That's ok, you're still learning to use computers. Once you get a bit more experience like us professionals, you'll realize those drives make excellent coffee cup holders.
Excellent idea. I'd also suggest some "classic" pictures in the rare, but important, case that the recipient happens to be an aspiring slashdot troll and history buff. Make sure the titles are misleading and the icon preview settings are turned off, as being helpful is always appreciated.
Let's say that's true. In that case, Occam's Razor is (logically, provable)
false. The number of assumptions in a theory proves nothing about whether it
is true or false; it is obviously possible to make an assumption which turns
out to be correct, even if there was no evidence for it at the time.
Remember, theology isn't science, and Occam's Razor is used as a logical principle - it can't be disproved. Theology interprets the nature and the intentions of its god using logic and assumptions primarily, and interpretation of texts and events.
For example, you could use Occam to decide how many angels can simultaneously dance on a pin. The existence of angels is beyond question, and the supposed knowledge about them is contained in sacred texts and discussions which are beyond question as well. So the whole exercise involves deriving statements from initial "facts", and if you can derive an answer using fewer of the available facts you win. Since there is never any direct experimental evidence, Occam is literally interpretable as a method to find the truth.
Perhaps a better way of phrasing it would be that Occam's Razor doesn't tell
you which model is right, or even more likely to be right in a probabilistic
sense. Instead, it tells you not bother with a model which make unnecessary
assumptions when there is a simpler model at hand with fewer assumptions which
still fits the data.
To go back to your original point, you're glossing over the important question of how to apply the last sentence, ie how to identify a simpler model which still fits the data. Two models generally don't give identical results, so to decide which to use requires both a measure of complexity, and a measure of fit. Both measures must be computable on both models, and usually they go in opposite directions: a more complex model usually fits slightly better than a simpler model, but has more parameters.
For example, the AIC literally counts and weighs the complexity of competing models (through the number of parameters), but also computes a likelihood function to measure fit. Then it adds the two numbers together and finds the model with the optimal AIC value. This is (one example) where Occam's Razor is used quantitatively in practice, not as a guide but as a formal decision criterion to identify "the" best model. It's not quite as bad as deciding what is "the" truth, but operationally it amounts to something close.
Because that approach (just do it) has always been used in the past with bad results. If the format doesn't become popular, there's no harm done. But if the format does become popular, problems inevitably surface that are hard to fix precisely because the format is popular.
The kind of problems that surface are: proprietary control that keeps changing (eg like MS Office file formats), forking and fragmentation (eg like when each browser decides to support their own version of HTML), unforeseen security and extensibility problems, etc.
Just do it is the right way if you don't anticipate that many people will use it, but if you have a reasonable expectation that this will form part of the backbone of the internet, you should have a proper engineering review even though it takes years.
Can you imagine all the people on the
train playing angry birds by voice command!?
Large metropolis, 7:30 Monday morning, a train heading towards the CBD:
Commuter1: "Caw"
Commuter2: "Caw-caw"
Commuter3: "Caw-caw, caw-caw"
Commuter1: "Caw-ca-caw-caw"
Commuter3: "Caw-caw-caw-caw"
Commuter1: "Caw-caw, Damn, hey Commuter3, keep it down will you? I'm in the middle of a game!"
Commuter3: "Shut up Commuter1, you stupid loser, Caw-caw-caw-caw-caw-caw....caw!"
Commuter2: "Hey guys, stop arguing, it's interfering with my game"
Commuter1: "Fuck you Commuter2, I just lost the high score, want me to come over there and help you? Caw-Caw! How d'you like them birds?"
???
News at 11: "This morning's tragedy in the commuter train will be remembered for a long time..."
The central point I was making -- which you seem to have completely missed --
is the OP was claiming Occam's Razor told us the heliocentric model was
correct and the Ptolemaic model was wrong. This is completely untrue. They
are each equally true from within their own frame of reference.
A frame of reference is a coordinate system, nothing more. It's not a point of view or some state of mind that makes something true or false. Both the geocentric and heliocentric models can (and must) be compared on their predictions. In case both models produce comparable predictions (which I'll grant you for argument's sake) then Occam does in fact apply and prefers heliocentrism.
The truth is there is no true state of affairs: it all depends on
your point of view. And that revelation lies at the heart of relativity.
I highlight this because I think this is where you fundamentally misunderstand relativity. Relativity isn't about a point of view, what's in the mind of the observer doesn't matter, only measurements matter. In classical physics, all models and theories are approximations of the truth, yet there are no multiple truths, there is just the one reality. There are multiple descriptions of reality, and the description with the best prediction rate is considered superior to all the others.
Yes, relativity concerns itself an awful lot with inertial reference frames.
However, it doesn't ever claim the inertial reference frame is true: it only
claims the math is easier.
No. It doesn't just claim the math is easier, it claims that the inertial frames are special, that the laws of physics have a particular form that is invariant. If you use any other reference frame, you will have spurious forces and ghost fields and maybe you'll have to invent new particles etc. The point is physics generally looks different in a non-inertial frame, so it's not just the math gets harder, it's your physical knowledge gets thrown out as well.
The only way to do physics without starting from scratch is by relating everything back to the special class of inertial frames of reference.
Heliocentrism is simpler and easier to get right. Those are enormous virtues,
and enough to recommend it. Why do we also need to believe that it is somehow
"more true" than the Ptolemaic version? There's no reason for us to do that,
and it's a gross abuse of Occam's Razor to make that claim.
Because the ptolemaic version doesn't offer the same accuracy of prediction that the heliocentric version offers and many incorrect predictions on top of that. In the geocentric model, sometimes planets go backwards for no apparent reason, then forwards again. That's not just a mathematical complication, it's a new physical behaviour that causes ripple effects in the physics.
If you accept the ptolemaic model as true, now you have to always assume any planet could do a U-turn at some time.
You also can't use Newton's laws in the ptolemaic model, and you can't use Maxwell's electromagnetism as is.
The only reasonable thing you can do with geocentrism is start with an inertial frame and work backwards, figuring out what things would look like from the ptolemaic point of view. But that's implicitly considering inertial frames as a privileged "source", so you might as well accept Einstein's claim and consider them to be special and superior to all others.
This bears repeating: according to
special relativity, there are no privileged frames of reference.
That's just not true.
Just to be clear: according to special relativity, there are privileged frames of reference, they are called inertial frames of reference.
What SR says is that provided you stay within the family of inertial frames of reference, those are all equivalent and none of them is superior to any other, but they are all superior to non-inertial frames of reference.
And that of course still isn't the full story, since for any given problem, there will usually be some inertial frame of reference which is superior (more convenient to use than any other).
What SR guarantees is that it's ok to use straightaway the most convenient inertial frame you can, because you would get the same answer with any other one (just with more work). However using a non-inertial frame is not acceptable, because the laws of physics are different in that case (you could derive them by relating it all to an inertial frame, but that's *way* more work and error prone, just say no if you can)
The Ptolemaic system uses a non-inertial reference frame.
I'm afraid that's somewhat revisionist thinking. Occam's Razor was a scholastic principle designed to discover The Truth(TM) in theological debates. There was never any experimental evidence when discussing the finer points of the mind of god, and those kinds of debates among priests were never about one model being more likely than another, but about truly choosing the correct description of the purpose and will of god.
So you're weakening the power and purpose of Occam's Razor enormously by demoting it to merely an experimental guide. Originally, in a scholastic context, a better way describe Occam is that the theory with fewer assumptions was (logically, provably) the truth and the one with more assumptions and equal outcomes was (logically, provably) false.
What? That's wrong. Look, you're being confused about reference frames and their purpose.
On the one hand, any description of physics whatsoever which predicts accurate results is true - reference frames don't matter, they're just a detail. I think that's what you're trying to say, but it's basically trivial.
Where you are confused is when you compare the historically successful theories of physics. Einstein did not say all reference frames are equivalent, he said all inertial reference frames have the same known laws of physics (in particular, the same speed of light).
By staying within the family of inertial reference frames, you can use all the old, tried and true, discoveries of Galileo, Faraday, Maxwell, etc. as is. You can switch from one inertial frame to another, and don't have to make any extra adjustments at all, justchange the coordinates using a Lorentz transformation.
If you use any other kind of reference frame (and let me be clear, a reference frame is just a coordinate system, nothing more, you can invent as many as you like), then the first problem you face is what are the laws of physics? And the answer is you don't know - there will be extra new forces, things will appear or disappear for no apparent reason, etc. You will have to start from scratch to figure it all out. That's bad.
So not all reference frames are equal, the inertial frames are privileged because 1) they have known laws and 2) those laws remain exactly the same among all the inertial reference frames.
The ptolemaic system is a great example of a bad reference frame that you should never use. All the extra circles and spheres required to correct for known observations are a real artifact of the bad choice of coordinate system. When a planet suddenly starts goin backwards in that model, that's a law of physics expressed in a non-inertial frame of reference. But it's a law that disappears (it doesn't transfer) if you change to a copernican model. So did you really advance science by figuring out that ptolemaic-only law? No.
The lesson of relativity is not all frames are equivalent or all points of view are equally valid, the lesson of relativity is only inertial frames should be used, or else you'll be building circles upon circles, and spheres upon spheres.
No. An inertial reference frame is any reference frame (= coordinate system) which isn't accelerating. It can be at rest, (which is what you're implicitly thinking when you visualize the barycenter, although you're still neglecting rotation about the galactic center etc) or it can be in motion provided the motion is not accelerated.
The crucial point is that there's not one privileged reference frame, there's a family of them and you can transform from one to the other using Lorentz transformations, which happen to be extremely simple. That allows you to do physics by imagining another viewpoint if it's convenient.
You can even do physics with non-inertial reference frames, by slicing time into very small intervals and approximating the accelerated motion by constant motion within the intervals.
It was a revelation
to me that when they said "probability space" it was only a space in the
mathematical sense (ie, something with N dimensions that could be graphed if
N were not too large).
Just be glad you didn't study advanced probability, it gets really confusing when they start talking about "wiener space".
Communication does contain redundant elements, but those elements are not
entirely functionless
I am not talking about function, but about value and purpose to the reader.
Your focus on message integrity doesn't allow comparison between disparate reading
methods, except trivially by measuring total retention rate in the reader.
My point is that reading a text always a purpose. The purpose could be pure enjoyment of course,
but more often it is to extract some kind of information. To compare reading methods is to
compare the process that leads to the sought after information, or more abstractly the
process that starts with looking at the text and ends with the completion of the reader's immediate purpose.
The information sought is often a small part of the full message. For example, in an article
about silicon valley businesses, only a small part of the article may be about Microsoft,
other parts may be about Google, Facebook etc. A reader collecting information about Microsoft
has a purpose which does not require the full facts about Google and Facebook, and so on.
One, you do not know that the second reader maintains a better model. That is
your assumption.
Second, since the parts are skipped before they are read, it is impossible to
know they have low value. The reader could have skipped important parts
simply because she thought they're uninteresting, or because of the way the
writing was organized.
Of course my assumption is: visual and hearing improves
understanding. However, your second point illustrates perfectly the
issue: You implicitly value accuracy and completeness
only. That's certainly a possible purpose for reading, but for most reading
situations it isn't the full story.
To compare the two methods you look at the full process after the text has
been read. [...]
That is another unfounded assumption. There is no reason at all why things
would happen the way you describe.
There is. The process isn't complete until the purpose for reading has been achieved
(or has failed). The input can be evaluated as it is being read, or can be evaluated
some time after reading, but reading without taking into account the purpose is simply
Shannon communication theory.
In communication theory, the purpose of the end points is unknowable. The communications
medium could be connected to a variety of sources and sinks. That leads to your metric
of accuracy and completeness (ie transmission errors), as the only viable metric.
Here however the reader (sink) is necessarily part of the model. You can compare any reading methods
provided the end-result in the reader's brain is comparable. So if the reader's purpose is
knowing about Microsoft, the other facts about Google and Facebook can be discarded at any stage.
And if the reader's purpose includes speed or timeliness, that can be traded off against transmission errors.
As an aside, I am not sure why you assume that reading whole words somehow
reduces one's ability to skip words, should one choose to do so.
I don't. You can read whole words and skip parts of the text, or you
can read and sound out words and skip parts of the text. In the
latter case, my thesis is that the rate of comprehension is higher,
thereby leading to more accurate skipping. And if you elect not to skip,
that's allright too, but you're likely doing more work than necessary for your purpose.
That does not make sense. You say she's better off because "she can speed up". In reality she can only read faster by ignoring parts of
the input (what you call "not essential"), which besides making the two processes not comparable, it also requires her making assumptions
on the fly about those parts before getting to read them.
On the contrary, the two processes are directly comparable. The reason is that human language is not context free.
Previous sentences tend to indicate what comes next. That's the key to allowing optimization.
In both cases, the text being read is processed linearly in a single
pass. The difference is that in the first case all sentences are evenly processed, while in the second case the reader maintains
a better current model of the content to optimize the processing - by skipping parts that are uninteresting or have low value.
To compare the two methods you look at the full process after the text has been read. In the first case, all sentences were read
and the mind only decides at the end what among the content is worth retaining or making use of in the future. In the second case, the choice occurs "on-line".
To give a simple analogy with HTML/XML. In the first case, you read and parse the full DOM of a webpage, and then decide you need to
extract some URLs or text nodes in some subset of the web page. In the second case, you parse the webpage on the fly skipping the bits
you are sure won't be of interest, and being careful (slowing down) with nodes of potential interest. The analogy isn't great because
computer languages are generally context free, but I think the gist is true.
This can - and often does - result in reading comprehension issues with all but
the simplest material.
These "inessential" parts are often skipped when her mind's made up and she's convinced she already "got it", for
example as in a conversation where she'd have the reply on its way out way before you're done talking.
Comprehension issues occur in both cases, it's only whether they happen during or at the end of the read. However, my point is
that by sounding out, her comprehension level is assisted during the read, so she has better available intelligence to choose what to skip,
compared with the alternative of visual only recognition.
To begin with, he has no need to sound words out, therefore it is not relevant how much he'd slow down had he tried to.
This is only relevant if he tries the online processing trick described above. If he wants to, he can skip parts of the text on the fly
with purely visual recognition, but all things being equal (assuming equal brain capacity etc) he'll make more mistakes than her. Reducing those mistakes by sounding out the words would have an immediate slowdown penalty.
That's because she has twice the input bandwidth you do. Think about it, you just see/recognize the words, but she sees/recognizes them, and also hears them being spoken to her. That's reinforcement using two senses, sight and hearing, against your single sense of sight.
Her slowness is due to the need to synchronize the simultaneous inputs.
But she's better off than you, because she can speed up if she decides to read "intelligently". Most sentences and phrases in paragraphs are not essential, and can be skipped with little loss. She can train herself to recognize what's inessential, and deliberately not read it to speed up, or slow down as necessary.
You however are in a bad position. If you try to sound words out, you'll slow down so much you'll need strong incentives to stick with it. So you're probably stuck in a local optimum whereas she's in reach of a better one.
If you're an academic, running a single hadoop job like that is not as useful as it sounds. In research, you never know what you want until you do something and realize that's not it. To write a paper you'd want to run at least 10-20 full jobs, all slightly different.
Luckily, lots of unis have their own clusters (aka beowulfs - I can't believe I have to point that out on slashdot...). It would really be great if the data could be duplicated so people could run the jobs on their own local setups.
Furthermore, that's how you ensure that research is duplicable. If there's only one source of data in the world then all your published results are at the mercy of the org that owns the data. Bad science.
Actually, I think I fucked that one up. The bit size is exact, but there is no guarantee that the typedef for a particular size exists, so technically you can't rely on uint16_t in your programs anyway.
The people on the standards committes are either saints, or abject evil scum. Maybe they're in a quantum superposition of both states, as long as you don't open the ISO report and take a look...
Wrong. On account of backwards compatibility, the available infrastructure was never seriously enforced, and on account of idiotic security flag combinations the options available did not promote real security.
XP SP2 was the first time a (laughable) effort was actually made to enforce some security.
With SP2, enforcing security on sensitive API calls meant something trivial like inserting a pop up dialog box to ask the user for confirmation. All you had to do to bypass it was send a button click message to that window's message pump, no user interaction required!
But the problems lie much deeper. For example, CreateFile has a lot of access flags, but it's such a boneheaded system that the only hassle free way to allow more than one process to read/write the same file concurrently is to always open the file with the strongest access rights available. That's because once a file is open, the next process that tries to open it can't override the existing access modes. Crazy, eh?
That's like having a door with a fancy biometric hand sensor but now your friend can't walk through it at the same time as you, because his DNA isn't compatible with yours. So you either walk through the door one person at a time and close it/reopen it halfway through, or the first person to get to the sensor disables the DNA checks so you can both walk through together.
MSDN is full of impressive sounding security systems that programmers do their best to ignore, because they're unusable.
Nonsense. Windows security flaws have historically been due to boneheaded design decisions. Windows was never meant to run as a node on the internet, that functionality was retrofitted when Bill decided to do his famous 180 about turn on the computer highway because he'd missed the internet on-ramp. "Security" was retrofitted ten years later.
Problems with the C standard library certainly do exist and can expose security issues, but Windows security problems exist because the OS design emphasises user friendliness and backwards compatibility over tight protections.
It's like inside your home, you don't lock all your cupboards, drawers and doors - that would be painful, eg to walk from the kitchen to the living room you'd take out your key, unlock the door, open go through shut, relock the door, each time. To make your home livable you keep it insecure, and that's how Windows was designed from the ground up.
But now suppose there's a magic internet wormhole that opens in your toilet room, and anybody can enter your house. Suddenly it makes sense to have locks on all the doors and cupboards etc, but it's too late. Windows + Internet = insecure.
Unix doesn't have this problem, because Unix was always designed as a hotel (multiuser OS) rather than a home. So there's locks on the rooms and the swimming pool needs an access card etc. If a wormhole opens in the hotel lobby or even in one of the guest rooms, there's limited access to most areas by design.
Slashdot Mirror
FORD: you can have any error message background you like as long as it is blue.
I think you won the WOOSH prize.
That's ok, you're still learning to use computers. Once you get a bit more experience like us professionals, you'll realize those drives make excellent coffee cup holders.
Excellent idea. I'd also suggest some "classic" pictures in the rare, but important, case that the recipient happens to be an aspiring slashdot troll and history buff. Make sure the titles are misleading and the icon preview settings are turned off, as being helpful is always appreciated.
Remember, theology isn't science, and Occam's Razor is used as a logical principle - it can't be disproved. Theology interprets the nature and the intentions of its god using logic and assumptions primarily, and interpretation of texts and events.
For example, you could use Occam to decide how many angels can simultaneously dance on a pin. The existence of angels is beyond question, and the supposed knowledge about them is contained in sacred texts and discussions which are beyond question as well. So the whole exercise involves deriving statements from initial "facts", and if you can derive an answer using fewer of the available facts you win. Since there is never any direct experimental evidence, Occam is literally interpretable as a method to find the truth.
To go back to your original point, you're glossing over the important question of how to apply the last sentence, ie how to identify a simpler model which still fits the data. Two models generally don't give identical results, so to decide which to use requires both a measure of complexity, and a measure of fit. Both measures must be computable on both models, and usually they go in opposite directions: a more complex model usually fits slightly better than a simpler model, but has more parameters.
For example, the AIC literally counts and weighs the complexity of competing models (through the number of parameters), but also computes a likelihood function to measure fit. Then it adds the two numbers together and finds the model with the optimal AIC value. This is (one example) where Occam's Razor is used quantitatively in practice, not as a guide but as a formal decision criterion to identify "the" best model. It's not quite as bad as deciding what is "the" truth, but operationally it amounts to something close.
The kind of problems that surface are: proprietary control that keeps changing (eg like MS Office file formats), forking and fragmentation (eg like when each browser decides to support their own version of HTML), unforeseen security and extensibility problems, etc.
Just do it is the right way if you don't anticipate that many people will use it, but if you have a reasonable expectation that this will form part of the backbone of the internet, you should have a proper engineering review even though it takes years.
Large metropolis, 7:30 Monday morning, a train heading towards the CBD:
Commuter1: "Caw"
Commuter2: "Caw-caw"
Commuter3: "Caw-caw, caw-caw"
Commuter1: "Caw-ca-caw-caw"
Commuter3: "Caw-caw-caw-caw"
Commuter1: "Caw-caw, Damn, hey Commuter3, keep it down will you? I'm in the middle of a game!"
Commuter3: "Shut up Commuter1, you stupid loser, Caw-caw-caw-caw-caw-caw....caw!"
Commuter2: "Hey guys, stop arguing, it's interfering with my game"
Commuter1: "Fuck you Commuter2, I just lost the high score, want me to come over there and help you? Caw-Caw! How d'you like them birds?"
???
News at 11: "This morning's tragedy in the commuter train will be remembered for a long time..."
A frame of reference is a coordinate system, nothing more. It's not a point of view or some state of mind that makes something true or false. Both the geocentric and heliocentric models can (and must) be compared on their predictions. In case both models produce comparable predictions (which I'll grant you for argument's sake) then Occam does in fact apply and prefers heliocentrism.
I highlight this because I think this is where you fundamentally misunderstand relativity. Relativity isn't about a point of view, what's in the mind of the observer doesn't matter, only measurements matter. In classical physics, all models and theories are approximations of the truth, yet there are no multiple truths, there is just the one reality. There are multiple descriptions of reality, and the description with the best prediction rate is considered superior to all the others.
No. It doesn't just claim the math is easier, it claims that the inertial frames are special, that the laws of physics have a particular form that is invariant. If you use any other reference frame, you will have spurious forces and ghost fields and maybe you'll have to invent new particles etc. The point is physics generally looks different in a non-inertial frame, so it's not just the math gets harder, it's your physical knowledge gets thrown out as well.
The only way to do physics without starting from scratch is by relating everything back to the special class of inertial frames of reference.
Because the ptolemaic version doesn't offer the same accuracy of prediction that the heliocentric version offers and many incorrect predictions on top of that. In the geocentric model, sometimes planets go backwards for no apparent reason, then forwards again. That's not just a mathematical complication, it's a new physical behaviour that causes ripple effects in the physics.
If you accept the ptolemaic model as true, now you have to always assume any planet could do a U-turn at some time. You also can't use Newton's laws in the ptolemaic model, and you can't use Maxwell's electromagnetism as is.
The only reasonable thing you can do with geocentrism is start with an inertial frame and work backwards, figuring out what things would look like from the ptolemaic point of view. But that's implicitly considering inertial frames as a privileged "source", so you might as well accept Einstein's claim and consider them to be special and superior to all others.
That's just not true. Just to be clear: according to special relativity, there are privileged frames of reference, they are called inertial frames of reference.
What SR says is that provided you stay within the family of inertial frames of reference, those are all equivalent and none of them is superior to any other, but they are all superior to non-inertial frames of reference.
And that of course still isn't the full story, since for any given problem, there will usually be some inertial frame of reference which is superior (more convenient to use than any other).
What SR guarantees is that it's ok to use straightaway the most convenient inertial frame you can, because you would get the same answer with any other one (just with more work). However using a non-inertial frame is not acceptable, because the laws of physics are different in that case (you could derive them by relating it all to an inertial frame, but that's *way* more work and error prone, just say no if you can)
The Ptolemaic system uses a non-inertial reference frame.
So you're weakening the power and purpose of Occam's Razor enormously by demoting it to merely an experimental guide. Originally, in a scholastic context, a better way describe Occam is that the theory with fewer assumptions was (logically, provably) the truth and the one with more assumptions and equal outcomes was (logically, provably) false.
On the one hand, any description of physics whatsoever which predicts accurate results is true - reference frames don't matter, they're just a detail. I think that's what you're trying to say, but it's basically trivial.
Where you are confused is when you compare the historically successful theories of physics. Einstein did not say all reference frames are equivalent, he said all inertial reference frames have the same known laws of physics (in particular, the same speed of light).
By staying within the family of inertial reference frames, you can use all the old, tried and true, discoveries of Galileo, Faraday, Maxwell, etc. as is. You can switch from one inertial frame to another, and don't have to make any extra adjustments at all, justchange the coordinates using a Lorentz transformation.
If you use any other kind of reference frame (and let me be clear, a reference frame is just a coordinate system, nothing more, you can invent as many as you like), then the first problem you face is what are the laws of physics? And the answer is you don't know - there will be extra new forces, things will appear or disappear for no apparent reason, etc. You will have to start from scratch to figure it all out. That's bad.
So not all reference frames are equal, the inertial frames are privileged because 1) they have known laws and 2) those laws remain exactly the same among all the inertial reference frames.
The ptolemaic system is a great example of a bad reference frame that you should never use. All the extra circles and spheres required to correct for known observations are a real artifact of the bad choice of coordinate system. When a planet suddenly starts goin backwards in that model, that's a law of physics expressed in a non-inertial frame of reference. But it's a law that disappears (it doesn't transfer) if you change to a copernican model. So did you really advance science by figuring out that ptolemaic-only law? No.
The lesson of relativity is not all frames are equivalent or all points of view are equally valid, the lesson of relativity is only inertial frames should be used, or else you'll be building circles upon circles, and spheres upon spheres.
The crucial point is that there's not one privileged reference frame, there's a family of them and you can transform from one to the other using Lorentz transformations, which happen to be extremely simple. That allows you to do physics by imagining another viewpoint if it's convenient.
You can even do physics with non-inertial reference frames, by slicing time into very small intervals and approximating the accelerated motion by constant motion within the intervals.
Just be glad you didn't study advanced probability, it gets really confusing when they start talking about "wiener space".
Tell 'em you're a farmer, and you grow app's.
I'm a CS professor, you insensitive clod!
(*) named Henery.
I am not talking about function, but about value and purpose to the reader. Your focus on message integrity doesn't allow comparison between disparate reading methods, except trivially by measuring total retention rate in the reader.
My point is that reading a text always a purpose. The purpose could be pure enjoyment of course, but more often it is to extract some kind of information. To compare reading methods is to compare the process that leads to the sought after information, or more abstractly the process that starts with looking at the text and ends with the completion of the reader's immediate purpose.
The information sought is often a small part of the full message. For example, in an article about silicon valley businesses, only a small part of the article may be about Microsoft, other parts may be about Google, Facebook etc. A reader collecting information about Microsoft has a purpose which does not require the full facts about Google and Facebook, and so on.
Of course my assumption is: visual and hearing improves understanding. However, your second point illustrates perfectly the issue: You implicitly value accuracy and completeness only. That's certainly a possible purpose for reading, but for most reading situations it isn't the full story.
There is. The process isn't complete until the purpose for reading has been achieved (or has failed). The input can be evaluated as it is being read, or can be evaluated some time after reading, but reading without taking into account the purpose is simply Shannon communication theory.
In communication theory, the purpose of the end points is unknowable. The communications medium could be connected to a variety of sources and sinks. That leads to your metric of accuracy and completeness (ie transmission errors), as the only viable metric.
Here however the reader (sink) is necessarily part of the model. You can compare any reading methods provided the end-result in the reader's brain is comparable. So if the reader's purpose is knowing about Microsoft, the other facts about Google and Facebook can be discarded at any stage. And if the reader's purpose includes speed or timeliness, that can be traded off against transmission errors.
I don't. You can read whole words and skip parts of the text, or you can read and sound out words and skip parts of the text. In the latter case, my thesis is that the rate of comprehension is higher, thereby leading to more accurate skipping. And if you elect not to skip, that's allright too, but you're likely doing more work than necessary for your purpose.
On the contrary, the two processes are directly comparable. The reason is that human language is not context free. Previous sentences tend to indicate what comes next. That's the key to allowing optimization.
In both cases, the text being read is processed linearly in a single pass. The difference is that in the first case all sentences are evenly processed, while in the second case the reader maintains a better current model of the content to optimize the processing - by skipping parts that are uninteresting or have low value.
To compare the two methods you look at the full process after the text has been read. In the first case, all sentences were read and the mind only decides at the end what among the content is worth retaining or making use of in the future. In the second case, the choice occurs "on-line".
To give a simple analogy with HTML/XML. In the first case, you read and parse the full DOM of a webpage, and then decide you need to extract some URLs or text nodes in some subset of the web page. In the second case, you parse the webpage on the fly skipping the bits you are sure won't be of interest, and being careful (slowing down) with nodes of potential interest. The analogy isn't great because computer languages are generally context free, but I think the gist is true.
Comprehension issues occur in both cases, it's only whether they happen during or at the end of the read. However, my point is that by sounding out, her comprehension level is assisted during the read, so she has better available intelligence to choose what to skip, compared with the alternative of visual only recognition.
This is only relevant if he tries the online processing trick described above. If he wants to, he can skip parts of the text on the fly with purely visual recognition, but all things being equal (assuming equal brain capacity etc) he'll make more mistakes than her. Reducing those mistakes by sounding out the words would have an immediate slowdown penalty.
Good! You're making an effort to use more than just your eyes...
Her slowness is due to the need to synchronize the simultaneous inputs.
But she's better off than you, because she can speed up if she decides to read "intelligently". Most sentences and phrases in paragraphs are not essential, and can be skipped with little loss. She can train herself to recognize what's inessential, and deliberately not read it to speed up, or slow down as necessary.
You however are in a bad position. If you try to sound words out, you'll slow down so much you'll need strong incentives to stick with it. So you're probably stuck in a local optimum whereas she's in reach of a better one.
If you're an academic, running a single hadoop job like that is not as useful as it sounds. In research, you never know what you want until you do something and realize that's not it. To write a paper you'd want to run at least 10-20 full jobs, all slightly different.
Luckily, lots of unis have their own clusters (aka beowulfs - I can't believe I have to point that out on slashdot...). It would really be great if the data could be duplicated so people could run the jobs on their own local setups. Furthermore, that's how you ensure that research is duplicable. If there's only one source of data in the world then all your published results are at the mercy of the org that owns the data. Bad science.
The people on the standards committes are either saints, or abject evil scum. Maybe they're in a quantum superposition of both states, as long as you don't open the ISO report and take a look...
XP SP2 was the first time a (laughable) effort was actually made to enforce some security. With SP2, enforcing security on sensitive API calls meant something trivial like inserting a pop up dialog box to ask the user for confirmation. All you had to do to bypass it was send a button click message to that window's message pump, no user interaction required!
But the problems lie much deeper. For example, CreateFile has a lot of access flags, but it's such a boneheaded system that the only hassle free way to allow more than one process to read/write the same file concurrently is to always open the file with the strongest access rights available. That's because once a file is open, the next process that tries to open it can't override the existing access modes. Crazy, eh?
That's like having a door with a fancy biometric hand sensor but now your friend can't walk through it at the same time as you, because his DNA isn't compatible with yours. So you either walk through the door one person at a time and close it/reopen it halfway through, or the first person to get to the sensor disables the DNA checks so you can both walk through together.
MSDN is full of impressive sounding security systems that programmers do their best to ignore, because they're unusable.
Somebody's going to get sued for false advertising...
Problems with the C standard library certainly do exist and can expose security issues, but Windows security problems exist because the OS design emphasises user friendliness and backwards compatibility over tight protections.
It's like inside your home, you don't lock all your cupboards, drawers and doors - that would be painful, eg to walk from the kitchen to the living room you'd take out your key, unlock the door, open go through shut, relock the door, each time. To make your home livable you keep it insecure, and that's how Windows was designed from the ground up.
But now suppose there's a magic internet wormhole that opens in your toilet room, and anybody can enter your house. Suddenly it makes sense to have locks on all the doors and cupboards etc, but it's too late. Windows + Internet = insecure.
Unix doesn't have this problem, because Unix was always designed as a hotel (multiuser OS) rather than a home. So there's locks on the rooms and the swimming pool needs an access card etc. If a wormhole opens in the hotel lobby or even in one of the guest rooms, there's limited access to most areas by design.