I have plenty of karma to burn going offtopic. (See my Mr. Burns signature.)
But I think you're wrong about most people agreeing with this statement. It helps sometimes if you target something specific -- for example, my father might agree if I used it to apply to Christianity, since he's a Jew and finds much of Christian philosophy to be stupid and inconsistent.
But the reality is, we live in a world of religious tolerance, which generally tolerates other religions (the ones deemed "legitimate") more easily than with lack of a religion. It does seem almost arbitrary to me that, for instance, JudeoChristianity/Islam, Hinduism, Taoism, Buddhism, Wiccan, and various Native American beliefs are things which must be treated with respect, yet Maharishi Mahesh Yogi, Uri Geller, the Psychic Hotline, and beliefs about UFOs must be ridiculed and disproved.
So my guess is, most religious people would have a hard time saying "yes" to that statement applied to other "legitimate" religions, unless they were willing to apply it to their own.
I admit my personal approach is arbitrary. I generally allow people to believe whatever they want, until they bring it up in conversation. Then, depending on how stupid I think it is or their application of it is, I may simply try to change the subject, or I may directly attack.
Actually, a lot of it hinges on how stupid I think the person is for believing whatever it is. So I do cut some slack to followers of "legitimate" religions, because they've likely been indoctrinated since birth, and it's hard to break away from that, even if you want to.
I think you do need it. It's a basic part of the theory in Object Oriented Programming.
I honestly have to say that no website or book I've seen about programming has ever had the words "noun" or "integral" even mentioned, let alone defined.
However, I think he said much better than I did:
He does make some very good points. And my feeling is, the lines get very blurry in computers, but if I had to choose, it'd probably be engineering. Except "Computer Engineering" classes are, if I understand them, more about the physical (hardware) aspects of computing than about engineering the software.
We're stilling waiting for the headline "Psychic Wins Lottery."
If I was a psychic, why would I stop at one?
"Psychic wins three different kinds of lotteries in three separate states, as well as substantial bets on horse races. Psychic gently asked to knock it the fuck off and let somebody else win something for a change."
What makes mathematics what it is, is the method of theorems, and logical proofs. Anywhere you can state something abstract as a theorem and use it to prove statements, there is mathematics.
You can actually go two distinct ways with pretty much the same thing: Mathematics or Philosophy. Math will tend to be mostly about numbers and how to analyze them, while Philosophy will tend to be mostly about the nature of existence.
Just because you can define each in terms of the other doesn't mean it's a branch of that field. Computer science may be very well informed by math, but comp sci is NOT simply a branch of mathematics, any more than philosophy is.
In that sense, all Knuth books are heavily mathematical.
What makes English what it is, is the method of subjects, objects, and verbs, used to construct sentences, which are then arranged into paragraphs. In that sense, all Knuth books are heavily literary.
Computer science is its own field. Entirely. It's time we started treating it as such. Yes, we need a little math, just as we need a little English to understand documentation, but we don't need to know what a noun is, and we shouldn't need to know what a derivative is, in order to get started in computer science.
Imagine if, the next time you were to buy a soda at your favorite fast food joint, you were offered the choice of a 10 oz., a 11 oz., or a 12, 13, 14, 15, 16, ad infinitum oz. drink, rather than small, medium, or large.
I only wish. Then maybe people would stop turning into whimpering piles of pus when given too much choice!
It really doesn't take a genius to look at those choices and figure out that a 10 oz is small, and a 30 or 40 oz is large. Yes, offering it in 1 oz increments is a bit extreme, but it also should not cause a problem for you.
Of course, the right way to do this is even easier: Offer small, medium, and large, and specify how much each is -- along with the option to ask for exactly how much you need. And maybe, too, it's a bad analogy. But I do find it pathetic that people need things dumbed down because there's too much choice. If it's really so bad, just close your eyes and pick one at random!
The iPhone already has accelerometers, and the reason is simple: It can then figure out when you flip it on its side. Thus, it automatically changes the interfaces to reflect whether you're holding it vertically or horizontally. Combine that with other sensors, and it can figure out when you're holding it to your ear (and thus disable the display).
Others have already mentioned a lot of the creative things that could be done with it beyond that, so I'll mention one -- scrolling. Put simply, if it's accurate enough, imagine having a document that is digitally the size of a wall -- rather than scrolling through it, you simply move the phone around the document.
Here's an example: Ring tones. Ring tones, wallpapers, screensavers, and assorted bullshit. The typical way you get this on a closed phone is, you dial some number you saw on TV, and you get charged something like $1 to download some tiny fragment of a song you doubtless already bought on CD.
On an open phone, you just rip the CD, then send the file to your phone -- like you would with an iPod, say.
The real reason for DRM -- not that people listen when I point this out -- is to be able to do crap like that. Sell you the same song five times -- once on a CD (which can ONLY be played on CD players, and not even all of them), again for your iPod, again as a ringtone on your phone, again as a soundtrack for your game console, and a fifth time because you'll lose one of the others and can't make backups.
But I don't think developers are going to create an amazing consumer application. If we do, someone will find a way to charge everyone on a "normal" phone to get the same thing. I'd settle for an amazing developer phone, and if we do create something useful, and users buy the phone and download our useful software, more power to them. I just want something I can hack.
The fact that it is stated in a functional programming language does not mean it's not an algorithm.
And yet, it is a completely different way to think about it.
Am I using the wrong word? I think of "algorithm" as meaning a set of clearly defined, step-by-step instructions -- in other words, a list of imperative commands. A functional language does deconstruct to that, but a functional programmer does not think that way.
I think you've just lost your credibility on the subject.
Really?
So tell me -- where would I have found a B-Tree useful to implement myself, other than in my own filesystem? Or maybe a database? Just about anything else I can think of is either simple enough to use the language's built-in hash in memory, or already solved in a database or a filesystem.
(Actually, I didn't go with B-Trees. I went with Trinary trees -- a kind of combination B-Tree and Radix tree, that also has interesting properties when it comes to searching -- you get directories for free.)
completely reworked their tortured code-paths and structures using a little calculus, and turned it into elegant, maintainable, efficient, and simple code.
I hate to be rude, but I'll believe that when I see it.
I often do the same thing, by the way, I'm just not using calculus, at least not consciously. I'm sure you have a theory to prove just how much more elegant, maintainable, efficient, and simple your code is... I just have eyeballing and benchmarks.
I suspect that I will want to go back and finish the computer science degree I started at some point, just not now. Frankly, I don't have the self-discipline for it. But if I do go back to school, it will be mostly for the degree, and not for the gaps in my knowledge -- which are there, but I usually fix those as I run into them.
if you're turning computer science into a course on programming then where are the actual computer scientists going to come from?
People who took the more advanced courses.
Consider: programming is used in business (where you need to understand a bit of accounting), aviation (where you need physics), game development (where you need math and artwork), systems design (where you need hardcore programming and people skills)...
Programming, in general, is too broad for a "computer scientist" to be limited to those exploring programming + math. I'd say, give them a course in programming (and not "just" programming, but really good, solid software design principles), and let them combine that with whatever else and make their own degree.
We don't need people claiming to be computer scientists when they don't know a Turing machine from a kick in the nut sack.
How do you show that your caching technique is working, and is efficent enough for your requirements?
Benchmark.
Can you be sure that you're reading them correctly?
Sure, build a unit test. It won't prove it mathematically, but I think you're a lot less likely to make an error in your unit test that leads to a faulty program passing the test than you are to make an error in your mathematical "proof".
Can you ensure that the file system is storing it correctly?
Same answer -- though in this case, the real answer is, use a known-good filesystem. 99% of the programming jobs out there do not involve writing a filesystem, and 99% of the programs out there do not really require any particular filesystem, other than one which is compatible with the platform they run on.
Are you doing this efficently(for a given value of efficency)? Can you prove it?
You've got me there. However, unless we're talking about re-implementing IP, you're generally doing something efficiently enough.
AI behavior is, at the moment, purely based in mathematical algorithms.
AI is also not a necessary component, though it is often used. You'd have better luck arguing about the graphics, but most of this work is already done for you -- either at the API level (OpenGL) or at the algorithm level (tricks like Carmack's Reverse).
You NEED a substantial mathematical background to even approach this problem-it's not just "it's X away, so we don't render y polys".
Actually, that is precisely what LoD is. I'm betting on a particular algorithm I've already seen which can take a 3D model and reduce its precision, but very slowly -- simply do this ahead of time, and cache the smaller model. All that's left is to figure out when to display the higher detail, and when the lower detail, which is mostly a matter of tweaking it till it looks right, and having the system dynamically adjust the ratio between your X and your y in order to hit a target framerate.
The system needs to decide which polys it can extrapolate over, which ones it doesn't need to render, and it needs to do this FAST.
That actually sounds a lot more like frustrum, backface, and occlusion culling. And here, there's a wide variety of algorithms to choose from.
Not having a strong mathematical background, and not being a genius, means I'm not particularly likely to invent such an algorithm. But I can put the pieces together, and in games in particular, this is often not done right at all. In particular, every time you see a loading screen (once you've started the game at all), it means someone at my "codemonkey" level got lazy.
The college I went to seemed to have another policy -- throw the really shitty teachers at the 1st-year, 1st-semester comp sci students. Take some snot-nosed kid off the street who likes computers cause he played Halo once, and give him Hello World in Java. Just so you remember:
class Hello { public static void main (String [] args) { System.out.println("Hello, world!"); } }
It was, very obviously, not designed to teach the kids anything. It was designed to find out who was good at doing one of:
- Already knowing the answer
- Looking it up online
- Learn it from the book
- Asking someone who can do the above
In other words, it was designed to figure out who already knew how to program, or who could learn quickly enough, and weed out everyone else. It was NOT designed to teach you anything -- that's what the second semester was for, if you survived the first.
That's all my subjective opinion, of course, but I really don't like that attitude -- we don't know how to teach, so instead we'll just fail you if you don't get it. I'm not saying that every Halo asshole is capable of learning Java, but we should be paying for an education, not a certification.
Also: I seem to be fairly good at programming, and fairly bad at calculus and math in general. Maybe I'm just bitter about failing Calculus, though that was my own damned fault (I didn't go to class)...
The Book of Stuff that Other People Already Invented.
I generally don't do the algorithms stuff, for the same reason that I don't program in C (or assembly). I like to have a decent understanding of why a particular algorithm is good, but generally, there's enough work already done for me that I can take someone's word for how well a particular algorithm scales.
Congratulations, you're an academic. You will spend a lifetime reinventing LISP and Haskell in order to more perfectly optimize your mathematical algorithms that run in O(lookatme) time, while the rest of us "code monkeys" do the real work -- including build the infrastructure necessary for your LISP wankery.
Or we could cut the name calling and actually talk.
is based on the idea that algorithms are the wrong model for program.
For some problems, I'd tend to agree. Talk to a functional programmer, for instance.
I don't do that, of course, because I've been doing imperative algorithms for long enough that it's difficult to wrap my head around things like Erlang.
You know how to use trees and graphs?
I actually had to learn trees recently. I did not know what a B-Tree was, except that most filesystems use them, so I ignored them until I started writing a filesystem.
You said you did CS in college so I'll guess that you're familiar with parsing / compiler writing / language design.
Well, I didn't finish. But I do know a bit about this, some of which isn't a good match for abstract algebra at all -- LISP derivatives, for example. (I understand you could call it a subset, but the fact is, you can do LISP without understanding anything about algebra or operator precedence.)
I would also suggest to you that, while some of these concepts may come from math originally, more often it's more like a loose analogy. You mention a few things that have to do with math; let me mention a few that I doubt have much to do with traditional math: caching, file formats, protocols, objects and inheritance, package managers, simulations (games) and things like dynamic level-of-detail...
And again, I'd suggest that a student who takes a philosophy class as a prerequisite to learn-to-program 101 is much more likely to get the concepts. I'd suggest this partly because much of the math that's required is simply not used, does not necessarily require the same skills as programming, and may actually be more difficult for a good coder to learn than the language itself -- among other things, it's easy to take Hello World out for a spin and actually make sure you understand what's going on. Not so easy with most math, beyond arithmetic and trigonometry.
If I were to design a CS curriculum, I'd require math up to and including algebra, and beyond that, only what's required for a particular domain (trig and maybe basic calc for game engines, for example). I'd also require english, philosophy, logic (if there is such a class), and any subject that requires the student to come up with a researched report (citing sources). Then I'd sit the kids down with the interpreter/compiler of the day, a manual, Google, and an assignment, and be there to offer help.
I went to Macchu Picchu about two weeks ago. They had computers lined up outside, hooked up to the Internet, Kiosk-moded to the survey website for this retarded thing.
I thought exactly the same thing as all of you, but I'd also argue that there's probably a good chunk of the votes coming from the shell-shocked people walking out of places like Macchu Picchu. I know I was tempted, because goddamn... It's beautiful, and amazing architecture, and a good bit of impressive history, and a little halucinogenic garden...
But this kind of shit really cheapens it. That, and the $70 ticket for the train to Aguascalientes (a small town of "hot" springs that are kind of lukewarm, that exists only to be a tourist town), then another, what, $20 for the bus ride and $50 for the ticket? I forget (I wasn't paying). The person who did pay (my father) was certainly convinced it was worth every penny, and I am too, but the commercialization is bullshit. Oh well, at least the rest of the country is cheap...
And I'm sorry but mathematics is just the way in which meaning is expressed for machines.
But not for humans.
And he's wrong about algorithms too - since a non-terminating algorithm is always expressible by deconstruction into a series of terminating algorithms.
I have no idea what you just said, and I've been coding for years.
I agree that computing is hard. Well, I find it easy, but I agree that, in general, if you're going to use a computer, you're going to learn some logic, and I will not help you to avoid thinking.
But 99% of the apps I write require little to no knowledge of mathematics, beyond basic algebra and arithmetic, and maybe a few binary tricks. In particular, Calculus was a required course for CS in college, and I have never used Calculus in my computing, even programming. Ever.
I have not read that book, but I would argue that a big reason computer science is stunted is this absurd relation to mathematics. You need better-than-average communication skills if you're ever going to work on a team -- at least to comment your code. You need philosophy -- yes, philosophy! -- at least as much as you need math, and a decent logic class would be even more useful. And you need research skills a bit beyond justfuckinggoogleit.com, if you're going to RTFM -- considering you may have to go looking for a manual, or even the right software to begin with.
I have a 20" monitor that I run at 1600x1200, and I use a Konsole with a slightly smaller-than-normal-sized font.
Why?
So I can have six terminal windows, an IM window, and a web browser all open at once. Or two terminal windows, an IM window, a web browser, and an MMO or a movie. Or any number of similar configurations.
Why so much crap?
Well, IM because I'm always on IM. Web browser for web-based documentation -- Google is your friend, and search.cpan.org if you're into Perl -- and maybe another tab open to the app itself, if it's a web app. One terminal for the source, maybe another terminal for a related file, of course a terminal to actually run the app in, a terminal to read a manpage in... and maybe a terminal or two for random stuff, like playing/controlling music, or doing admin stuff (aptitude to install some library that might help, or a cpan shell if it's Perl and not in Ubuntu yet)...
Yes, I know about multiple monitors, but then I have to turn my head, and also buy another monitor. And yes, I know about multiple desktops, but then, I have email on one of them, and I can always throw other stuff onto another one, but that doesn't buy me the ability to actually have the documentation page and my own code on the screen at the same time. Switching back and forth is, frankly, little better than alt+tab.
Another reason is, with Beryl, if I feel like, say, reading a book, I can give my eyes a rest and win+mousewheel, zoom in on the terminal, and get text I can read from across the room, or as far as my keyboard's USB cable will reach. (I know about wireless, and I don't like it, and I've got a few 6-foot extenders anyway. Plus, no batteries to run out.)
If you're having problems with that, the simplest suggestion I have is, change your tabstop (and have the project change it, also). I find that with a tabstop of 4 and an 80-character width, I rarely nest deep enough to wrap, and when I do, I take a serious look at what I'm doing -- looking to refactor it, maybe slice some of those deeply-nested things out and put them into another function.
The stuff I sell for money because I and my family like to eat is a different story.
I don't mean to judge you here, but let's be clear: You sell stuff for money. With the GPL, I trade code for code.
I understand that there are practical differences, but I don't see what makes yours an honest living, and mine some sort of hidden threat of a lawsuit.
Nor am I claiming the opposite -- that I'm some altruistic saint, and you're an evil proprietary developer (developer! developer! developer!) -- I'm just pointing out that I always have, and always will see the GPL as one of the clearest, fairest licenses out there, especially when compared to proprietary software. BSD may be better, and it may be worse, but "give me source or I'll sue" is not at all the tone of the GPL, it's entirely a response to people who abuse it.
I don't think the "deal" it offers is likely to sway anyone who didn't agree with the moral argument already.
I'd argue that it already has: Lawyers.
In short: Large companies like IBM and Oracle use and develop Linux. If Linux was BSD-licensed, the individual programmers might be fine with giving code back, but you might have a corporate lawyer who decides that you shouldn't give away anything you don't have to. People could sort of agree in principle, but be bound by an NDA.
This has happened before -- read the story of the printer driver that pissed RMS off so much in the first place that he started this whole thing. He did eventually track down the developer most responsible for writing the software, but the guy didn't have the source code, and even if he did, had signed legal agreements preventing him from releasing it.
Linux, though, is a different story. Sure, IBM could've gone with FreeBSD, but Linux is bigger, has more developers, more drivers, and I'd imagine it actually performs better, too. It certainly has been ported to more platforms. And because of the GPL, if you can make the (very impressive) business case (including GPL), then the deal is sealed -- your lawyers cannot come to you later and suggest you close everything up.
The only thing I consider ineffective here is that there are too many loopholes -- the existence of proprietary kernel modules is proof of that. You could argue that nVidia may never have developed a Linux driver at all, and you may be right. I'd argue that the situation is worse now -- there are tons of licenses and patents on that driver, probably most of which are not owned by nVidia. Had they been forced into GPL or nothing, we might've gotten nothing, but they might've also worked harder to ensure that they didn't license anything for use in their driver that they couldn't control.
For that matter, nVidia did pretty well keeping up for awhile, so maybe they never intended to go open until it became apparent just how fast Linux moves.
I think those who will contribute to open source because it's a superior model for developing commodity code far outnumber those who will contribute because they think proprietary is evil, and hence that BSD code will tend to go farther faster than more restricted code.
We can argue all day about whether that should be true, but in the real world, it isn't. Linux is much, much farther along than any BSD save OS X, which does not contribute much of its BSD code back to the community, and often has to have its arm twisted to give back its GPL'd and LGPL'd stuff.
I suppose Solaris arguably comes close... but hey, Solaris is also GPL'd!
You could also argue that the BSD network stack has gone farther (as parts of it are apparently in Windows), but I don't think that is good for the community, or an effective way to promote open source. It's good for Microsoft, and for Windows, and for people who depend on Windows, but it ends there.
Slashdot Mirror
I have plenty of karma to burn going offtopic. (See my Mr. Burns signature.)
But I think you're wrong about most people agreeing with this statement. It helps sometimes if you target something specific -- for example, my father might agree if I used it to apply to Christianity, since he's a Jew and finds much of Christian philosophy to be stupid and inconsistent.
But the reality is, we live in a world of religious tolerance, which generally tolerates other religions (the ones deemed "legitimate") more easily than with lack of a religion. It does seem almost arbitrary to me that, for instance, JudeoChristianity/Islam, Hinduism, Taoism, Buddhism, Wiccan, and various Native American beliefs are things which must be treated with respect, yet Maharishi Mahesh Yogi, Uri Geller, the Psychic Hotline, and beliefs about UFOs must be ridiculed and disproved.
So my guess is, most religious people would have a hard time saying "yes" to that statement applied to other "legitimate" religions, unless they were willing to apply it to their own.
I admit my personal approach is arbitrary. I generally allow people to believe whatever they want, until they bring it up in conversation. Then, depending on how stupid I think it is or their application of it is, I may simply try to change the subject, or I may directly attack.
Actually, a lot of it hinges on how stupid I think the person is for believing whatever it is. So I do cut some slack to followers of "legitimate" religions, because they've likely been indoctrinated since birth, and it's hard to break away from that, even if you want to.
I honestly have to say that no website or book I've seen about programming has ever had the words "noun" or "integral" even mentioned, let alone defined.
He does make some very good points. And my feeling is, the lines get very blurry in computers, but if I had to choose, it'd probably be engineering. Except "Computer Engineering" classes are, if I understand them, more about the physical (hardware) aspects of computing than about engineering the software.
If I was a psychic, why would I stop at one?
"Psychic wins three different kinds of lotteries in three separate states, as well as substantial bets on horse races. Psychic gently asked to knock it the fuck off and let somebody else win something for a change."
I have a question: Does this statement apply to religion?
If not, why not?
Does your explanation of "why not" apply to Geller, also?
You can actually go two distinct ways with pretty much the same thing: Mathematics or Philosophy. Math will tend to be mostly about numbers and how to analyze them, while Philosophy will tend to be mostly about the nature of existence.
Just because you can define each in terms of the other doesn't mean it's a branch of that field. Computer science may be very well informed by math, but comp sci is NOT simply a branch of mathematics, any more than philosophy is.
What makes English what it is, is the method of subjects, objects, and verbs, used to construct sentences, which are then arranged into paragraphs. In that sense, all Knuth books are heavily literary.
Computer science is its own field. Entirely. It's time we started treating it as such. Yes, we need a little math, just as we need a little English to understand documentation, but we don't need to know what a noun is, and we shouldn't need to know what a derivative is, in order to get started in computer science.
The only "reality check" I see here is that the new version is actually going to be about as expensive as the iPhone.
What makes you think it's not a competitor, other than not seeing ads every three seconds on cable TV?
I only wish. Then maybe people would stop turning into whimpering piles of pus when given too much choice!
It really doesn't take a genius to look at those choices and figure out that a 10 oz is small, and a 30 or 40 oz is large. Yes, offering it in 1 oz increments is a bit extreme, but it also should not cause a problem for you.
Of course, the right way to do this is even easier: Offer small, medium, and large, and specify how much each is -- along with the option to ask for exactly how much you need. And maybe, too, it's a bad analogy. But I do find it pathetic that people need things dumbed down because there's too much choice. If it's really so bad, just close your eyes and pick one at random!
Assuming you want it to sync to a desktop at all, what the hell is iTunes doing syncing your calendar?
Here's another question: Does it have to be iTunes? Would you move to another app if the other app could sync with your iPod and devices like this?
The iPhone already has accelerometers, and the reason is simple: It can then figure out when you flip it on its side. Thus, it automatically changes the interfaces to reflect whether you're holding it vertically or horizontally. Combine that with other sensors, and it can figure out when you're holding it to your ear (and thus disable the display).
Others have already mentioned a lot of the creative things that could be done with it beyond that, so I'll mention one -- scrolling. Put simply, if it's accurate enough, imagine having a document that is digitally the size of a wall -- rather than scrolling through it, you simply move the phone around the document.
Here's an example: Ring tones. Ring tones, wallpapers, screensavers, and assorted bullshit. The typical way you get this on a closed phone is, you dial some number you saw on TV, and you get charged something like $1 to download some tiny fragment of a song you doubtless already bought on CD.
On an open phone, you just rip the CD, then send the file to your phone -- like you would with an iPod, say.
The real reason for DRM -- not that people listen when I point this out -- is to be able to do crap like that. Sell you the same song five times -- once on a CD (which can ONLY be played on CD players, and not even all of them), again for your iPod, again as a ringtone on your phone, again as a soundtrack for your game console, and a fifth time because you'll lose one of the others and can't make backups.
But I don't think developers are going to create an amazing consumer application. If we do, someone will find a way to charge everyone on a "normal" phone to get the same thing. I'd settle for an amazing developer phone, and if we do create something useful, and users buy the phone and download our useful software, more power to them. I just want something I can hack.
Web browsers, I'd agree with you. But file browsers? You just need practice.
And yet, it is a completely different way to think about it.
Am I using the wrong word? I think of "algorithm" as meaning a set of clearly defined, step-by-step instructions -- in other words, a list of imperative commands. A functional language does deconstruct to that, but a functional programmer does not think that way.
Really?
So tell me -- where would I have found a B-Tree useful to implement myself, other than in my own filesystem? Or maybe a database? Just about anything else I can think of is either simple enough to use the language's built-in hash in memory, or already solved in a database or a filesystem.
(Actually, I didn't go with B-Trees. I went with Trinary trees -- a kind of combination B-Tree and Radix tree, that also has interesting properties when it comes to searching -- you get directories for free.)
The "Well, it's hard, mmmkay?" sounds like somebody concerned with keeping job security...
I realize there are some hard problems out there, but some of the hardest (like threading) have some really easy solutions (like stackless Python).
I hate to be rude, but I'll believe that when I see it.
I often do the same thing, by the way, I'm just not using calculus, at least not consciously. I'm sure you have a theory to prove just how much more elegant, maintainable, efficient, and simple your code is... I just have eyeballing and benchmarks.
I suspect that I will want to go back and finish the computer science degree I started at some point, just not now. Frankly, I don't have the self-discipline for it. But if I do go back to school, it will be mostly for the degree, and not for the gaps in my knowledge -- which are there, but I usually fix those as I run into them.
People who took the more advanced courses.
Consider: programming is used in business (where you need to understand a bit of accounting), aviation (where you need physics), game development (where you need math and artwork), systems design (where you need hardcore programming and people skills)...
Programming, in general, is too broad for a "computer scientist" to be limited to those exploring programming + math. I'd say, give them a course in programming (and not "just" programming, but really good, solid software design principles), and let them combine that with whatever else and make their own degree.
Call them software engineers, then. Whatever.
I don't have to. I can implement my app with a generic "sort" function, substitute in bubblesort and quicksort, and build some graphs.
That one, I probably could, if I were particularly interested. But I don't do ethernet. The lowest I'll probably ever go is TCP or UDP.
And sex is a kind of physics, but Human Sexuality is not in the physics department.
Benchmark.
Sure, build a unit test. It won't prove it mathematically, but I think you're a lot less likely to make an error in your unit test that leads to a faulty program passing the test than you are to make an error in your mathematical "proof".
Same answer -- though in this case, the real answer is, use a known-good filesystem. 99% of the programming jobs out there do not involve writing a filesystem, and 99% of the programs out there do not really require any particular filesystem, other than one which is compatible with the platform they run on.
You've got me there. However, unless we're talking about re-implementing IP, you're generally doing something efficiently enough.
AI is also not a necessary component, though it is often used. You'd have better luck arguing about the graphics, but most of this work is already done for you -- either at the API level (OpenGL) or at the algorithm level (tricks like Carmack's Reverse).
Actually, that is precisely what LoD is. I'm betting on a particular algorithm I've already seen which can take a 3D model and reduce its precision, but very slowly -- simply do this ahead of time, and cache the smaller model. All that's left is to figure out when to display the higher detail, and when the lower detail, which is mostly a matter of tweaking it till it looks right, and having the system dynamically adjust the ratio between your X and your y in order to hit a target framerate.
That actually sounds a lot more like frustrum, backface, and occlusion culling. And here, there's a wide variety of algorithms to choose from.
Not having a strong mathematical background, and not being a genius, means I'm not particularly likely to invent such an algorithm. But I can put the pieces together, and in games in particular, this is often not done right at all. In particular, every time you see a loading screen (once you've started the game at all), it means someone at my "codemonkey" level got lazy.
I really don't like that attitude.
The college I went to seemed to have another policy -- throw the really shitty teachers at the 1st-year, 1st-semester comp sci students. Take some snot-nosed kid off the street who likes computers cause he played Halo once, and give him Hello World in Java. Just so you remember:
class Hello {
public static void main (String [] args) {
System.out.println("Hello, world!");
}
}
It was, very obviously, not designed to teach the kids anything. It was designed to find out who was good at doing one of:
- Already knowing the answer
- Looking it up online
- Learn it from the book
- Asking someone who can do the above
In other words, it was designed to figure out who already knew how to program, or who could learn quickly enough, and weed out everyone else. It was NOT designed to teach you anything -- that's what the second semester was for, if you survived the first.
That's all my subjective opinion, of course, but I really don't like that attitude -- we don't know how to teach, so instead we'll just fail you if you don't get it. I'm not saying that every Halo asshole is capable of learning Java, but we should be paying for an education, not a certification.
Also: I seem to be fairly good at programming, and fairly bad at calculus and math in general. Maybe I'm just bitter about failing Calculus, though that was my own damned fault (I didn't go to class)...
The Book of Stuff that Other People Already Invented.
I generally don't do the algorithms stuff, for the same reason that I don't program in C (or assembly). I like to have a decent understanding of why a particular algorithm is good, but generally, there's enough work already done for me that I can take someone's word for how well a particular algorithm scales.
Congratulations, you're an academic. You will spend a lifetime reinventing LISP and Haskell in order to more perfectly optimize your mathematical algorithms that run in O(lookatme) time, while the rest of us "code monkeys" do the real work -- including build the infrastructure necessary for your LISP wankery.
Or we could cut the name calling and actually talk.
For some problems, I'd tend to agree. Talk to a functional programmer, for instance.
I don't do that, of course, because I've been doing imperative algorithms for long enough that it's difficult to wrap my head around things like Erlang.
I actually had to learn trees recently. I did not know what a B-Tree was, except that most filesystems use them, so I ignored them until I started writing a filesystem.
Well, I didn't finish. But I do know a bit about this, some of which isn't a good match for abstract algebra at all -- LISP derivatives, for example. (I understand you could call it a subset, but the fact is, you can do LISP without understanding anything about algebra or operator precedence.)
I would also suggest to you that, while some of these concepts may come from math originally, more often it's more like a loose analogy. You mention a few things that have to do with math; let me mention a few that I doubt have much to do with traditional math: caching, file formats, protocols, objects and inheritance, package managers, simulations (games) and things like dynamic level-of-detail...
And again, I'd suggest that a student who takes a philosophy class as a prerequisite to learn-to-program 101 is much more likely to get the concepts. I'd suggest this partly because much of the math that's required is simply not used, does not necessarily require the same skills as programming, and may actually be more difficult for a good coder to learn than the language itself -- among other things, it's easy to take Hello World out for a spin and actually make sure you understand what's going on. Not so easy with most math, beyond arithmetic and trigonometry.
If I were to design a CS curriculum, I'd require math up to and including algebra, and beyond that, only what's required for a particular domain (trig and maybe basic calc for game engines, for example). I'd also require english, philosophy, logic (if there is such a class), and any subject that requires the student to come up with a researched report (citing sources). Then I'd sit the kids down with the interpreter/compiler of the day, a manual, Google, and an assignment, and be there to offer help.
I went to Macchu Picchu about two weeks ago. They had computers lined up outside, hooked up to the Internet, Kiosk-moded to the survey website for this retarded thing.
I thought exactly the same thing as all of you, but I'd also argue that there's probably a good chunk of the votes coming from the shell-shocked people walking out of places like Macchu Picchu. I know I was tempted, because goddamn... It's beautiful, and amazing architecture, and a good bit of impressive history, and a little halucinogenic garden...
But this kind of shit really cheapens it. That, and the $70 ticket for the train to Aguascalientes (a small town of "hot" springs that are kind of lukewarm, that exists only to be a tourist town), then another, what, $20 for the bus ride and $50 for the ticket? I forget (I wasn't paying). The person who did pay (my father) was certainly convinced it was worth every penny, and I am too, but the commercialization is bullshit. Oh well, at least the rest of the country is cheap...
But not for humans.
I have no idea what you just said, and I've been coding for years.
I agree that computing is hard. Well, I find it easy, but I agree that, in general, if you're going to use a computer, you're going to learn some logic, and I will not help you to avoid thinking.
But 99% of the apps I write require little to no knowledge of mathematics, beyond basic algebra and arithmetic, and maybe a few binary tricks. In particular, Calculus was a required course for CS in college, and I have never used Calculus in my computing, even programming. Ever.
I have not read that book, but I would argue that a big reason computer science is stunted is this absurd relation to mathematics. You need better-than-average communication skills if you're ever going to work on a team -- at least to comment your code. You need philosophy -- yes, philosophy! -- at least as much as you need math, and a decent logic class would be even more useful. And you need research skills a bit beyond justfuckinggoogleit.com, if you're going to RTFM -- considering you may have to go looking for a manual, or even the right software to begin with.
I have a 20" monitor that I run at 1600x1200, and I use a Konsole with a slightly smaller-than-normal-sized font.
Why?
So I can have six terminal windows, an IM window, and a web browser all open at once. Or two terminal windows, an IM window, a web browser, and an MMO or a movie. Or any number of similar configurations.
Why so much crap?
Well, IM because I'm always on IM. Web browser for web-based documentation -- Google is your friend, and search.cpan.org if you're into Perl -- and maybe another tab open to the app itself, if it's a web app. One terminal for the source, maybe another terminal for a related file, of course a terminal to actually run the app in, a terminal to read a manpage in... and maybe a terminal or two for random stuff, like playing/controlling music, or doing admin stuff (aptitude to install some library that might help, or a cpan shell if it's Perl and not in Ubuntu yet)...
Yes, I know about multiple monitors, but then I have to turn my head, and also buy another monitor. And yes, I know about multiple desktops, but then, I have email on one of them, and I can always throw other stuff onto another one, but that doesn't buy me the ability to actually have the documentation page and my own code on the screen at the same time. Switching back and forth is, frankly, little better than alt+tab.
Another reason is, with Beryl, if I feel like, say, reading a book, I can give my eyes a rest and win+mousewheel, zoom in on the terminal, and get text I can read from across the room, or as far as my keyboard's USB cable will reach. (I know about wireless, and I don't like it, and I've got a few 6-foot extenders anyway. Plus, no batteries to run out.)
If you're having problems with that, the simplest suggestion I have is, change your tabstop (and have the project change it, also). I find that with a tabstop of 4 and an 80-character width, I rarely nest deep enough to wrap, and when I do, I take a serious look at what I'm doing -- looking to refactor it, maybe slice some of those deeply-nested things out and put them into another function.
I don't mean to judge you here, but let's be clear: You sell stuff for money. With the GPL, I trade code for code.
I understand that there are practical differences, but I don't see what makes yours an honest living, and mine some sort of hidden threat of a lawsuit.
Nor am I claiming the opposite -- that I'm some altruistic saint, and you're an evil proprietary developer (developer! developer! developer!) -- I'm just pointing out that I always have, and always will see the GPL as one of the clearest, fairest licenses out there, especially when compared to proprietary software. BSD may be better, and it may be worse, but "give me source or I'll sue" is not at all the tone of the GPL, it's entirely a response to people who abuse it.
I'd argue that it already has: Lawyers.
In short: Large companies like IBM and Oracle use and develop Linux. If Linux was BSD-licensed, the individual programmers might be fine with giving code back, but you might have a corporate lawyer who decides that you shouldn't give away anything you don't have to. People could sort of agree in principle, but be bound by an NDA.
This has happened before -- read the story of the printer driver that pissed RMS off so much in the first place that he started this whole thing. He did eventually track down the developer most responsible for writing the software, but the guy didn't have the source code, and even if he did, had signed legal agreements preventing him from releasing it.
Linux, though, is a different story. Sure, IBM could've gone with FreeBSD, but Linux is bigger, has more developers, more drivers, and I'd imagine it actually performs better, too. It certainly has been ported to more platforms. And because of the GPL, if you can make the (very impressive) business case (including GPL), then the deal is sealed -- your lawyers cannot come to you later and suggest you close everything up.
The only thing I consider ineffective here is that there are too many loopholes -- the existence of proprietary kernel modules is proof of that. You could argue that nVidia may never have developed a Linux driver at all, and you may be right. I'd argue that the situation is worse now -- there are tons of licenses and patents on that driver, probably most of which are not owned by nVidia. Had they been forced into GPL or nothing, we might've gotten nothing, but they might've also worked harder to ensure that they didn't license anything for use in their driver that they couldn't control.
For that matter, nVidia did pretty well keeping up for awhile, so maybe they never intended to go open until it became apparent just how fast Linux moves.
We can argue all day about whether that should be true, but in the real world, it isn't. Linux is much, much farther along than any BSD save OS X, which does not contribute much of its BSD code back to the community, and often has to have its arm twisted to give back its GPL'd and LGPL'd stuff.
I suppose Solaris arguably comes close... but hey, Solaris is also GPL'd!
You could also argue that the BSD network stack has gone farther (as parts of it are apparently in Windows), but I don't think that is good for the community, or an effective way to promote open source. It's good for Microsoft, and for Windows, and for people who depend on Windows, but it ends there.