I can give an approximate for this. It really all depends on the situation, but if we're talking about a VLSI design of some sort, reverse-engineering is usually impractical compared to reimplementation. It can take up to 100 times longer to reverse-engineer something, and, at the lowest level, the way things are designed in these sorts of systems is pretty well-defined (Eg. there are only so many ways you can make a floating-point multiplier faster). As for reverseing it from specs, that's simply not going to happen.
The only thing that having somebody else's specs does is let you see how, in high-level terms, they made it work. But nowadays, most of this is publicized (eg. nVidia improved memory bandwidth on the GeForce3 with a crossbar memory controller). So all that it boils down to is the ability to make cards that are 100% compatible, but not identical.
But any engineer will tell you that there are always things they would like to do different. So, while a new feature ("framebuffer blit engine" for older video features, for exxample) would be cloned by everybody if the word of it existing got out (and it got out fast, and was subsequently cloned, so now virtually all video cards have one of these), the implementations will almost always be different.
And it's always easier to hire 5 engineers than 1 good reverse-engineer, and reverseing takes a lot more effort (around 100x for some cases, meaning more staff).
I should point out here that reverse-engineering software (IE the drivers) is actually much simpler than hardware, although still challenging at best. In most cases, this is what happens, the actual competitors dismantle the drivers to see how it works, so really it doesn't protect them any.
Well, it doesn't clip as in "High-pass filter". It's because there's so little data in the low-frequency subband, it has only a few types of *BOOM* noises. I'm personally a fan of just about everything electronic, so my stuff pushes a lot of base too.
192kbps MP3's are a lot better than 128 for this, so that's most of your experience. Still stack a 128kbit MP3 against a comparable Vorbis file. The difference is noticable.
Interesting. Personally I've noticed that my equalizer settings make more difference to sound quality than compression, unless it's a poorly-encoded file.
The point is that humans only hear a certain range of frequencies, but not only that, they only hear prominent frequencies. Vorbis took advantage of this moreso than MP3 by implementing a Bark-scale to differentiate the frequency subbands, instead of an arbitrary scale.
The point is that, if you hear a note at 400Hz at 50dB and one at 401Hz at 2dB, you won't notice the second. Your ear just cannot receive it. Lossiness comes in when this "masking" effect (where a frequency blocks out one of a similar frequency but lower amplitude) does not mask the noise introduced by the frequency quantization. The more advanced the psychoaccoustics, the more of this noise will fall below the "masking threshold". Some people have better masking thresholds than others.
The salient point here is that the more data you throw out, the harder it becomes to keep a perfect SnR (by the human ear that is, you obviously lose a lot of actual signal). The 128kbit is probably going to stay the standard, period, since bandwidth is getting so much cheaper. With improving psychoaccoustical models, the quantization noise should fall well below the masking thresholds of even the most astute listeners.
Without more sophisticated pickups (IE something better than human ears), files with some compression will eventually sound identical from the listeners perspective.
I encoded a few of my songs into Vorbis and really liked the results. Not using a low-bitrate option or any such nonsense, of course... I set it for the compression rate that uses just more than 128kbps, so it was a smidge bigger than an mp3...
And suddenly, the floor started shaking, noticably. You see, MP3 seems to clip anything below 60Hz. Human hearing goes down to 40Hz normally, 20Hz for some people, but you can still detect a 10Hz waveform... Vorbis allocates much more data to the low-frequency end of the spectra than MP3, as was evidenced by the much higher utilization of my subwoofer.
Too bad the specification is still lacking... Vorbis is a nice format and I've been trying to write a player. Hard to do it all from sources.
The chest structure is identical, the hand structures are identical, etc... But the thumb on the right hand is in a different position. The face is in a different position as well...
It's also very revealing to look at it in a skewed aspect ratio (my monitor is 2:1 in high-res, not 4:3), at which point hilighting like in regular CG starts showing up. These don't show up in normal images. The skintone is VERY different, showing much more natural tone with blemishes in the "harsh-lit" version, which are lost in the "fuzzy" lighting. I doubt those could be reconstructed.
I'm having quite a job believing that this is anything other than CG straight out of the machines that did the movie.
I think at this point I have yet to find a movie that Kats and I agreed on, to the point where I actually can count on any movie that Katz HATES to be GREAT.
First of all, I don't think Katz has EVER seen any anime. This movie honestly reminds me of Akira, and who wants to bet he'd hate that one too? Anybody who thinks that this is not an anime should be slapped.
Personally I'm a little tired of the Japanese mystical plots, they start looking like a formula after a while (did anybody catch the "protoculture"-likeness of their energy sources?) but other than that this was a great movie.
And what's with the complaint of insufficient action? Last I checked, Final Fantasy wasn't a 3rd-person shooter. I think he'd get bored with the turn-based combat in the real games (comparing it to Tomb Raider ?!?!?!?)
The first Dune game was made by a company named Cryo and distributed by Virgin (if I remember correctly), not by Westwood. Dune 2 was the original realtime strategy, but Dune 1 was a very different, interesting game, albeit a little old.
I don't ever remember slicing myself on Meccano, but then again, I had one of the later sets that looked galvanized or chromed (or painted, for a few of the larger pieces). Mine aren't rusty yet.
As for what I did as a kid myself (and therefore, what I swear by), my favorite toys encompassed the following:
Meccano - I had a fairly large set with a motor, but the gears never meshed very well, even in the designs they listed.
Construx - Now THIS one was fun, especially since the holes in the interconnecting blocks were the same diameter as the screws from the meccano set. I ended up making a few really weird hybrid designs. The other nice thing about it was that it could make very large things... One year I built a (lifesize) model of Johnny 5 (from Short Circuit) that could move back and forth and blink lots of lights. This set could also, if you were careful, mount the radio-shack project kits.
Radio Shack Project Kits - The electronics kits. Too bad these don't exist in Canada anymore. I had 4 or 5 of these things, and the 200 even had some logic circuits. Fun devices. I got my first one at age 7.
My Commodore 128 - Yup, nice machine. Great for hacking, I turned the joystick port into an output port one year, and ran a robot (Construx with relays and transistors breadboarded to the top) under computer control. I must have been about 11 at the time.
Breadboards + 7400-series logic - Got some of these when I was 11, learned lots about digital circuits, got me well-started in digital design. I surprised the community-college instructor when I showed him my design for an adder circuit (the standard design, of course, but I figured it out;)
I remember that I thought Lego technic was really cool, but normal lego was pretty restrictive. I also had a set called Robotix, which was really nice for building complicated robotics with, but pretty limited in other ways. Bah, I did manage to build a pick-and-place out of it...
Now I'm taking engineering, and everybody is like "How do you know so much?". Bah, when they were out playing with their dolls (or whatever the guys were doing, since there ARE a lot more of them in engineering), I was building stages for mine with working amplified sounds and lighting systems;)
Anybody who thinks that kids don't learn from their toys should think again.
I'm just wondering what I'm missing. Is this the "Acer" that makes DVD-ROM drives? I'm confused.
And Asus? Huh? They're not worried about the dangers of K7 motherboards, but an MPEG-2 decoder chip is dangerous? But an Athlon can merrily perform all the same functions, if not a bit more!
I think these "sources" sound like a couple teenage kids with a little too much [favorite drug here] on their brains.
Wouldn't it be more dangerous if China got ahold of just one or two well-trained EE's from around here?
No, it really IS the case that you can't trademark a SINGLE dictionary word, or else we would lose our language. Now, you _can_ trademark strings.
For example, I cannot trademark either "Troll" or "shredder", but I could certainly trademark "The Troll Shredder", since it's more than one word. Not to mention it sounds like a handy product (although I can think of other words besides "shredder" I'd like to use).
You know, you have to run a trademark check when you register one, to make sure it's valid and all... Part of that is that you cannot trademark dictionary words. So this script may help:
grep -i WORD/usr/dict/words
So for illustrator:
grep -i Illustrator/usr/dict/words
You better believe it finds something.
The trick here is that they really DID trademark "Adobe Illustrator". Interestingly, they actually own trademarks neither on "adobe" (which is a type of brick) or "illustrator" (which is a person who draws), but on "Adobe Corporation" (or similar) and "Adobe Illustrator".
Illustrator might not be as common a word as, say, "Office", but since it is a single, dictionary word (not a hybrid like PrintWriter, or somesuch), the word itself may be used in names, even of competing products (or, at least, it's true here in Canada).
A cute example is "Draw". If you send somebody a "Draw file", you're talking Corel Draw. Why hasn't Corel sued anybody over "Microsoft Draw" (I seem to recall such a product), or similar? Same reason.
I think I'll patent the dictionary-word trademark.
I thought I heard that Ghostscript had a PDF interpreter recently (PDF and PS are related formats, they're quite similar I recall).
And I _KNOW_ that there's a windows version of it, although it's a bit tricky to install. I had to use it once because, well, standard windows doesn't have a thing that reads postscript files.
This is actually in reference to the C++ template facility, not the old carry-over C preprocessor that is left in the C++ environment. The template system can actually do some looping if you're careful, it has to do something with using recursive template calls with an end-cap, similar to how you do loops in M4.
I saw a paper a while back on numerical computing (I tried to look it up, but alas, I don't remember enough of it) that used these kinds of tricks to do all kinds of neat numerical analysis while compiling. Strange use of templates.
Slashdot Mirror
The only thing that having somebody else's specs does is let you see how, in high-level terms, they made it work. But nowadays, most of this is publicized (eg. nVidia improved memory bandwidth on the GeForce3 with a crossbar memory controller). So all that it boils down to is the ability to make cards that are 100% compatible, but not identical.
But any engineer will tell you that there are always things they would like to do different. So, while a new feature ("framebuffer blit engine" for older video features, for exxample) would be cloned by everybody if the word of it existing got out (and it got out fast, and was subsequently cloned, so now virtually all video cards have one of these), the implementations will almost always be different.
And it's always easier to hire 5 engineers than 1 good reverse-engineer, and reverseing takes a lot more effort (around 100x for some cases, meaning more staff).
I should point out here that reverse-engineering software (IE the drivers) is actually much simpler than hardware, although still challenging at best. In most cases, this is what happens, the actual competitors dismantle the drivers to see how it works, so really it doesn't protect them any.
192kbps MP3's are a lot better than 128 for this, so that's most of your experience. Still stack a 128kbit MP3 against a comparable Vorbis file. The difference is noticable.
The point is that humans only hear a certain range of frequencies, but not only that, they only hear prominent frequencies. Vorbis took advantage of this moreso than MP3 by implementing a Bark-scale to differentiate the frequency subbands, instead of an arbitrary scale.
The point is that, if you hear a note at 400Hz at 50dB and one at 401Hz at 2dB, you won't notice the second. Your ear just cannot receive it. Lossiness comes in when this "masking" effect (where a frequency blocks out one of a similar frequency but lower amplitude) does not mask the noise introduced by the frequency quantization. The more advanced the psychoaccoustics, the more of this noise will fall below the "masking threshold". Some people have better masking thresholds than others.
The salient point here is that the more data you throw out, the harder it becomes to keep a perfect SnR (by the human ear that is, you obviously lose a lot of actual signal). The 128kbit is probably going to stay the standard, period, since bandwidth is getting so much cheaper. With improving psychoaccoustical models, the quantization noise should fall well below the masking thresholds of even the most astute listeners.
Without more sophisticated pickups (IE something better than human ears), files with some compression will eventually sound identical from the listeners perspective.
And suddenly, the floor started shaking, noticably. You see, MP3 seems to clip anything below 60Hz. Human hearing goes down to 40Hz normally, 20Hz for some people, but you can still detect a 10Hz waveform... Vorbis allocates much more data to the low-frequency end of the spectra than MP3, as was evidenced by the much higher utilization of my subwoofer.
Too bad the specification is still lacking... Vorbis is a nice format and I've been trying to write a player. Hard to do it all from sources.
It's also very revealing to look at it in a skewed aspect ratio (my monitor is 2:1 in high-res, not 4:3), at which point hilighting like in regular CG starts showing up. These don't show up in normal images. The skintone is VERY different, showing much more natural tone with blemishes in the "harsh-lit" version, which are lost in the "fuzzy" lighting. I doubt those could be reconstructed.
I'm having quite a job believing that this is anything other than CG straight out of the machines that did the movie.
First of all, I don't think Katz has EVER seen any anime. This movie honestly reminds me of Akira, and who wants to bet he'd hate that one too? Anybody who thinks that this is not an anime should be slapped.
Personally I'm a little tired of the Japanese mystical plots, they start looking like a formula after a while (did anybody catch the "protoculture"-likeness of their energy sources?) but other than that this was a great movie.
And what's with the complaint of insufficient action? Last I checked, Final Fantasy wasn't a 3rd-person shooter. I think he'd get bored with the turn-based combat in the real games (comparing it to Tomb Raider ?!?!?!?)
As for what I did as a kid myself (and therefore, what I swear by), my favorite toys encompassed the following:
- Meccano - I had a fairly large set with a motor, but the gears never meshed very well, even in the designs they listed.
- Construx - Now THIS one was fun, especially since the holes in the interconnecting blocks were the same diameter as the screws from the meccano set. I ended up making a few really weird hybrid designs. The other nice thing about it was that it could make very large things... One year I built a (lifesize) model of Johnny 5 (from Short Circuit) that could move back and forth and blink lots of lights. This set could also, if you were careful, mount the radio-shack project kits.
- Radio Shack Project Kits - The electronics kits. Too bad these don't exist in Canada anymore. I had 4 or 5 of these things, and the 200 even had some logic circuits. Fun devices. I got my first one at age 7.
- My Commodore 128 - Yup, nice machine. Great for hacking, I turned the joystick port into an output port one year, and ran a robot (Construx with relays and transistors breadboarded to the top) under computer control. I must have been about 11 at the time.
- Breadboards + 7400-series logic - Got some of these when I was 11, learned lots about digital circuits, got me well-started in digital design. I surprised the community-college instructor when I showed him my design for an adder circuit (the standard design, of course, but I figured it out
;)
I remember that I thought Lego technic was really cool, but normal lego was pretty restrictive. I also had a set called Robotix, which was really nice for building complicated robotics with, but pretty limited in other ways. Bah, I did manage to build a pick-and-place out of it...Now I'm taking engineering, and everybody is like "How do you know so much?". Bah, when they were out playing with their dolls (or whatever the guys were doing, since there ARE a lot more of them in engineering), I was building stages for mine with working amplified sounds and lighting systems ;)
Anybody who thinks that kids don't learn from their toys should think again.
And Asus? Huh? They're not worried about the dangers of K7 motherboards, but an MPEG-2 decoder chip is dangerous? But an Athlon can merrily perform all the same functions, if not a bit more!
I think these "sources" sound like a couple teenage kids with a little too much [favorite drug here] on their brains.
Wouldn't it be more dangerous if China got ahold of just one or two well-trained EE's from around here?
For example, I cannot trademark either "Troll" or "shredder", but I could certainly trademark "The Troll Shredder", since it's more than one word. Not to mention it sounds like a handy product (although I can think of other words besides "shredder" I'd like to use).
grep -i WORD /usr/dict/words
So for illustrator:
grep -i Illustrator /usr/dict/words
You better believe it finds something.
The trick here is that they really DID trademark "Adobe Illustrator". Interestingly, they actually own trademarks neither on "adobe" (which is a type of brick) or "illustrator" (which is a person who draws), but on "Adobe Corporation" (or similar) and "Adobe Illustrator".
Illustrator might not be as common a word as, say, "Office", but since it is a single, dictionary word (not a hybrid like PrintWriter, or somesuch), the word itself may be used in names, even of competing products (or, at least, it's true here in Canada).
A cute example is "Draw". If you send somebody a "Draw file", you're talking Corel Draw. Why hasn't Corel sued anybody over "Microsoft Draw" (I seem to recall such a product), or similar? Same reason.
I think I'll patent the dictionary-word trademark.
And I _KNOW_ that there's a windows version of it, although it's a bit tricky to install. I had to use it once because, well, standard windows doesn't have a thing that reads postscript files.
I saw a paper a while back on numerical computing (I tried to look it up, but alas, I don't remember enough of it) that used these kinds of tricks to do all kinds of neat numerical analysis while compiling. Strange use of templates.