Yeah, I saw those, but realistically if the TSA has the masters, what are the chances other folks do as well?
Probably pretty high. But as you point out with regards to just stealing the suitcase, you're trying to protect against a relatively limited form of crime of opportunism.
Without getting into value judgments about the situation, you can get locks which you can safely put on your luggage. The TSA has master keys that will open them so they can still search, but it will at least prevent average Joe from getting in. They aren't expensive either.
A real megabyte is divisible by an order of magnitude in the relevant counting system.
Why is it always 2 that's the relevant counting system?
For instance, take the hard drive example. Why are we seeing drives with 200 billion bytes, 250 billion bytes, 500 billion bytes, etc. if 2 is the relevant counting system? If there was something about hard drives that made powers of 2 easier or more natural and the manufacturers kept advertising in base 10, you would expect to see hard drives with 192 GB, 256 GB, or 512 GB.
This confusion did not exist until the likes of Seagate chose to create it for their own benefit.
What about the network guys. They use powers of 10. And floppy disks used 1 MB = 1000 * 1024 bytes.
Not to mention that developing for PC's eliminates the licensing/lockout/content restriction issues inherent in all major consoles.
It doesn't eliminate it... look at Steam for example. Okay, you might be able to find a cracked version of whatever piece of software you want to run and play it, while doing the same on your console requires a HW mod. But the PC arena isn't exactly rosy.
That said, I agree that PC gaming isn't going away. There are a lot of games that simply won't work on a console unless you can hook up essentially a keyboard and mouse. Let's see someone get 300 APM in Starcraft with a PS3 controller. PC gaming will become less and less of a player, but there will still be big-name titles.
It doesn't 'just work' in windows in anything except the end user point of view.
"Anything except the end user point of view"? What other point of view is there?
Okay, you can say that MS pays royalties that you pay when you buy Windows, fine. This is true. And I'm not sure what the post a couple up was talking about with the lack of MP3 playing out of the box. But the end users' point of view is the only thing that matters.
Especially when, god forbid, your driver isn't Digitally Signed and Certified by Microsoft, at which point Vista just refuses to install it.
(1) MS isn't directly involved with driver signing; you get keys from Verisign (2) "Only" the x64 editions of Vista refuse to load drivers that aren't signed
But I don't think we have the same things in mind when we say failure.
No, I mean that the machines are more prone to breaking down than marking a ballot with a marker. (Though probably less so than some punch machines.)
To me, failure is not in failing to elect the most popular candidate (by whatever measure of popularity). To me, failure is when the elections process puts someone in office who is more interesting in using the system to his or her own purposes than in being a public servant. That means that the system should discourage gaming at any level. (Yes, the current US system is a dismal failure.)
If this is what you mean by failure, I would argue that virtually all of the problem is outside of the actual vote counting (Florida in 2000 being the exception), so using an electronic voting machine can't really improve or hurt the problems here.
Have you checked wikipedia on the subject? Or are you only thinking of the popular vote in the presidential elections?
I'm thinking of the presidential elections, elections for the US Senate and House, and elections for the state Governor and congress in the places I've lived and voted in.
However, Wikipedia confirms me. It states that there have been 33 instant-runoff elections in the US. There are a number of other places that have accepted but not implemented it, but still only a baker's dozen cities.
What do you mean by so far apart that the outcome can't change? 10% difference? 30%? Change by means of what?
What I was thinking was that at the very least, if someone got more votes than everyone else combined then there was no need to run through the rest of the runoff process. In retrospect, I realize that what I was thinking makes no sense and was stupid.:-) The election would still be essentially an instant runoff election, just without the need to run through every step. But since you wouldn't know that was the case until after votes were counted...
You aren't suggesting that voiding a ballot is presently not possible, are you?
If a voiding machine doesn't allow a voter to void his or her ballot, it isn't really a voting machine.
The question is at what point is the ballot voidable. For instance, you can currently void paper ballots, of course, up to a point. Once you stick it in the box though, this is no longer possible.
So the relevance to the current discussion is in what order does "print the ballot" and "give a final confirmation" come in the final machine. There would need to be a mechanism that would allow the printing to occur before the final confirmation is all I'm saying.
(In other words, treat the paper printout as a vital part of the verification rather than just a receipt that is printed after-the-fact.)
Openness is completely opaque to someone who doesn't even know where the door is. Technology is a wall.
I couldn't tell you how an optical counter works, except guessing and in fairly general terms. I can think of a few ways in which it might work.
But I would much rather have an optical counter than, say, a bunch of people putting ballots into stacks. (Which I could easily understand.)
But if it has to be done, it has to be GPL, or maybe Apache license, or maybe some new government license that requires it to be available to all to both examine and implement, but only if any code used in elections is under the same license.
I don't think that would be necessary. Source would need to be examinable, but I don't think that it's a necessary characteristic that the government could use the code in a third party machine, or another company could use the code, etc.
Being able to examine and test it is important of course.
I don't have to check the optical scanner for two reasons.
I don't buy your first reason (since the machine is st
Failure in a voting system is acceptable? That's why you should have a backup voting method. I think the benefits of e-voting are enough that it is worth it.
Voting is a statistical process. If the vote is so close that a 1% error could effect the outcome, you can't really say that the voters have chosen one or the other. That's the reason some places have runoff elections when no candidate is a clear winner, and runoff elections are the better solution here. I agree. Unfortunately, I'm pretty sure almost nowhere in the US does actually have runoff elections.
I would *love* to see an instant runoff system, or other modified voting system, in place throughout the country for all elections. (Not just when it's close, unless it's so far apart that the outcome can't change.) However, I'm not holding my breath for this to happen.
Besides, what happens if it reaches two candidates and there is still a virtual tie? Declare both of them winners and have two presidents? If you only have one seat, you have to have a winner. So I think e-voting would be good even with instant runoff.
So the city runs temporary cables between the county offices and all the polling places just before every election?
You really think that man-in-the-middle is just a theoretical problem? 1) You don't need temporary cables, the standard phone system would work fine. 2) As I said, the "call-in" count would be preliminary anyway; you would audit the systems later anyway. 3) Network people in CS figured out ways of preventing man-in-the middle attacks long ago (see SSL), and intelligence people long before that.
What does calling in have to do with anything here? If you're willing to trust the net, there is no reason to even bother with voice. The 'net would work too. What is important to me is that there is a separation between the Internet and the voting machines, and machines that hold the counts should never be hooked up to the Internet. If you wanted to have another computer that the precinct supervisor enters the tallies into that is online, that's okay, but the voting machines should be physically isolated. (Or more likely there is some sort of card that is removed from each of the voting machines (with a mechanical interlock to prevent reinsertion) and inserted into the reporting computer.)
But what are we going to do when a voter claims that the printer messed up? (Okay, seems like not such a big deal, but it will happen.) Have a way for the user to trash the ballot before the final vote is counted. (I've seen proposals of machines where this is possible.) This is probably a desirable property anyway.
We agree on the conclusion. Maybe I shouldn't have spent the time arguing. I don't think we do. Despite that ending to my post, I am in favor of e-voting, provided it's done well.
We don't want to teach voters to trust what they can't see. One of the criteria that I use for "done well" is openness. It doesn't have to be GPL'd or anything like that, but the source to the machines should be open and publicly visible. Same with the hardware designs.
The end result -- you in the booth with the machine -- doesn't give you complete trust that the source you looked at is what is running, but the current system also has parts that are hidden. I wonder if you could get schematics for one of those optical scan machines. I kinda doubt it.
It's more expensive, more prone to failure, and doesn't actual provide, better, faster, or more verifiable results. Really? I agree that it's more expensive and probably more prone to failure, but I would argue that a system should provide better, faster, AND more verifiable results.
Better: If you have a "voter marks a ballot, machine counts ballot" system, that will have recognition errors. These can be upwards of 99%, but there are important elections where the margin is smaller than that. A computer voting system should have NO error. The computer won't occasionally add 257 + 1 and get 258. (Bizarre quantum effects and energetic particles hitting the RAM notwithstanding; and you could always have it do every calculation twice if you really want to worry about those.) There are still other sources of inaccuracy and fraud in election, but why not remove one part?
Faster: It should be virtually instant. Even assuming that the machines aren't connected to an outside network (which is how it should be), precincts should be able to report almost instant vote totals. For instance, at election close, someone at each precinct calls the statewide election office and reports the total for each machine (perhaps in encrypted form). Mutual authentication ensures that the person calling is the designated representative. I can imagine several other schemes where perfectly accurate (assuming subsequent audits are clean) statewide results can be available within 5 or 10 minutes of the close of elections. None of this waiting several hours for Cleveland to count their ballots to even get the first number.
Verifiable: A paper trail provides essentially as much verification as any other system. Because it would be printed by the computer, quality control could ensure that the paper ballots are clear in their intention and all valid. It would be impossible to create a paper ballot that had two votes for the same office, and squabbles about voter intention should all but disappear.
I think a much better argument would be that the "better" result is a tiny part of voter inequity and isn't worth the extra money, and faster isn't really a worthwhile goal.
Algorithms are ranked on their resource usage. Not always. Approximation algorithms are often ranked on their accuracy. Online algorithms are often ranked on something called the competitive ratio. Randomized algorithms are usually ranked on their resource uses, but all three of these needn't be optimal (in the context of an optimization problem) -- or produce correct results (in the context of a decision problem).
Algorithms must have the same correct results by definition. [citation needed]
Lets not be overly pedantic: a heuristic is a type of algorithm, in casual speech.
"In casual speech"? That's just wrong... a heuristic is a type of algorithm, period. (Assuming it meets the other requirements of being an algorithm, such as termination.) That it doesn't produce an optimal result doesn't enter into it. [In this post I say "doesn't produce" as a shorthand for "isn't guaranteed to produce".]
CS theorists talk about randomized algorithms. They don't produce an optimal result. CS theorists talk about online algorithms. They don't produce an optimal result. CS theorists talk about approximation algorithms. They don't produce an optimal result.
Producing an optimal result isn't a requirement of being an algorithm. Heuristics are just algorithms that tend to produce useful results most of the time. In fact, Wikipedia page for the CS notion of a heuristic is called "heuristic algorithm."
For people who have more electronics knowledge than I have:
Make a circuit that beeps every 30 seconds or so. Add a photoresistor that turns on and off the beeping, so it beeps when it's dark. Put in victim's bedroom.
Laugh at the though that when they go to bed, it will start beeping, frequently and quietly enough to be annoying, but infrequently enough that it's hard to find. But when they turn the lights back on... the beeping stops!
You pierce the side of a regular lightbulb and it'll maybe shatter and you might get some hot glass on you. Worst case you manage to brand yourself with the tungsten wire.
Well, worst case is the broken glass slices your wrist or something like that.
Large enough that EPA standards imply you need the famous $2,000 clean up crew.
Even the EPA doesn't say that.
Anyone care to suggest what [I assume is] very high pressure plasma, at 6,000K (~10,000F) will do when there's a small hole between it and a much lower pressure room?
Probably not much. There is such a tiny amount the actual amount of heat energy contained would be small. Perhaps lower than what is in your incandescent.
This is true, but at least some people feel that halogens are rather more of a fire risk than incandescent. I don't know if the data supports this, but I do know that many colleges ban halogen lights in dorm rooms for that reason.
Since you talk about Git, I have a question for you. I'm in grad school right now, and use Subversion for my version control. One feature it would be nice to have is I "checkout" a local copy of the repository, make changes when I'm disconnected, then import that whole history into the main repository. (E.g. if I make 3 commits while disconnected, I want to see 3 commits after the merge.)
Most of the advantages I've seen of Git have talked about its branching abilities and such, and I'm not sure I really care about that. But my sense is that Git would give me the above, and that I do care about, and it very well may be enough to get me to switch to Git.
Is my impression correct? Most of what I've seen on Git has either been to long for me to go through or too short to explain in enough detail what is going on.
The nice thing about having a RAM disk is that you may be able to predict the future better than the OS can. For instance, the buffer cache only comes into play when you first read a file. That's all well and good if you are doing repeated access or prefetching or such, but it will do diddly squat for the file that you are/about/ to open if you haven't loaded it since boot.
You're right that it's not a clear win, and in fact there are plenty of times when it'll be a loss because you have more stuff than you have memory, and stuff that you aren't using that's in your ramdisk will take up space that could be used by the buffer cache for stuff that you are. But, if there is a time when you can keep all of your data that isn't a video or music file in RAM, that would be better than having a buffer cache. (Video and music are both big space hogs and I think it's unreasonable to expect that they will reside in RAM, both are read sequentially so caching what has been read isn't worth it, and both have predictable read patterns that are extremely amenable to prefetching. If your hard drive can provide the bandwidth, memory doesn't need to do much. So you could have a "media" partition that just has a small buffer cache bit for prefetching, and your ramdisk can have your data, programs, etc. on it.)
The machine gives a per-person result and a total number who voted result. For the Republican primaries, the total of the former is greater than the latter.
I suppose I could have said "He used the "look at the vote totals the machine printed and add" method."
Slashdot Mirror
Yeah, I saw those, but realistically if the TSA has the masters, what are the chances other folks do as well?
Probably pretty high. But as you point out with regards to just stealing the suitcase, you're trying to protect against a relatively limited form of crime of opportunism.
(and WTH is with that "premier multi-protocol instant messaging client" remark? Nobody uses that on Windows and Mac OS X).
I do. I think it's the best of a bad lot.
Without getting into value judgments about the situation, you can get locks which you can safely put on your luggage. The TSA has master keys that will open them so they can still search, but it will at least prevent average Joe from getting in. They aren't expensive either.
A real megabyte is divisible by an order of magnitude in the relevant
counting system.
Why is it always 2 that's the relevant counting system?
For instance, take the hard drive example. Why are we seeing drives with 200 billion bytes, 250 billion bytes, 500 billion bytes, etc. if 2 is the relevant counting system? If there was something about hard drives that made powers of 2 easier or more natural and the manufacturers kept advertising in base 10, you would expect to see hard drives with 192 GB, 256 GB, or 512 GB.
This confusion did not exist until the likes of Seagate chose to create
it for their own benefit.
What about the network guys. They use powers of 10. And floppy disks used 1 MB = 1000 * 1024 bytes.
Will Windows play DVD's as a fresh install?
Yes, it will. (The Home and Ultimate versions anyway; not Business.) Who's strawmanning now?
Not to mention that developing for PC's eliminates the licensing/lockout/content restriction issues inherent in all major consoles.
It doesn't eliminate it... look at Steam for example. Okay, you might be able to find a cracked version of whatever piece of software you want to run and play it, while doing the same on your console requires a HW mod. But the PC arena isn't exactly rosy.
That said, I agree that PC gaming isn't going away. There are a lot of games that simply won't work on a console unless you can hook up essentially a keyboard and mouse. Let's see someone get 300 APM in Starcraft with a PS3 controller. PC gaming will become less and less of a player, but there will still be big-name titles.
Install Windows -- it won't play DVDs out of the box.
Vista does, at least some flavors.
"Anything except the end user point of view"? What other point of view is there?
Okay, you can say that MS pays royalties that you pay when you buy Windows, fine. This is true. And I'm not sure what the post a couple up was talking about with the lack of MP3 playing out of the box. But the end users' point of view is the only thing that matters.
Especially when, god forbid, your driver isn't Digitally Signed and Certified by Microsoft, at which point Vista just refuses to install it.
(1) MS isn't directly involved with driver signing; you get keys from Verisign
(2) "Only" the x64 editions of Vista refuse to load drivers that aren't signed
It worked for Windows, which just eliminates the previous MBR without asking any questions at all.
Only on initial install and not, as your parent claims Ubuntu will do, at updates.
Backup method?
:-) The election would still be essentially an instant runoff election, just without the need to run through every step. But since you wouldn't know that was the case until after votes were counted...
But I don't think we have the same things in mind when we say failure.
No, I mean that the machines are more prone to breaking down than marking a ballot with a marker. (Though probably less so than some punch machines.)
To me, failure is not in failing to elect the most popular candidate (by whatever measure of popularity). To me, failure is when the elections process puts someone in office who is more interesting in using the system to his or her own purposes than in being a public servant. That means that the system should discourage gaming at any level. (Yes, the current US system is a dismal failure.)
If this is what you mean by failure, I would argue that virtually all of the problem is outside of the actual vote counting (Florida in 2000 being the exception), so using an electronic voting machine can't really improve or hurt the problems here.
Have you checked wikipedia on the subject? Or are you only thinking of the popular vote in the presidential elections?
I'm thinking of the presidential elections, elections for the US Senate and House, and elections for the state Governor and congress in the places I've lived and voted in.
However, Wikipedia confirms me. It states that there have been 33 instant-runoff elections in the US. There are a number of other places that have accepted but not implemented it, but still only a baker's dozen cities.
What do you mean by so far apart that the outcome can't change? 10% difference? 30%? Change by means of what?
What I was thinking was that at the very least, if someone got more votes than everyone else combined then there was no need to run through the rest of the runoff process. In retrospect, I realize that what I was thinking makes no sense and was stupid.
You aren't suggesting that voiding a ballot is presently not possible, are you?
If a voiding machine doesn't allow a voter to void his or her ballot, it isn't really a voting machine.
The question is at what point is the ballot voidable. For instance, you can currently void paper ballots, of course, up to a point. Once you stick it in the box though, this is no longer possible.
So the relevance to the current discussion is in what order does "print the ballot" and "give a final confirmation" come in the final machine. There would need to be a mechanism that would allow the printing to occur before the final confirmation is all I'm saying.
(In other words, treat the paper printout as a vital part of the verification rather than just a receipt that is printed after-the-fact.)
Openness is completely opaque to someone who doesn't even know where the door is. Technology is a wall.
I couldn't tell you how an optical counter works, except guessing and in fairly general terms. I can think of a few ways in which it might work.
But I would much rather have an optical counter than, say, a bunch of people putting ballots into stacks. (Which I could easily understand.)
But if it has to be done, it has to be GPL, or maybe Apache license, or maybe some new government license that requires it to be available to all to both examine and implement, but only if any code used in elections is under the same license.
I don't think that would be necessary. Source would need to be examinable, but I don't think that it's a necessary characteristic that the government could use the code in a third party machine, or another company could use the code, etc.
Being able to examine and test it is important of course.
I don't have to check the optical scanner for two reasons.
I don't buy your first reason (since the machine is st
Failure in a voting system is acceptable?
That's why you should have a backup voting method. I think the benefits of e-voting are enough that it is worth it.
Voting is a statistical process. If the vote is so close that a 1% error could effect the outcome, you can't really say that the voters have chosen one or the other. That's the reason some places have runoff elections when no candidate is a clear winner, and runoff elections are the better solution here.
I agree. Unfortunately, I'm pretty sure almost nowhere in the US does actually have runoff elections.
I would *love* to see an instant runoff system, or other modified voting system, in place throughout the country for all elections. (Not just when it's close, unless it's so far apart that the outcome can't change.) However, I'm not holding my breath for this to happen.
Besides, what happens if it reaches two candidates and there is still a virtual tie? Declare both of them winners and have two presidents? If you only have one seat, you have to have a winner. So I think e-voting would be good even with instant runoff.
So the city runs temporary cables between the county offices and all the polling places just before every election?
You really think that man-in-the-middle is just a theoretical problem?
1) You don't need temporary cables, the standard phone system would work fine.
2) As I said, the "call-in" count would be preliminary anyway; you would audit the systems later anyway.
3) Network people in CS figured out ways of preventing man-in-the middle attacks long ago (see SSL), and intelligence people long before that.
What does calling in have to do with anything here? If you're willing to trust the net, there is no reason to even bother with voice.
The 'net would work too. What is important to me is that there is a separation between the Internet and the voting machines, and machines that hold the counts should never be hooked up to the Internet. If you wanted to have another computer that the precinct supervisor enters the tallies into that is online, that's okay, but the voting machines should be physically isolated. (Or more likely there is some sort of card that is removed from each of the voting machines (with a mechanical interlock to prevent reinsertion) and inserted into the reporting computer.)
But what are we going to do when a voter claims that the printer messed up? (Okay, seems like not such a big deal, but it will happen.)
Have a way for the user to trash the ballot before the final vote is counted. (I've seen proposals of machines where this is possible.) This is probably a desirable property anyway.
We agree on the conclusion. Maybe I shouldn't have spent the time arguing.
I don't think we do. Despite that ending to my post, I am in favor of e-voting, provided it's done well.
We don't want to teach voters to trust what they can't see.
One of the criteria that I use for "done well" is openness. It doesn't have to be GPL'd or anything like that, but the source to the machines should be open and publicly visible. Same with the hardware designs.
The end result -- you in the booth with the machine -- doesn't give you complete trust that the source you looked at is what is running, but the current system also has parts that are hidden. I wonder if you could get schematics for one of those optical scan machines. I kinda doubt it.
It's more expensive, more prone to failure, and doesn't actual provide, better, faster, or more verifiable results.
Really? I agree that it's more expensive and probably more prone to failure, but I would argue that a system should provide better, faster, AND more verifiable results.
Better: If you have a "voter marks a ballot, machine counts ballot" system, that will have recognition errors. These can be upwards of 99%, but there are important elections where the margin is smaller than that. A computer voting system should have NO error. The computer won't occasionally add 257 + 1 and get 258. (Bizarre quantum effects and energetic particles hitting the RAM notwithstanding; and you could always have it do every calculation twice if you really want to worry about those.) There are still other sources of inaccuracy and fraud in election, but why not remove one part?
Faster: It should be virtually instant. Even assuming that the machines aren't connected to an outside network (which is how it should be), precincts should be able to report almost instant vote totals. For instance, at election close, someone at each precinct calls the statewide election office and reports the total for each machine (perhaps in encrypted form). Mutual authentication ensures that the person calling is the designated representative. I can imagine several other schemes where perfectly accurate (assuming subsequent audits are clean) statewide results can be available within 5 or 10 minutes of the close of elections. None of this waiting several hours for Cleveland to count their ballots to even get the first number.
Verifiable: A paper trail provides essentially as much verification as any other system. Because it would be printed by the computer, quality control could ensure that the paper ballots are clear in their intention and all valid. It would be impossible to create a paper ballot that had two votes for the same office, and squabbles about voter intention should all but disappear.
I think a much better argument would be that the "better" result is a tiny part of voter inequity and isn't worth the extra money, and faster isn't really a worthwhile goal.
Algorithms are ranked on their resource usage.
Not always. Approximation algorithms are often ranked on their accuracy. Online algorithms are often ranked on something called the competitive ratio. Randomized algorithms are usually ranked on their resource uses, but all three of these needn't be optimal (in the context of an optimization problem) -- or produce correct results (in the context of a decision problem).
Algorithms must have the same correct results by definition.
[citation needed]
Lets not be overly pedantic: a heuristic is a type of algorithm, in casual speech.
"In casual speech"? That's just wrong... a heuristic is a type of algorithm, period. (Assuming it meets the other requirements of being an algorithm, such as termination.) That it doesn't produce an optimal result doesn't enter into it. [In this post I say "doesn't produce" as a shorthand for "isn't guaranteed to produce".]
CS theorists talk about randomized algorithms. They don't produce an optimal result. CS theorists talk about online algorithms. They don't produce an optimal result. CS theorists talk about approximation algorithms. They don't produce an optimal result.
Producing an optimal result isn't a requirement of being an algorithm. Heuristics are just algorithms that tend to produce useful results most of the time. In fact, Wikipedia page for the CS notion of a heuristic is called "heuristic algorithm."
One would hope that the thing that calculates the heuristic is an algorithm. See wikipedia.
The links in the article look legit...
Is this the day of actual stories that look like April fools stories but aren't, or am I still too dense?
It's not exactly the same thing because it's not light-sensitive.
For people who have more electronics knowledge than I have:
Make a circuit that beeps every 30 seconds or so. Add a photoresistor that turns on and off the beeping, so it beeps when it's dark. Put in victim's bedroom.
Laugh at the though that when they go to bed, it will start beeping, frequently and quietly enough to be annoying, but infrequently enough that it's hard to find. But when they turn the lights back on... the beeping stops!
You pierce the side of a regular lightbulb and it'll maybe shatter and you might get some hot glass on you. Worst case you manage to brand yourself with the tungsten wire.
Well, worst case is the broken glass slices your wrist or something like that.
Large enough that EPA standards imply you need the famous $2,000 clean up crew.
Even the EPA doesn't say that.
Anyone care to suggest what [I assume is] very high pressure plasma, at 6,000K (~10,000F) will do when there's a small hole between it and a much lower pressure room?
Probably not much. There is such a tiny amount the actual amount of heat energy contained would be small. Perhaps lower than what is in your incandescent.
This is true, but at least some people feel that halogens are rather more of a fire risk than incandescent. I don't know if the data supports this, but I do know that many colleges ban halogen lights in dorm rooms for that reason.
We need more accidents, and less people.
Fewer. "Fewer people."
Since you talk about Git, I have a question for you. I'm in grad school right now, and use Subversion for my version control. One feature it would be nice to have is I "checkout" a local copy of the repository, make changes when I'm disconnected, then import that whole history into the main repository. (E.g. if I make 3 commits while disconnected, I want to see 3 commits after the merge.)
Most of the advantages I've seen of Git have talked about its branching abilities and such, and I'm not sure I really care about that. But my sense is that Git would give me the above, and that I do care about, and it very well may be enough to get me to switch to Git.
Is my impression correct? Most of what I've seen on Git has either been to long for me to go through or too short to explain in enough detail what is going on.
The nice thing about having a RAM disk is that you may be able to predict the future better than the OS can. For instance, the buffer cache only comes into play when you first read a file. That's all well and good if you are doing repeated access or prefetching or such, but it will do diddly squat for the file that you are /about/ to open if you haven't loaded it since boot.
You're right that it's not a clear win, and in fact there are plenty of times when it'll be a loss because you have more stuff than you have memory, and stuff that you aren't using that's in your ramdisk will take up space that could be used by the buffer cache for stuff that you are. But, if there is a time when you can keep all of your data that isn't a video or music file in RAM, that would be better than having a buffer cache. (Video and music are both big space hogs and I think it's unreasonable to expect that they will reside in RAM, both are read sequentially so caching what has been read isn't worth it, and both have predictable read patterns that are extremely amenable to prefetching. If your hard drive can provide the bandwidth, memory doesn't need to do much. So you could have a "media" partition that just has a small buffer cache bit for prefetching, and your ramdisk can have your data, programs, etc. on it.)
The machine gives a per-person result and a total number who voted result. For the Republican primaries, the total of the former is greater than the latter.
I suppose I could have said "He used the "look at the vote totals the machine printed and add" method."