You need to educate yourself. Micro-kernels failed because of messaging overhead, otherwise known as "kernel crossings".
You need to educate YOURself. Old-style microkernels are doing well in the marketplace (e.g. NT), though they tend to perform poorly compared with monolithic kernels because they aren't "micro" enough. New-style microkernels (e.g. QNX, ChorusOS) are also doing well, though a few (e.g. BeOS) failed for business reasons.
What you call "messaging overhead" is, in a modern microkernel, basically just copying data between address spaces. This is no more "overhead" than you would find in a monolithic kernel, where data has to be copied between "user space" and "kernel space" and back again.
Or you may have meant the overhead of context switching. Yes, you're right. Context switching is expensive on an IA-32, because the chip flushes the TLB on every address space switch. This is a flaw in the x86 design, so you might say that monolithic kernels are better suited to Intel IA-32 CPUs, and I'd say you probably had a good case there. In defence of microkernels, though, I should note that a) there are ways around this (e.g. L4's small address space optimisation), and b) monolithic kernels like Linux are hitting this limitation too on 32-bit machines with lots of RAM (see, for example this patch).
That's true, but your implication that the original poster is living his life "in [...] terror of death" is based on a shaky assumption, namely, that mraymer has the same definition of "a good life" as you do.
I don't drink, for example. This is mostly because I don't enjoy it. (Admit it, you hated the taste of alcohol the first time you tried it. It's the world's most popular acquired taste.) However this has the fortunate side-effect that, because alcohol is expensive, not imbibing frees up funds to support my non-chemical habits.
I'm glad you find yourself having what you believe to be "a good life". However, I'm even more glad that I'm living my life and not yours. I'd probably find yours quite dull in places.
Actually, this reminds me of an Aesop fable. A dog and an ass were walking along the road. The ass finds a scroll, picks it up, and starts to read it. "Let's see now... location of buried treasure... answer to life the universe and everything..." The dog interrupts, "hang on, is there anything about meat or bones there?" The ass checks, and says "nope". The dog replies, "well throw it away then; it's useless!"
The difference is that the extinction of the stupid is entirely voluntary. Not feeding a starving child is cruel because chances are the child did not choose to starve. On the other hand, someone who drives like a lunatic is voluntarily removing themselves from the gene pool. Such an altruistic act of self-sacrifice is to be applauded indeed.
Web cacheing is apparently illegal in Australia too is you take one of the laws to it's logical conclusion.
If you meant proxying, not any more. This was fixed in the Copyright Amendment (Digital Agenda) Act, aka the "Australian DMCA". In the US, this was also fixed in the DMCA. No, the DMCA isn't all bad...
Well, that's just the sort of blinkered, phillistine pig-ignorance I've come to expect from you non-creative garbage. You sit there on your loathsome spotty behind, squeezing blackheads, not caring a tinker's cuss for a struggling artist, you EXCREMENT! You whining, hypocritical toady with your colour TV set and your Tony Jacklin golf clubs and your bleeding masonic secret handshakes. You wouldn't let me join, would you, you blackballing BASTARD! Well I wouldn't become a Freemason now if you got down on your lousy stinking knees and begged me!
The BIND patch and related things can only be a temporary measure, because Verisign will have the patch too, and be able to do something which works around it. Then BIND will work around that and so on.
Basically, you have a technological arms race, and an arms race is a race that nobody can possibly win. Legal recourse is handy for breaking the cycle.
Many ISPs over a certain size run more than one DNS server, at least one of which is a different kind than the others. That way, if an exploit is found in one kind of server, the other will still be available.
DNS is a sufficiently critical service that this is almost always worth it.
Re:Why does a professional have to clock in?
on
Are You On Time To Work?
·
· Score: 2, Insightful
Even when you're on a salary, your hours can be billable to a customer.
Even when I'm a salaried professional, I don't have a problem with "clocking in" and/or keeping a timesheet. But saying that 9:01am is late, no questions asked, is just plain wrong unless there's a good reason for it (e.g. customer support contracts requiring the phone/email to be manned during certain hours). Even then, though, I'd want it negotiable.
To see why, consider the word "beginning". (Yes, I used Genesis for my test purposes too.) In alphabetical order, that would be "begiinnng", but a more optimal word for compression is "beninging" because of the repeated three-letter sequence, which most compression algorithms find.
Huffman compression uses Huffman Coding based on the *entire* input string [...]
Well that's easy to implement. The input string will appear with probability 1, so it can be coded in 0 bits.
What you describe is character-based Huffman coding, where each octet is given its own Huffman code. There are many otehr kinds of "Huffman compression", such as word-based, where each English word (and inter-word space/punctuation combination) gets its own code.
While a random permutation for the middle letters is less compressible (for reasons mentioned by others; you can't compress noise), you should, in theory, be able to choose a set of permutations for the middle letters which maximises compressibility, and probably which does better than plain English alone.
The research programme:
Implement an algorithm which does this. You will need to pick some metric for compressibility. Choosing one that is independent of individual letter frequencies may be useful.
Here's the kicker: Are these non-random permutations more or less readable than random permutations? My guess is that they would be less readable, because the permutations themselves would start looking like other English words.
Re:Yes, a cat's got my tongue, OK?
on
Can You Raed Tihs?
·
· Score: 2, Interesting
Nit: Huffman coding is just a technique for taking a symbol alphabet with associated probability model and generating a minimal-entropy prefix-free binary code.
It is not a compression algorithm, though it often appears as the last step in a compression algorithm. In particular, it doesn't deal with the problem of how you generate the probability model, or what your symbol alphabet is.
The gzip algorithm, for example, uses "Huffman compression" just fine, but it still does poorly on scrambled text.
Seriously, Australia as a nation seems to have gotten most things right. The notable exception is the complete stupidity in the way we manage and legislate technology-related stuff, but that's just Senator Alston being his usual self.
Slashdot Mirror
You need to educate YOURself. Old-style microkernels are doing well in the marketplace (e.g. NT), though they tend to perform poorly compared with monolithic kernels because they aren't "micro" enough. New-style microkernels (e.g. QNX, ChorusOS) are also doing well, though a few (e.g. BeOS) failed for business reasons.
What you call "messaging overhead" is, in a modern microkernel, basically just copying data between address spaces. This is no more "overhead" than you would find in a monolithic kernel, where data has to be copied between "user space" and "kernel space" and back again.
Or you may have meant the overhead of context switching. Yes, you're right. Context switching is expensive on an IA-32, because the chip flushes the TLB on every address space switch. This is a flaw in the x86 design, so you might say that monolithic kernels are better suited to Intel IA-32 CPUs, and I'd say you probably had a good case there. In defence of microkernels, though, I should note that a) there are ways around this (e.g. L4's small address space optimisation), and b) monolithic kernels like Linux are hitting this limitation too on 32-bit machines with lots of RAM (see, for example this patch).
That's true, but your implication that the original poster is living his life "in [...] terror of death" is based on a shaky assumption, namely, that mraymer has the same definition of "a good life" as you do.
I don't drink, for example. This is mostly because I don't enjoy it. (Admit it, you hated the taste of alcohol the first time you tried it. It's the world's most popular acquired taste.) However this has the fortunate side-effect that, because alcohol is expensive, not imbibing frees up funds to support my non-chemical habits.
I'm glad you find yourself having what you believe to be "a good life". However, I'm even more glad that I'm living my life and not yours. I'd probably find yours quite dull in places.
Actually, this reminds me of an Aesop fable. A dog and an ass were walking along the road. The ass finds a scroll, picks it up, and starts to read it. "Let's see now... location of buried treasure... answer to life the universe and everything..." The dog interrupts, "hang on, is there anything about meat or bones there?" The ass checks, and says "nope". The dog replies, "well throw it away then; it's useless!"
Sounds like it: yes.
Is like it: no.
The difference is that the extinction of the stupid is entirely voluntary. Not feeding a starving child is cruel because chances are the child did not choose to starve. On the other hand, someone who drives like a lunatic is voluntarily removing themselves from the gene pool. Such an altruistic act of self-sacrifice is to be applauded indeed.
While the fiaso with Lucas and the DGA didn't exactly hurt his bank balance, there's a reason he didn't direct any movies between 1977 and 1999.
Geez, don't you know how to research anything?
Australia uses the British Monopoly board, so you would in fact be buying Mayfair.
I thought that was SCO. Or maybe the RIAA. Damn, who is it we hate this week?
Oh, man. I just got flashes of a paperclip with an eyebrow ring and a goatee.
Quick meta-question.
USB seems like a good technology for homebrew hardware. Does anyone know of a good way to interface USB to a cheap microcontroller?
If you meant proxying, not any more. This was fixed in the Copyright Amendment (Digital Agenda) Act, aka the "Australian DMCA". In the US, this was also fixed in the DMCA. No, the DMCA isn't all bad...
So? Freddie Mercury, Sir Mixalot and David St. Hubbins all seem to be in agreement that this isn't a bad thing necessarily.
Well, that's just the sort of blinkered, phillistine pig-ignorance I've come to expect from you non-creative garbage. You sit there on your loathsome spotty behind, squeezing blackheads, not caring a tinker's cuss for a struggling artist, you EXCREMENT! You whining, hypocritical toady with your colour TV set and your Tony Jacklin golf clubs and your bleeding masonic secret handshakes. You wouldn't let me join, would you, you blackballing BASTARD! Well I wouldn't become a Freemason now if you got down on your lousy stinking knees and begged me!
Sorry, I don't know what came over me.
The BIND patch and related things can only be a temporary measure, because Verisign will have the patch too, and be able to do something which works around it. Then BIND will work around that and so on.
Basically, you have a technological arms race, and an arms race is a race that nobody can possibly win. Legal recourse is handy for breaking the cycle.
Many ISPs over a certain size run more than one DNS server, at least one of which is a different kind than the others. That way, if an exploit is found in one kind of server, the other will still be available.
DNS is a sufficiently critical service that this is almost always worth it.
Even when you're on a salary, your hours can be billable to a customer.
Even when I'm a salaried professional, I don't have a problem with "clocking in" and/or keeping a timesheet. But saying that 9:01am is late, no questions asked, is just plain wrong unless there's a good reason for it (e.g. customer support contracts requiring the phone/email to be manned during certain hours). Even then, though, I'd want it negotiable.
Tried that. Doesn't help.
To see why, consider the word "beginning". (Yes, I used Genesis for my test purposes too.) In alphabetical order, that would be "begiinnng", but a more optimal word for compression is "beninging" because of the repeated three-letter sequence, which most compression algorithms find.
Well that's easy to implement. The input string will appear with probability 1, so it can be coded in 0 bits.
What you describe is character-based Huffman coding, where each octet is given its own Huffman code. There are many otehr kinds of "Huffman compression", such as word-based, where each English word (and inter-word space/punctuation combination) gets its own code.
While a random permutation for the middle letters is less compressible (for reasons mentioned by others; you can't compress noise), you should, in theory, be able to choose a set of permutations for the middle letters which maximises compressibility, and probably which does better than plain English alone.
The research programme:
Nit: Huffman coding is just a technique for taking a symbol alphabet with associated probability model and generating a minimal-entropy prefix-free binary code.
It is not a compression algorithm, though it often appears as the last step in a compression algorithm. In particular, it doesn't deal with the problem of how you generate the probability model, or what your symbol alphabet is.
The gzip algorithm, for example, uses "Huffman compression" just fine, but it still does poorly on scrambled text.
I would sue people who stand in water!
Ooooooh!
No, there are different politics.
Forget Troll Tech, he's a director of Lineo!
Now who do you think bought the SCO licences?
It's just you.
Seriously, Australia as a nation seems to have gotten most things right. The notable exception is the complete stupidity in the way we manage and legislate technology-related stuff, but that's just Senator Alston being his usual self.
Fine, so a religion degree would help qualify you as an editor of a religious affairs magazine or teacher of comparative religions.
How useful is a degree in mediaeval history? Or pure mathematics?
So? I've been quitting programs for a decade or so using the "File" menu. Since when has quitting a program been a file operation?
The semantics of "Start" is that to do anything, you "start here". That actually makes more sense to me than putting Quit under the File menu.