It is reasonable to assume that non-significant digits will be uniform, given a sufficiently large sample. On the other hand, it is not reasonable to expect that mere second digits will be uniform in data that is as highly biased as poll results.
In other words, I don't expect any particular distribution, but I don't believe that the mere presence of a non-uniform distribution is enough to prove wrong-doing.
Moreover, (as Nate states) over enough data, even the effect of the undecided percentage on the trailing digit should be random.
Except that in this case, the trailing digit is merely the second digit. A bias in the second digit of what is after all highly biased data (you don't have a lot of 98-2 results in polls) is not unlikely, even in samples much larger than what he's using.
Not saying that the company is honest, but Silver's argument is not sufficient to condemn them.
Just to be clear -- you cannot expect second digits in what are two-digit results to be uniform. You can expect fourth digits to be uniform, but that data is not available.
I'm not sure I understand what Silver is claiming about the data.
He shows that the distribution of second digits in the results of Pollster's polls doesn't follow a uniform distribution -- and from that he somehow deduces it's not random.
If you look at the figure in the second article, it looks to my untrained eyes like a gaussian curve with maximum around 8 -- since when are gaussians not random?
Now, I know from experience that open source software is significantly less expensive on a per seat basis, builds local skills and support, and offers flexibility you just can't get from other options.
That is not the main point.
The main point is that by using Free software, OLPC will get millions of children exposed to the idea that computers are tweakable -- that it is okay to look under the hood of your computer. By using proprietary software, OLPC will get millions of children exposed to the notion that computers are frozen, that you're allowed to look but not to touch.
Which is basically why my profs just hit 'clear all' themselves and move on to their next student/victim
I'm fairly positive we're not allowed to do that. We do have the right to ask a student to turn his calculator/mobile phone/mp3 player/whatever off, we are allowed to confiscate it and give it back at the end of the lecture, but we're certainly not allowed to clear the data off a studen't calculator.
Surprising as it may sound, students are human beings, and enjoy many of the same rights as we do.
[The Qur'an] calls for the shedding of the blood of the kafir
And how exactly that that differ from the Bible?
At least, nowhere in the Qur'an will you find the suggestion that it is a good idea to murder the Gentiles when they are recovering from circumcision (Exodus, if memory serves), and nowhere in the Qur'an will you find the notion that you should repudiate your wife because she refused to perform a strip-tease for your drunken friends (Book of Esther).
On the other hand, nowhere is the Qur'an as colourful as the Bible, when it raves about the joys of collective sex (Genesis) or homosexual love between consenting adults (the Gospel according to John). Or promote the drinking of wine (Saint Paul's epistle to Timothy).
All they have to do is block the known Tor entry points
Tor bridges are designed to provide F2F access to the tor network in such a situation.
or set up their own hacked TOR routers.
Tor traffic is encrypted and authenticated end-to-end within the tor network, so unless they manage to get the users to download hacked tor binaries, this shouldn't be an issue.
If it's that much of a concern, why don't the schools have a set of TI calculators that are available for standardized tests? Stagger the testing properly, and you don't actually need one for each student.
And spend two days on tests that could be done in four hours? Thanks, I'll pass.
If the school is concerned with the costs involved, I'm sure they could work a deal with TI to receive the "in-class only" calculators for free or at cost.
TI does indeed provide us with free calculators, if you manage to get through the mess that is TI's management. (Different managers appear to disagree on whether or not they are giving us calculators.)
[some lecturers] make up problems such that you shouldn't need a calculator to solve them
You wouldn't believe how difficult it is to convince undergrads that they do not need a calculator to multiply 1024 by 4.
(Most students want to use their mobile phone as a calculator -- which is obviously forbidden during the test, we don't want them to be sending SMSes asking for help.)
Andy Grove said, "You can't just sit on your a** and give everyone the finger." And later he added, "Hey you kids, get off my damn lawn!";-)
But Mr. Grove is correct - government often makes things stagnate and hold steady, such as when AT&T had a government-protected monopoly over the phone lines and computer modems.
I think there's a confusion between two things. There's government-backed projects, which tend to give excellent results -- the Internet, Bell Labs, NASA, the French railway system, etc.
Then, there's government backed monopolies and subventions to dying industries, which tend to promote obsolete technology, and do so at a very high cost -- the AT&T telephone system, General Motors, the European steel industry, and, some might argue, the current banking system.
There are some restrictions on when macro expanders are allowed to be run which cannot be satisfied without performing a pre-pass. Sorry, I don't recall the details.
I thought that some other Free CL implementations are interpreters, but I might be confused, CLISP does compile to bytecodes, and GCL uses gcc...
CMUCL, CCL (formerly MCL), SBCL, Lispworks and ACL are native compilers. KCL, AKCL and GCL compile to C. CLISP compiles to bytecode. Heck, even Emacs Lisp is compiled (bytecode).
(By the way, have you tried CCL yet? It's a very nice implementation, and since it's only been ported to Linux recently, a lot of people are not aware that it exists.)
I really don't know much about LISP, but I'm having a hard time understanding why "performing a pass over each function" rules out being a "pure interpreter." Maybe we just have different definitions of what that means. Could you please clarify?
Sure. A pure interpreter is an implementation that executes the code as it parses it, or at least as it walks the parse tree. An intepreter with a pre-pass is one that first walks the code, does some transformations or caches some data, and only then executes the code.
Suppose you're implementing a BASIC interpreter. A pure interpreter would walk your BASIC program, and execute statements as it parses them. However, it will surely be more efficient to perform a pre-pass and remember the addresses of every line number, so that every GOTO doesn't need to scan the whole program to find its target.
Actually, Common Lisp cannot be implemented as a pure interpreter -- there are a few features of the language that you cannot implement without performing a pass over each function.
(Scheme, the other dominant dialect of Lisp, can be implemented as a pure interpreter, a pure compiler, or a hybrid design.)
Slashdot Mirror
It is reasonable to assume that non-significant digits will be uniform, given a sufficiently large sample. On the other hand, it is not reasonable to expect that mere second digits will be uniform in data that is as highly biased as poll results.
In other words, I don't expect any particular distribution, but I don't believe that the mere presence of a non-uniform distribution is enough to prove wrong-doing.
Moreover, (as Nate states) over enough data, even the effect of the undecided percentage on the trailing digit should be random.
Except that in this case, the trailing digit is merely the second digit. A bias in the second digit of what is after all highly biased data (you don't have a lot of 98-2 results in polls) is not unlikely, even in samples much larger than what he's using. Not saying that the company is honest, but Silver's argument is not sufficient to condemn them.
Just to be clear -- you cannot expect second digits in what are two-digit results to be uniform. You can expect fourth digits to be uniform, but that data is not available.
I'm not sure I understand what Silver is claiming about the data.
He shows that the distribution of second digits in the results of Pollster's polls doesn't follow a uniform distribution -- and from that he somehow deduces it's not random.
If you look at the figure in the second article, it looks to my untrained eyes like a gaussian curve with maximum around 8 -- since when are gaussians not random?
Most of us keep our machines running all the time.
Yes, we do, and that is wasteful. With faster boot and support for wake-on-lan in routers, we could be making significant energy savings.
I would think a quicker return from suspend or hibernate would be more useful.
Returning from hibernate performs a full hardware boot (including BIOS POST) -- hibernate merely restores the user-space memory from disk.
Now, I know from experience that open source software is significantly less expensive on a per seat basis, builds local skills and support, and offers flexibility you just can't get from other options.
That is not the main point.
The main point is that by using Free software, OLPC will get millions of children exposed to the idea that computers are tweakable -- that it is okay to look under the hood of your computer. By using proprietary software, OLPC will get millions of children exposed to the notion that computers are frozen, that you're allowed to look but not to touch.
The choice is political, not economic.
Anyone care to step up to the plate to defend BIND's security credentials? Anyone? Is this thing on?
Interestingly enough, Bind9 was originally written by Nominium...
Way, way back when, Nominum employees successfully performed a denial of service attack on PowerDNS.
Does anyone know what this refers to?
Fucking Nixon and his brinksmanship, fucking LeMay and trying to start WWIII during the Cuban Missile Crisis, and fucking Reagan...
A pity you didn't mention Kennedy in what is otherwise a perfectly respectable rant.
Which is basically why my profs just hit 'clear all' themselves and move on to their next student/victim
I'm fairly positive we're not allowed to do that. We do have the right to ask a student to turn his calculator/mobile phone/mp3 player/whatever off, we are allowed to confiscate it and give it back at the end of the lecture, but we're certainly not allowed to clear the data off a studen't calculator.
Surprising as it may sound, students are human beings, and enjoy many of the same rights as we do.
what is the answer?
42.
[The Qur'an] calls for the shedding of the blood of the kafir
And how exactly that that differ from the Bible?
At least, nowhere in the Qur'an will you find the suggestion that it is a good idea to murder the Gentiles when they are recovering from circumcision (Exodus, if memory serves), and nowhere in the Qur'an will you find the notion that you should repudiate your wife because she refused to perform a strip-tease for your drunken friends (Book of Esther).
On the other hand, nowhere is the Qur'an as colourful as the Bible, when it raves about the joys of collective sex (Genesis) or homosexual love between consenting adults (the Gospel according to John). Or promote the drinking of wine (Saint Paul's epistle to Timothy).
All they have to do is block the known Tor entry points
Tor bridges are designed to provide F2F access to the tor network in such a situation.
or set up their own hacked TOR routers.
Tor traffic is encrypted and authenticated end-to-end within the tor network, so unless they manage to get the users to download hacked tor binaries, this shouldn't be an issue.
Hang on - it's 2009 and we're still arguing about calculators?
Vi sucks.
If it's that much of a concern, why don't the schools have a set of TI calculators that are available for standardized tests? Stagger the testing properly, and you don't actually need one for each student.
And spend two days on tests that could be done in four hours? Thanks, I'll pass.
If the school is concerned with the costs involved, I'm sure they could work a deal with TI to receive the "in-class only" calculators for free or at cost.
TI does indeed provide us with free calculators, if you manage to get through the mess that is TI's management. (Different managers appear to disagree on whether or not they are giving us calculators.)
[some lecturers] make up problems such that you shouldn't need a calculator to solve them
You wouldn't believe how difficult it is to convince undergrads that they do not need a calculator to multiply 1024 by 4.
(Most students want to use their mobile phone as a calculator -- which is obviously forbidden during the test, we don't want them to be sending SMSes asking for help.)
Worse is better!
how is designing software to do more things badly superior...
Whoosh
Andy Grove said, "You can't just sit on your a** and give everyone the finger." And later he added, "Hey you kids, get off my damn lawn!" ;-)
But Mr. Grove is correct - government often makes things stagnate and hold steady, such as when AT&T had a government-protected monopoly over the phone lines and computer modems.
I think there's a confusion between two things. There's government-backed projects, which tend to give excellent results -- the Internet, Bell Labs, NASA, the French railway system, etc.
Then, there's government backed monopolies and subventions to dying industries, which tend to promote obsolete technology, and do so at a very high cost -- the AT&T telephone system, General Motors, the European steel industry, and, some might argue, the current banking system.
I know thats why it was originally invented, but I don't think the modern internet is emp resistant.
That's an urban legend.
When did that transition happen? Late 70s?
Winter 1982/1983. On 7 December 1982, 130 out of 315 hosts speak TCP/IP (RFC 832). On 22 February 1983, that's 230 out of 320 (RFC 846).
A lot of the other torrent sites relies on the Pirate Bay tracker. What other large, public bittorrent trackers (not just .torrent mirrors) are there?
When a tracker goes down, most BitTorrent implementations will automatically (and invisibly to the user) switch to the DHT.
What would be an example?
There are some restrictions on when macro expanders are allowed to be run which cannot be satisfied without performing a pre-pass. Sorry, I don't recall the details.
I thought that some other Free CL implementations are interpreters, but I might be confused, CLISP does compile to bytecodes, and GCL uses gcc...
CMUCL, CCL (formerly MCL), SBCL, Lispworks and ACL are native compilers. KCL, AKCL and GCL compile to C. CLISP compiles to bytecode. Heck, even Emacs Lisp is compiled (bytecode).
(By the way, have you tried CCL yet? It's a very nice implementation, and since it's only been ported to Linux recently, a lot of people are not aware that it exists.)
I really don't know much about LISP, but I'm having a hard time understanding why "performing a pass over each function" rules out being a "pure interpreter." Maybe we just have different definitions of what that means. Could you please clarify?
Sure. A pure interpreter is an implementation that executes the code as it parses it, or at least as it walks the parse tree. An intepreter with a pre-pass is one that first walks the code, does some transformations or caches some data, and only then executes the code.
Suppose you're implementing a BASIC interpreter. A pure interpreter would walk your BASIC program, and execute statements as it parses them. However, it will surely be more efficient to perform a pre-pass and remember the addresses of every line number, so that every GOTO doesn't need to scan the whole program to find its target.
[Lisp] can be [interpreted]
Actually, Common Lisp cannot be implemented as a pure interpreter -- there are a few features of the language that you cannot implement without performing a pass over each function.
(Scheme, the other dominant dialect of Lisp, can be implemented as a pure interpreter, a pure compiler, or a hybrid design.)
C is more readable, though.