deal with many smaller programs that were designed to be strung together in arbitrary ways.
If I end up needing to reuse it, I can just copy/paste it into a file and "tada"! A new shell script. (cat > script, [highlight then middle-click], [Ctrl-D], chmod +x script)
But, most of the time when I write such long involved command lines, it's without the intent of reuse, believe it or not. Or only the limited reuse that's facilitated by the command history buffer. Ask my coworkers -- this is how I use UNIX, and I find it exceptionally powerful.
You posit that "it's exactly how you would use GUI tools." No, it isn't. It's not even close.
See, I built everything into that command up-front. The whole pipeline of operations is in one command. I don't have to load program A, do steps 1-2, load program B, do step 3, load program C, do step 4.
With the command line I gave above, I don't have to keep re-selecting files, or anything. I don't have to babysit the process. Once I hit enter and everything looks underway, I can go do something else entirely while that processes. It's zero-touch after hitting enter, for the whole pipeline of operations.
So, yeah, you could complete the task entirely without the CLI. But, you'll be spending much more time handholding the computer through the process.