The only thing that could be misleading in that sense are the 7% and 10% numbers I mentioned; but those are both relative to the group's prior numbers -- so the bottom 50% of earners had a significantly higher relative increase in AGI than the top 50% (or 1%) did. The shifting share of income tax paid, especially given the shift in AGI in the opposite direction, is not susceptible to that misunderstanding.
You are also moving goalposts. You posted about tax burdens, not total ROI from the government or other forms of economic health. I have better things to do than chase down numbers for you and explain why per capita income tax or median wage are at least as susceptible to manipulation or misunderstanding as fraction of total AGI and income tax paid (they certainly have less to do with tax burden than do the numbers I cited). Since you present no support for your tax burden argument, I interpret your post as an admission of error.
Both Republicans and Democrats use vi -- each wants behavioral modes where different rules apply at different times and only the person in charge knows which rules apply at any given time. The Republicans happen to be farther along making those modes "sticky" based on the user's party affiliation. Libertarians use emacs: they hold positions that make perfect sense to them, but hardly anyone else can figure out exactly why the libertarian believes that, and the details vary significantly from one libertarian to the next. Communists use TECO because it facilitates central planning of operations; they just ignore the system's vulnerability to bad input. Most socialists use ed because ED IS THE STANDARD TEXT EDITOR and this obviously means everyone must use it.
Between 2000 and 2004 (the last year for which data is available from the IRS), the fractions of US income tax paid by the top 5% and 10% of earners grew. The fractions of income tax paid by the bottom 50% and bottom 75% of earners shrank. Over the same time period, pre-tax income for the top 1% grew by 7%, and pre-tax income for the bottom 50% grew by 10%. The Tax Foundation's report has more details, but I see no evidence there of an increased tax burden on the middle class.
Bush 43 cannot claim much credit for this -- moving the tax burden up the scale is a trend that goes back as far as The Tax Foundation's charts show -- but he may be able to claim credit for increasing the bottom 50% earners' share of total Adjusted Gross Income, which had been steadily decreasing between 1980 and 2000. Surely doing that, while still reducing their tax burden, counts for something. (Judging from those AGI numbers, at least, overall income inequality has been DECREASING since 2000. Go figure.)
Any competent pollster structures their study so that they have a broad sample set. However, not all the results are weighted equally: if they end up getting answers from ten illegal immigrant transsexuals and ten soccer fathers, the soccer father set will be weighted more heavily because they constitute a larger fraction of the relevant (in this case, voting) population. Pollsters' models of how to weight these groups are based largely on previous turnout numbers, although they could be (and probably are) adjusted if some demographic groups show a large change in likelihood to vote.
Thus, when there is a sudden change in voter demographics, the extrapolations from poll sample set to voting public are less accurate than they otherwise would be.
I am all for independently auditable election results, and for paper trails as an easy way to get there. I am against those who think we can get by without easy, robust recounts. Go back and read the comment to which I replied. It was about party affiliation, so I responded in counterpoint.
(I also doubt that the Democrats figured out better ways to hack the system -- the problems in the primaries were mostly intra-party, and related more to whether the systems accepted votes than to how accurately they counted. Instead, I suspect that the Democrats wanted to keep the Diebold machines simply because a Republican governor wanted to shelve them this year.)
Even using Republican math (I prefer small-l libertarian math), the US has five presidential elections in 20 years. Being wrong one out of those five times is (drum roll...) a 20% error rate. If you mean to say that Democratic math reaches a different result, perhaps you should check that rather than laugh about Republican math.
Even if you can't read what the Tribune or I actually wrote, two seconds of thought should indicate that it would be rather hard to use the same political correspondent for 20 presidential elections in a row.
The Chicago Tribune itself says that the numbers were based on a combination of early election results and a Washington correspondent who was wrong just once in the previous 20 years (20% error?). Private polls did show a significant Dewey lead before the election. Between the two, it was surprising Truman won the election, but I do not think even die-hard Republicans would claim the election was stolen.
And just reflect: That headline was based on early election results, which lack a number of biases present in exit or telephone polling. (The major hard-to-address biases early results have are limited sample size and geographical bias.)
American exit polls have never been that accurate. Their margins of error have come down somewhat, but statistically speaking they have never been "accurate to a very slim margin".
Maryland Governor Bob Ehrlich, a Republican, ran into stiff opposition after (Diebold?) voting machines caused major problems in the state's primary elections this year. Ehrlich wanted to switch to paper-based methods that were known to be reliable. The opposition was NOT from his own party, but from the state's Democratic majority and career bureaucrats.
No; quite the opposite. Open source development that accepts third-party contributions makes it extraordinarily difficult to meet ISO 9000 process requirements. The standard is not just about documenting the process, but making sure you follow it consistently. When you have workers whose compliance is not audited, you cannot pass the audit. This means that continuously accepting third-party contributions has to be designed into the documented process -- something alien to most ISO 9000 practitioners -- and, in particular, quality controls and testing have to be designed to accomodate that. In practice, you would also want some fraction (at least 5-10%; higher if your process isn't tuned exactly to the open source model you use) of the time spent on the project to be internal process auditing by people who know both ISO 9000 and the documented process.
By posting your comment, you have accepted the Slashdot Comment Contract, which gives CowboyNeal the right to camp out indefinitely in your front yard, kitchen, bedroom or other equivalent area, at his option. This Contract may be revised at any time with notice provided by CmdrTaco scribbling the revisions on his cocktail napkin. You have the opportunity to review the changes by reading them before condensation or spill renders them illegible, or before Hemos eats the napkin on a dare.
Seriously, though, no court will enforce a contract that is changed unilaterally. A court may set the bar absurdly low on what constitutes acceptance by the other party -- for example, use of an online service after the usage contract changed without the user knowing it changed -- but there must be some specific action by each party to indicate acceptance of the contract.
What planet are you from where you get a factor of 100 difference by adding 100 rather than multiplying? Try using standard arithmetic.
If you do, you will find that the crossover is around n=647 for a relative speed of 100x (647*647 = 418609 vs 100*647*log(647)=418760), or around n=282 for a relative speed of 50x.
However, your broader point stands, since 50-100x ratios are what you might expect for a poorly interpreted language, and JITs or other optimizing compilers perform at the same order of magnitude as native code (and usually within a factor of two).
The United States alone can account for over 100 million Windows XP licenses (that would need just one license per three people, which is low: 70% of Americans are online from home, 35% from work). According to IDC, 208 million PCs were shipped in 2005, and they project an 10% increase in 2006. Windows XP was released in 2001; I stand by my numbers.
"100% of the people who get these are not innocent." Does that 100% include Mr. Thurrott, or did you mean that some (unknown) fraction of the people who get this are not innocent?
Microsoft is showing all the signs of botching the sensitivity tradeoff; even if the false alarm rate is one in 10,000, that will affect tens or hundreds of thousands of innocent users. False alarms are also likely to occur in clusters where the direct impact (e.g. one company that has enough licenses but some procedural problem has all its machines flagged) or indirect impact (e.g. IT managers get together and trade stories over beer) make the problem visible enough to deter people from Microsoft products. Users can rationally blame Microsoft directly for this more than for third-party malware or for software bugs: Microsoft claims to be closing security holes and fixing bugs, but cannot claim that WGA is there by accident.
I for one welcomed my not-so-new penguin overlords and have not looked back.
Answering for missing top-level zones was such a resounding success when Network Solutions tried it that everyone else decided they had to get in on the action! Don't blame these cutting-edge Internet Innovators And Entrepreneurs just because they beat you to the punch (bowl)!
You fail at English as badly as the article's author. What do you honestly think "one in the same" means? Here, have an "and" -- if you have overdosed on prepositions, conjunctions are an entirely different grammar sensation!
Even more than "opposites attract", like charges repel -- and there are like charges closer than opposite charges. A charge tends to distribute fairly evenly across the surface it is on, although shape has some effect.
If new releases and termination of security updates were predictable, having a year-long window would be more than enough. The problem is that the release schedule is hugely unpredictable under Debian (remember the last-minute delay of sarge?), so the effective planning window is significantly shorter than a year. Since it's common for major efforts to be planned at least a year in advance, this often means that the IT department must support a mixed environment because there is no one time that every system can be upgraded without impacting someone's deadline.
As the other commenter pointed out, there is a lower barrier to entry for the well-supported version.
Beyond all the reasons that running a mixture of stable and testing is in itself risky, there is also no predictable long-term support for Debian stable releases: you have to upgrade after one or two release cycles (unpredictably[1]) but so far that has been more often than every three years. Debian puts a lot of effort into making upgrades work well, but any dist upgrade has risks, and enterprise customers want to minimize the number of times they have to upgrade everyone. Perhaps more importantly, they want the timing to be predictable, so they can plan and budget to reduce the indirect costs of upgrades (such as lost productivity during the upgrade or when fixing problems aftwards).
[1]- Debian's woody release page says woody will stop getting security updates once etch starts getting its security updates; potato stopped getting security updates two years before sarge was released; the pages for hamm and slink do not say when their security updates stopped.
Today most of the software that is used to fly planes (both fighter jets and passenger) is based on a microkernel architecture.
Sort of -- in the same way that OS X and similar systems are "based on a microkernel architecture". ARINC-653, which drives that software architecture, specifies a partitioning kernel that separates safety-critical tasks from non-safety-critical tasks (or other safety-critical tasks). Most DO-178B compliant software vendors then run a monolithic kernel in each partition. The partitioning kernel is usually more like an extremely rigorously verified virtualization service than a traditional microkernel.
As you would know if you had the slightest clue about Windows NT or its successors, IFS is Microsoft's proprietary Installable File System API. The only thing it tests is whether a software vendor complies with Microsoft, so there is no point for anyone else to bother attending.
Also, ITSC Plugfest is NOT the one TFA refers to (your Google fu is weak). As you might expect, Tridgell was talking about one relevant to Samba: the CIFS Conference & Plugfest, operated by the Storage Networking Industry Association. Most of the vendors who sponsor it do not use Linux on their storage solutions.
Sure, it's what we would expect them to do. However, it supports the claim that Microsoft does not care about interoperability and wants to illegally leverage its monopoly in some areas to dominate related areas as well. It is a strong clue that Microsoft software fails to interoperate -- if it worked well with others, sending a few engineers for the relevant two-week test session would be a cheap form of PR and would totally defuse complaints about interoperability.
That really begs for someone to ask *why*, if their education programs are so good, those grad students move to US schools instead of choosing domestic graduate programs. Whatever fraction of Indian and Chinese undergraduates attend grad school in the US, a much lower fraction of (I think effectively zero) US undergraduates attend grad school in India or China.
The US has a shockingly poor primary and secondary education system. It has a rather weak undergraduate system. But those traits are not what this kind of contest measures.
This particular contest tests preparation and aptitude for this kind of contest. As an ancestor post mentioned, US universities do not emphasize this much. Instead of high-intensity coding sprints, they emphasize longer term projects and, to some extent, entrepreneurial opportunities. I submit that the difference in focus is a rather larger factor in the results than the general qualities of the respective education systems.
This was certainly my experience in similar but more local competitions during high school. In French, Latin, programming and mathematics, the schools that did particularly well were the ones with organized and well-supported programs that spent a lot of time drilling their teams in mock competitions. Ignoring polymath individuals, the correlation of excellence in different programs was low.
There was also considerable momentum in the form of the staff and faculty members who coached the teams; they generally had well-adapted and proven ways to train their students. Pressure to live up to a program's history could also be a significant drive.
The hard part -- at least for people without a decade of intensive C background -- is knowing the "appropriate places" for size_t, ssize_t, socklen_t, and all those other fun types that are used only sometimes or in some places or on some platforms. System libraries are getting better about consistency and compatibility, and compilers are getting better about type conversion warnings, but source code that wants to run on different OSes released before 2001 ends up growing a maze of twisty little #ifdefs, all alike (but sometimes all different). If you only care about platforms from 2004 and later, you can get away with a small thicket of fairly straightforward little #ifdefs.
Slashdot Mirror
The only thing that could be misleading in that sense are the 7% and 10% numbers I mentioned; but those are both relative to the group's prior numbers -- so the bottom 50% of earners had a significantly higher relative increase in AGI than the top 50% (or 1%) did. The shifting share of income tax paid, especially given the shift in AGI in the opposite direction, is not susceptible to that misunderstanding.
You are also moving goalposts. You posted about tax burdens, not total ROI from the government or other forms of economic health. I have better things to do than chase down numbers for you and explain why per capita income tax or median wage are at least as susceptible to manipulation or misunderstanding as fraction of total AGI and income tax paid (they certainly have less to do with tax burden than do the numbers I cited). Since you present no support for your tax burden argument, I interpret your post as an admission of error.
Both Republicans and Democrats use vi -- each wants behavioral modes where different rules apply at different times and only the person in charge knows which rules apply at any given time. The Republicans happen to be farther along making those modes "sticky" based on the user's party affiliation. Libertarians use emacs: they hold positions that make perfect sense to them, but hardly anyone else can figure out exactly why the libertarian believes that, and the details vary significantly from one libertarian to the next. Communists use TECO because it facilitates central planning of operations; they just ignore the system's vulnerability to bad input. Most socialists use ed because ED IS THE STANDARD TEXT EDITOR and this obviously means everyone must use it.
Bush 43 cannot claim much credit for this -- moving the tax burden up the scale is a trend that goes back as far as The Tax Foundation's charts show -- but he may be able to claim credit for increasing the bottom 50% earners' share of total Adjusted Gross Income, which had been steadily decreasing between 1980 and 2000. Surely doing that, while still reducing their tax burden, counts for something. (Judging from those AGI numbers, at least, overall income inequality has been DECREASING since 2000. Go figure.)
Any competent pollster structures their study so that they have a broad sample set. However, not all the results are weighted equally: if they end up getting answers from ten illegal immigrant transsexuals and ten soccer fathers, the soccer father set will be weighted more heavily because they constitute a larger fraction of the relevant (in this case, voting) population. Pollsters' models of how to weight these groups are based largely on previous turnout numbers, although they could be (and probably are) adjusted if some demographic groups show a large change in likelihood to vote.
Thus, when there is a sudden change in voter demographics, the extrapolations from poll sample set to voting public are less accurate than they otherwise would be.
I am all for independently auditable election results, and for paper trails as an easy way to get there. I am against those who think we can get by without easy, robust recounts. Go back and read the comment to which I replied. It was about party affiliation, so I responded in counterpoint. (I also doubt that the Democrats figured out better ways to hack the system -- the problems in the primaries were mostly intra-party, and related more to whether the systems accepted votes than to how accurately they counted. Instead, I suspect that the Democrats wanted to keep the Diebold machines simply because a Republican governor wanted to shelve them this year.)
Even using Republican math (I prefer small-l libertarian math), the US has five presidential elections in 20 years. Being wrong one out of those five times is (drum roll...) a 20% error rate. If you mean to say that Democratic math reaches a different result, perhaps you should check that rather than laugh about Republican math.
Even if you can't read what the Tribune or I actually wrote, two seconds of thought should indicate that it would be rather hard to use the same political correspondent for 20 presidential elections in a row.
The Chicago Tribune itself says that the numbers were based on a combination of early election results and a Washington correspondent who was wrong just once in the previous 20 years (20% error?). Private polls did show a significant Dewey lead before the election. Between the two, it was surprising Truman won the election, but I do not think even die-hard Republicans would claim the election was stolen. And just reflect: That headline was based on early election results, which lack a number of biases present in exit or telephone polling. (The major hard-to-address biases early results have are limited sample size and geographical bias.)
DEWEY DEFEATS TRUMAN
American exit polls have never been that accurate. Their margins of error have come down somewhat, but statistically speaking they have never been "accurate to a very slim margin".
Maryland Governor Bob Ehrlich, a Republican, ran into stiff opposition after (Diebold?) voting machines caused major problems in the state's primary elections this year. Ehrlich wanted to switch to paper-based methods that were known to be reliable. The opposition was NOT from his own party, but from the state's Democratic majority and career bureaucrats.
No; quite the opposite. Open source development that accepts third-party contributions makes it extraordinarily difficult to meet ISO 9000 process requirements. The standard is not just about documenting the process, but making sure you follow it consistently. When you have workers whose compliance is not audited, you cannot pass the audit. This means that continuously accepting third-party contributions has to be designed into the documented process -- something alien to most ISO 9000 practitioners -- and, in particular, quality controls and testing have to be designed to accomodate that. In practice, you would also want some fraction (at least 5-10%; higher if your process isn't tuned exactly to the open source model you use) of the time spent on the project to be internal process auditing by people who know both ISO 9000 and the documented process.
By posting your comment, you have accepted the Slashdot Comment Contract, which gives CowboyNeal the right to camp out indefinitely in your front yard, kitchen, bedroom or other equivalent area, at his option. This Contract may be revised at any time with notice provided by CmdrTaco scribbling the revisions on his cocktail napkin. You have the opportunity to review the changes by reading them before condensation or spill renders them illegible, or before Hemos eats the napkin on a dare.
Seriously, though, no court will enforce a contract that is changed unilaterally. A court may set the bar absurdly low on what constitutes acceptance by the other party -- for example, use of an online service after the usage contract changed without the user knowing it changed -- but there must be some specific action by each party to indicate acceptance of the contract.
What planet are you from where you get a factor of 100 difference by adding 100 rather than multiplying? Try using standard arithmetic.
If you do, you will find that the crossover is around n=647 for a relative speed of 100x (647*647 = 418609 vs 100*647*log(647)=418760), or around n=282 for a relative speed of 50x.
However, your broader point stands, since 50-100x ratios are what you might expect for a poorly interpreted language, and JITs or other optimizing compilers perform at the same order of magnitude as native code (and usually within a factor of two).
The United States alone can account for over 100 million Windows XP licenses (that would need just one license per three people, which is low: 70% of Americans are online from home, 35% from work). According to IDC, 208 million PCs were shipped in 2005, and they project an 10% increase in 2006. Windows XP was released in 2001; I stand by my numbers.
"100% of the people who get these are not innocent." Does that 100% include Mr. Thurrott, or did you mean that some (unknown) fraction of the people who get this are not innocent?
Microsoft is showing all the signs of botching the sensitivity tradeoff; even if the false alarm rate is one in 10,000, that will affect tens or hundreds of thousands of innocent users. False alarms are also likely to occur in clusters where the direct impact (e.g. one company that has enough licenses but some procedural problem has all its machines flagged) or indirect impact (e.g. IT managers get together and trade stories over beer) make the problem visible enough to deter people from Microsoft products. Users can rationally blame Microsoft directly for this more than for third-party malware or for software bugs: Microsoft claims to be closing security holes and fixing bugs, but cannot claim that WGA is there by accident.
I for one welcomed my not-so-new penguin overlords and have not looked back.
Answering for missing top-level zones was such a resounding success when Network Solutions tried it that everyone else decided they had to get in on the action! Don't blame these cutting-edge Internet Innovators And Entrepreneurs just because they beat you to the punch (bowl)!
You fail at English as badly as the article's author. What do you honestly think "one in the same" means? Here, have an "and" -- if you have overdosed on prepositions, conjunctions are an entirely different grammar sensation!
Even more than "opposites attract", like charges repel -- and there are like charges closer than opposite charges. A charge tends to distribute fairly evenly across the surface it is on, although shape has some effect.
Your explanation addresses exactly why Japan and Germany are so radicalized these days.
Or do they not teach recent history on your side of the puddle?
If new releases and termination of security updates were predictable, having a year-long window would be more than enough. The problem is that the release schedule is hugely unpredictable under Debian (remember the last-minute delay of sarge?), so the effective planning window is significantly shorter than a year. Since it's common for major efforts to be planned at least a year in advance, this often means that the IT department must support a mixed environment because there is no one time that every system can be upgraded without impacting someone's deadline.
As the other commenter pointed out, there is a lower barrier to entry for the well-supported version.
Beyond all the reasons that running a mixture of stable and testing is in itself risky, there is also no predictable long-term support for Debian stable releases: you have to upgrade after one or two release cycles (unpredictably[1]) but so far that has been more often than every three years. Debian puts a lot of effort into making upgrades work well, but any dist upgrade has risks, and enterprise customers want to minimize the number of times they have to upgrade everyone. Perhaps more importantly, they want the timing to be predictable, so they can plan and budget to reduce the indirect costs of upgrades (such as lost productivity during the upgrade or when fixing problems aftwards).
[1]- Debian's woody release page says woody will stop getting security updates once etch starts getting its security updates; potato stopped getting security updates two years before sarge was released; the pages for hamm and slink do not say when their security updates stopped.
As you would know if you had the slightest clue about Windows NT or its successors, IFS is Microsoft's proprietary Installable File System API. The only thing it tests is whether a software vendor complies with Microsoft, so there is no point for anyone else to bother attending.
Also, ITSC Plugfest is NOT the one TFA refers to (your Google fu is weak). As you might expect, Tridgell was talking about one relevant to Samba: the CIFS Conference & Plugfest, operated by the Storage Networking Industry Association. Most of the vendors who sponsor it do not use Linux on their storage solutions.
Sure, it's what we would expect them to do. However, it supports the claim that Microsoft does not care about interoperability and wants to illegally leverage its monopoly in some areas to dominate related areas as well. It is a strong clue that Microsoft software fails to interoperate -- if it worked well with others, sending a few engineers for the relevant two-week test session would be a cheap form of PR and would totally defuse complaints about interoperability.
That really begs for someone to ask *why*, if their education programs are so good, those grad students move to US schools instead of choosing domestic graduate programs. Whatever fraction of Indian and Chinese undergraduates attend grad school in the US, a much lower fraction of (I think effectively zero) US undergraduates attend grad school in India or China.
The US has a shockingly poor primary and secondary education system. It has a rather weak undergraduate system. But those traits are not what this kind of contest measures.
This particular contest tests preparation and aptitude for this kind of contest. As an ancestor post mentioned, US universities do not emphasize this much. Instead of high-intensity coding sprints, they emphasize longer term projects and, to some extent, entrepreneurial opportunities. I submit that the difference in focus is a rather larger factor in the results than the general qualities of the respective education systems.
This was certainly my experience in similar but more local competitions during high school. In French, Latin, programming and mathematics, the schools that did particularly well were the ones with organized and well-supported programs that spent a lot of time drilling their teams in mock competitions. Ignoring polymath individuals, the correlation of excellence in different programs was low.
There was also considerable momentum in the form of the staff and faculty members who coached the teams; they generally had well-adapted and proven ways to train their students. Pressure to live up to a program's history could also be a significant drive.
The hard part -- at least for people without a decade of intensive C background -- is knowing the "appropriate places" for size_t, ssize_t, socklen_t, and all those other fun types that are used only sometimes or in some places or on some platforms. System libraries are getting better about consistency and compatibility, and compilers are getting better about type conversion warnings, but source code that wants to run on different OSes released before 2001 ends up growing a maze of twisty little #ifdefs, all alike (but sometimes all different). If you only care about platforms from 2004 and later, you can get away with a small thicket of fairly straightforward little #ifdefs.