$3350 is peanuts for a motivated teenager. My son raised about twice this amount by the time he was 16 to finance a year-long exchange trip to Germany during his junior year in high school. His job: walking dogs.
The tools used to make professional movies, even computer-animated ones, are mostly not software. The talent and industry of the artists that make the movies are far more important. The idea that cheap software will allow just anybody to make professional animation is ludicrous on its face. That's just not the reason that making professional movies is expensive.
The cost of software is a drop in the bucket compared to salaries for 250 people working for 4 years (roughly what Pixar or PDI or Disney deploys.)
Here's a good guess at the numbers:
Operating system software for 250 computer animators at $1,000 per station is about $0.25 million. Loaded salaries (i.e. salaries plus benefits plus pro-rata overhead) of 250 people for 4 years, at $200,000 per year, is $200 million. Nobody in the business cares what the system software costs. (In fact, the places that don't write their own application software in-house don't even care what that costs -- $25,000 per workstation for Maya and Renderman is still peanuts.)
Linux is taking over because the previous market leader was SGI, not Microsoft. The programmers that write the software are naturally less interested in switching to Win32 than Linux, when presented with the inevitability of an Intel future. If Microsoft were a real player in this market (and, oh do they wish they were!) the outcome would not be so clear.
The traditional way to deal with this problem is to distribute binaries as well as source code, and only support customers who can reproduce their complaints on verifiably original binaries. That's what everybody used to do back before the closed-source era. (Yes, that's right, open source came first -- as late as the 1970s IBM was still distributing source code to their operating systems free to all users. Software was viewed as a loss-leader for the hardware. Only with the advent of `unbundling', the separate sale of hardware and software, implemented to satisfy anti-trust complaints, did closed-source become a common feature of the software business.)
The first multiprocessor BSD was running in 1981, a decade or more before the first SMP Linux. (The reference mostly describes the hardware, which was an even better hack than the software.)
Politicians are preferentially drawn from the bullying classes, and so see this as a problem that won't go away -- it's how people like them naturally behave, so obviously there's nothing to be done.
This is obviously a self-reinforcing situation. As long as we allow ourselves to be led by bullies, nobody will do anything about bullying. This will go on until we *really* start taking character into account (rather than just the demagogic pot vs. kettle name-calling that goes on these days) when we choose our representatives.
Back in the late 1970s I worked at the NYIT Computer Graphics laboratory on Long Island. At the time the Graphics Lab was among the world's best CG research facilities (its founding director, who I still work for, is now president of Pixar.)
The main focus of our research was a system for doing cartoon animation. In 1978 or '79 Bill Hanna, then approaching 70 years old, came out to see what we were up to, and seeing exciting possibilities for his cost-driven business, put together a team led by Mark Levoy, a Cornell graduate student (who I believe we recommended to him, and who is now a CS professor at Stanford) to build what would be the first commercially successful computer animation system, which Hanna-Barbera used to produce The Flintstones, Scooby Doo, etc. throughout the '80s and '90s.
> Now if you use < you have to hit up arrow,
> and do some retyping, cutting and pasting..
>but with cat you can hit up arrow,
>move to the right place, and type frob
> and a pipe!
Redirection doesn't have to go at the end
of the arg list. You're allowed to say
<file qrpff | extract_mpeg2 | mpeg2_dec
Now interpolating something before qrpff
is as simple as point + type.
one can replace "speciation" and its relatives with "correct predictions" and its ilk; and "evolution" with "astrology", and the last paragraph will be just as sensical as it is now.
Of course, because that paragraph deals with falsifiability, not with truth. The difference between astrology and natural selection is not whether they're falsifiable, but that when we try to falsify one we succeed, and with the other we fail.
I'm not particularly interested in arguing the truth or falsity of evolution. Specialists are better equipped for that. But it's pretty obviously a falsifiable theory, since there is pretty obviously possible evidence that would contradict it.
Due to the absurd length of time for natural selection (speciation, not peppered-moth-style adaptation) to run its course, predictions based on the theory of natural selection are not possible, and therefore it is not falsifiable.
It's a mistake to think that because some particular experiment can't be done that there is no way to test a theory. Instead of sitting and waiting for speciation to happen, we can also, for example, examine the fossil record for evidence of the ancestry of current species. When we do, what do we see? Zillions of species, appearing and dying, the new obviously related to the old, all arranged in neat cladographic hierarchies.
I should point out that just sitting and waiting for speciation is a perfectly good test of the theory. Every day that goes by without a new species appearing whittles away infinitesimally at the credibility of speciation by natural selection. If it goes on for long enough, the theory will be discarded.
(Of course, as soon as a new species appears,
all the whittling will be instantly undone.)
It's unfortunate for you and me that `long enough' is a good deal longer than our tenure on this planet, but the theory doesn't care. Some experiments run longer than others. It's a shame that you and I will never get to read the paper, but it has no bearing on the falsifiability of the theory.
This is just nonsense. There are plenty of observations that would falsify evolution. For example:
Absence of morphologically related species.
Dissimilarity of DNA of morphologically similar species.
Absence of correlation of fossil forms with time in the fossil record.
No divergence of geographically isolated populations.
Book of Genesis coded in ASCII in human introns
...
If we observed any of these things it would lead us to believe that evolution was a bad theory. That is what it means to be falsifiable.
That none of these falsifiers is true leads us to believe the theory.
I for one find it very telling that Linux started in Finland.
Linux started in New Jersey under a different name and eventually moved to Berkeley. A clone appeared Holland (under the aegis of an expatriate America) and eventually found its way to Finland.
Doug's, Flint's and Everett & Jones are all good, but if you're willing to go up San Pablo to Richmond, Bobby's Backdoor Cajun BBQ is unparalleled. Another local Southern highlight is Roscoe's Chicken & Waffles (only 2 things on the menu!) near Lake Merritt. (Roscoe's was better before they moved, but still first rate.)
anticypher didn't mention the BBQ goat and turkey at Doug's, both of which are excellent.
When Pascal replaces C, how do I convert my C functions into Pascal functions? Eh?
When that happens, I'll use the concommitant
flying pigs and infernal snowballs to solve all my programming problems.
Genesis 1:1 "In the beginning God created the heavens and the earth."
Before you get your flamethrowers in a bunch trying to hose me for
being an idiot, I suggest you consider the evidence of specific creation
based on the concept of intelligent design.
Henry Massalin's PhD thesis `Synthesis: An
Efficient Implementation of Fundamental
Operating System Services' has all kinds
of good stuff about high-performance system
software, including fairly cool lock-free
synchronization ideas.
I can't find a url for it -- Columbia University
has apparently dumped Henry's pages from their
server.
>Among other things, the Copyright notice
>says that ``This book shall not... be lent
>out...''.
This quote out of context is misleading.
The actual condition is that the book
not be sold or lent out with a different
cover than the one provided by the publisher.
The reason for this proviso is that the standard
way for retailers to return unsold paperbacks
to their publishers for credit is to tear the
covers off and send them back, destroying the
books, thus saving substantial postage. Whenever
you buy or borrow a used book that is missing
its cover, the publishers and authors have not
been payed for the book, and either the retailer
has misappropriated it or someone has picked
through trash that ought to have been destroyed.
They claim to be able to fit 100 128K bit/sec
streams into an FM radio channel. The bandwidth
of an FM channel is 100 KHz, not 200 KHz as stated
in the article. Nevertheless, we'll give them
the benefit of the doubt.
The information-carrying capacity (C) of a channel
is a function of its bandwidth B and signal (S)
to noise (N) ratio. The formula (this is
elementary information theory) is
C=B log2(1+S/N)
We can plug in the given values for B (200 KHz)
and C (12.8M bit/sec) and determine the required
signal to noise ratio:
12.8e6=2e5 log2(1+S/N)
log2(1+S/N)=64
S/N=2^64 - 1
Each bit of signal-to-noise ratio corresponds
to 3 decibels (6 dB if you're talking about
power, which I won't, just to be charitable),
so the required S/N ratio is around 190 dB.
Is this achievable? In a word, no. It's not
even close. They're off by more than 100 dB.
When these guys claim that `when properly implemented and used under the right
conditions, the digital modulation scheme reportedly delivers 90 bps/Hz-over the air'
they're blowing smoke. They're saying they
need a 270 dB S/N ratio. (That's larger than
my number because I gave them some extra slack.)
If they could get it,
their scheme would work, but they can't, so it
won't.
>a quarter of a million americans >are killed in car accidents a year.
No, 42,000 people are killed in car accidents in the US every year. Unless there's a lot of Americans being killed abroad, you're off by a factor of 6. 250,000 americans suffer serious injuries in crashes every year, but only 1 in 6 dies. Reference
Uhh, Plan 9 is an operating system, not an application -- there's no chance that it would appear in anybody's process tables.
I just grepped my copy, and the string `plan9.conf' appears nowhere in the Plan 9 distribution.
The `original incantation' [sic] of Plan 9 was Plan 9 -- it was a new system from the ground up, developed in the Computing Science Research Center at Bell Labs, not by any bunch of Deccies working spare time. (Dave Presotto, largely responsible for Plan 9 networking, worked at DEC once, but that was before he went to grad school, I believe.)
I speak having worked on Plan 9 at Bell Labs for 6 years (and I have the Bowling Shirt to prove it.) Among other things, I wrote the shell and a whole load of graphics junk.
His Daddy also works for the NSA (how do you think he knew enough about the flaws in sendmail etc... that made the WORM work?)
This is pretty low. I know rtm and (slightly) his father, rhm. rtm is smart enough to find holes in sendmail on his own, and his father (who is now retired, BTW) is not the sort that would leak NSA information, even to his son. Have you ever read his report on breaking into 4.2BSD TCP/IP, written BEFORE his father went to work at NSA? (The funniest part of that research was the machine at Bell Labs named ucbvax, used to exploit a backdoor.)
This post gives the impression that 18 was universally the age of majority starting at the Big Bang up to the Disco Era. In fact, reduction of the age of majority from 21 to 18, and the accompanying abrogation of the in loco parentis doctrine by colleges happened during the Nixon administration (in 1971, when I was 18!), when the government decided that if US citizens at age 18 were old enough to send off to Viet Nam to die, then they were certainly old enough to vote (and drink, and take responsibility for their own sex lives.)
The retrenchment back to considering 21 the age of majority, at least for drinking purposes, happened gradually over the next few years. But really, 18 year olds were only legally considered adults for a very short period of time. (Except for voting, which would require repealing the 26th amendment.)
There is no news for computer science professionals in this review, nor apparently in the book reviewed. Furthermore, the review alleges professional ignorance, laziness or hubris, contrary to fact. If the reviewer had seen fit to ask a couple of professionals about his allegations (this is standard reportorial conduct) he would have known that they aren't true. The review should be marked Editorial lest someone confuse it for actual reportage.
First of all, this quote is a real laugh:
We now know that not all algorithmic problems are
solvable by computers, even with unlimited
access to resources like time and memory
space.
The implication is that the realization that there are things that computers can't do is somehow new or unexpected.
In fact, the very first computer science paper ever written (Alan Turing: "On Computable Numbers, with an Application to the Entsheidungsproblem" Proc. London Math. Soc. 42, 230-65, 1937) is about problems that have a definite, mathematical answer, but whose solution by a computer is mathematically, intrinsically impossible.
So, for more than sixty years, a good five or ten years before the first computers were built (depending on whose definition you accept) we already knew about problems that they couldn't solve.
Harel's book is not one of the first
books in recent memory that focuses on the
limits of computers. Any textbook on the theory of computation concentrates on this. The classic of the field is Computability and Unsolvability by Martin Davis, first published in 1958, which set the standard for how computing theory is taught. If you suffered through a theory course based on Turing Machines, it's Martin's fault. His book gets to its first proof that a problem is uncomputable half way down the third page of the introduction, before Chapter 1 even starts!
The idea that few people, even with advance [sic] computer science degrees understand these problems is laughable. Anybody with an advanced computer science degree from a major institution has sat through a theory course in which undecidable problems are broached within the first few weeks, and an algorithms course in which intractable problems (those that aren't impossible, but for which no feasibly fast algorithm exists) are a major topic.
>Or they must be at least >writing a compiler for it...
Steve Johnson (who wrote yacc and pcc) works there, so they probably have that one in hand.
Note: back at Bell Labs, Steve was Dave Ditzel's boss for a while when Ditzel and Rae Maclellan were designing the Crisp microprocessor, whose branch-folding scheme smacks of the sort of code-rewriting that the most interesting Transmeta rumors feature.
Another former Labbie at Transmeta is Pat Parseghian.
Slashdot Mirror
$3350 is peanuts for a motivated teenager. My son raised about twice this amount by the time he was 16 to finance a year-long exchange trip to Germany during his junior year in high school. His job: walking dogs.
The cost of software is a drop in the bucket compared to salaries for 250 people working for 4 years (roughly what Pixar or PDI or Disney deploys.) Here's a good guess at the numbers:
Operating system software for 250 computer animators at $1,000 per station is about $0.25 million. Loaded salaries (i.e. salaries plus benefits plus pro-rata overhead) of 250 people for 4 years, at $200,000 per year, is $200 million. Nobody in the business cares what the system software costs. (In fact, the places that don't write their own application software in-house don't even care what that costs -- $25,000 per workstation for Maya and Renderman is still peanuts.)
Linux is taking over because the previous market leader was SGI, not Microsoft. The programmers that write the software are naturally less interested in switching to Win32 than Linux, when presented with the inevitability of an Intel future. If Microsoft were a real player in this market (and, oh do they wish they were!) the outcome would not be so clear.
The traditional way to deal with this problem is to distribute binaries as well as source code, and only support customers who can reproduce their complaints on verifiably original binaries. That's what everybody used to do back before the closed-source era. (Yes, that's right, open source came first -- as late as the 1970s IBM was still distributing source code to their operating systems free to all users. Software was viewed as a loss-leader for the hardware. Only with the advent of `unbundling', the separate sale of hardware and software, implemented to satisfy anti-trust complaints, did closed-source become a common feature of the software business.)
The first multiprocessor BSD was running in 1981, a decade or more before the first SMP Linux. (The reference mostly describes the hardware, which was an even better hack than the software.)
Politicians are preferentially drawn from the bullying classes, and so see this as a problem that won't go away -- it's how people like them naturally behave, so obviously there's nothing to be done.
This is obviously a self-reinforcing situation. As long as we allow ourselves to be led by bullies, nobody will do anything about bullying. This will go on until we *really* start taking character into account (rather than just the demagogic pot vs. kettle name-calling that goes on these days) when we choose our representatives.
Back in the late 1970s I worked at the NYIT Computer Graphics laboratory on Long Island. At the time the Graphics Lab was among the world's best CG research facilities (its founding director, who I still work for, is now president of Pixar.)
The main focus of our research was a system for doing cartoon animation. In 1978 or '79 Bill Hanna, then approaching 70 years old, came out to see what we were up to, and seeing exciting possibilities for his cost-driven business, put together a team led by Mark Levoy, a Cornell graduate student (who I believe we recommended to him, and who is now a CS professor at Stanford) to build what would be the first commercially successful computer animation system, which Hanna-Barbera used to produce The Flintstones, Scooby Doo, etc. throughout the '80s and '90s.
> Now if you use < you have to hit up arrow,
> and do some retyping, cutting and pasting..
>but with cat you can hit up arrow,
>move to the right place, and type frob
> and a pipe!
Redirection doesn't have to go at the end
of the arg list. You're allowed to say
<file qrpff | extract_mpeg2 | mpeg2_dec
Now interpolating something before qrpff
is as simple as point + type.
Of course, because that paragraph deals with falsifiability, not with truth. The difference between astrology and natural selection is not whether they're falsifiable, but that when we try to falsify one we succeed, and with the other we fail.
I'm not particularly interested in arguing the truth or falsity of evolution. Specialists are better equipped for that. But it's pretty obviously a falsifiable theory, since there is pretty obviously possible evidence that would contradict it.
It's a mistake to think that because some particular experiment can't be done that there is no way to test a theory. Instead of sitting and waiting for speciation to happen, we can also, for example, examine the fossil record for evidence of the ancestry of current species. When we do, what do we see? Zillions of species, appearing and dying, the new obviously related to the old, all arranged in neat cladographic hierarchies.
I should point out that just sitting and waiting for speciation is a perfectly good test of the theory. Every day that goes by without a new species appearing whittles away infinitesimally at the credibility of speciation by natural selection. If it goes on for long enough, the theory will be discarded. (Of course, as soon as a new species appears, all the whittling will be instantly undone.) It's unfortunate for you and me that `long enough' is a good deal longer than our tenure on this planet, but the theory doesn't care. Some experiments run longer than others. It's a shame that you and I will never get to read the paper, but it has no bearing on the falsifiability of the theory.
This is just nonsense. There are plenty of observations that would falsify evolution. For example:
If we observed any of these things it would lead us to believe that evolution was a bad theory. That is what it means to be falsifiable. That none of these falsifiers is true leads us to believe the theory.
Linux started in New Jersey under a different name and eventually moved to Berkeley. A clone appeared Holland (under the aegis of an expatriate America) and eventually found its way to Finland.
Doug's, Flint's and Everett & Jones are all good, but if you're willing to go up San Pablo to Richmond, Bobby's Backdoor Cajun BBQ is unparalleled. Another local Southern highlight is Roscoe's Chicken & Waffles (only 2 things on the menu!) near Lake Merritt. (Roscoe's was better before they moved, but still first rate.)
anticypher didn't mention the BBQ goat and turkey at Doug's, both of which are excellent.
When Pascal replaces C, how do I convert my C functions into Pascal functions? Eh?
When that happens, I'll use the concommitant flying pigs and infernal snowballs to solve all my programming problems.
Do you not see how this begs the question ?
Henry Massalin's PhD thesis `Synthesis: An
Efficient Implementation of Fundamental
Operating System Services' has all kinds
of good stuff about high-performance system
software, including fairly cool lock-free
synchronization ideas.
I can't find a url for it -- Columbia University
has apparently dumped Henry's pages from their
server.
>Among other things, the Copyright notice ... be lent
...''.
>says that ``This book shall not
>out
This quote out of context is misleading.
The actual condition is that the book
not be sold or lent out with a different
cover than the one provided by the publisher.
The reason for this proviso is that the standard
way for retailers to return unsold paperbacks
to their publishers for credit is to tear the
covers off and send them back, destroying the
books, thus saving substantial postage. Whenever
you buy or borrow a used book that is missing
its cover, the publishers and authors have not
been payed for the book, and either the retailer
has misappropriated it or someone has picked
through trash that ought to have been destroyed.
The information-carrying capacity (C) of a channel is a function of its bandwidth B and signal (S) to noise (N) ratio. The formula (this is elementary information theory) is
C=B log2(1+S/N)
We can plug in the given values for B (200 KHz) and C (12.8M bit/sec) and determine the required signal to noise ratio:
12.8e6=2e5 log2(1+S/N)
log2(1+S/N)=64
S/N=2^64 - 1
Each bit of signal-to-noise ratio corresponds to 3 decibels (6 dB if you're talking about power, which I won't, just to be charitable), so the required S/N ratio is around 190 dB.
Is this achievable? In a word, no. It's not even close. They're off by more than 100 dB.
When these guys claim that `when properly implemented and used under the right conditions, the digital modulation scheme reportedly delivers 90 bps/Hz-over the air' they're blowing smoke. They're saying they need a 270 dB S/N ratio. (That's larger than my number because I gave them some extra slack.) If they could get it, their scheme would work, but they can't, so it won't.
>a quarter of a million americans
>are killed in car accidents a year.
No, 42,000 people are killed in car accidents in the US every year. Unless there's a lot of Americans being killed abroad, you're off by a factor of 6. 250,000 americans suffer serious injuries in crashes every year, but only 1 in 6 dies. Reference
Uhh, Plan 9 is an operating system, not an
application -- there's no chance that it would
appear in anybody's process tables.
I just grepped my copy, and the string
`plan9.conf' appears nowhere in the
Plan 9 distribution.
The `original incantation' [sic] of Plan 9 was
Plan 9 -- it was a new system from the ground
up, developed in the Computing Science
Research Center at Bell Labs, not by any bunch
of Deccies working spare time. (Dave Presotto,
largely responsible for Plan 9 networking, worked
at DEC once, but that was before he went to grad
school, I believe.)
I speak having worked on Plan 9 at Bell
Labs for 6 years (and I have the Bowling
Shirt to prove it.) Among other things,
I wrote the shell and a whole load of
graphics junk.
My goal is to die having made an even number of sign errors.
This is pretty low. I know rtm and (slightly) his father, rhm. rtm is smart enough to find holes in sendmail on his own, and his father (who is now retired, BTW) is not the sort that would leak NSA information, even to his son. Have you ever read his report on breaking into 4.2BSD TCP/IP, written BEFORE his father went to work at NSA? (The funniest part of that research was the machine at Bell Labs named ucbvax, used to exploit a backdoor.)
This post gives the impression that 18 was universally the age of majority starting at the Big Bang up to the Disco Era. In fact, reduction of the age of majority from 21 to 18, and the accompanying abrogation of the in loco parentis doctrine by colleges happened during the Nixon administration (in 1971, when I was 18!), when the government decided that if US citizens at age 18 were old enough to send off to Viet Nam to die, then they were certainly old enough to vote (and drink, and take responsibility for their own sex lives.)
The retrenchment back to considering 21 the age of majority, at least for drinking purposes, happened gradually over the next few years. But really, 18 year olds were only legally considered adults for a very short period of time. (Except for voting, which would require repealing the 26th amendment.)
First of all, this quote is a real laugh:
The implication is that the realization that there are things that computers can't do is somehow new or unexpected.In fact, the very first computer science paper ever written (Alan Turing: "On Computable Numbers, with an Application to the Entsheidungsproblem" Proc. London Math. Soc. 42, 230-65, 1937) is about problems that have a definite, mathematical answer, but whose solution by a computer is mathematically, intrinsically impossible.
So, for more than sixty years, a good five or ten years before the first computers were built (depending on whose definition you accept) we already knew about problems that they couldn't solve.
Harel's book is not one of the first books in recent memory that focuses on the limits of computers. Any textbook on the theory of computation concentrates on this. The classic of the field is Computability and Unsolvability by Martin Davis, first published in 1958, which set the standard for how computing theory is taught. If you suffered through a theory course based on Turing Machines, it's Martin's fault. His book gets to its first proof that a problem is uncomputable half way down the third page of the introduction, before Chapter 1 even starts!
The idea that few people, even with advance [sic] computer science degrees understand these problems is laughable. Anybody with an advanced computer science degree from a major institution has sat through a theory course in which undecidable problems are broached within the first few weeks, and an algorithms course in which intractable problems (those that aren't impossible, but for which no feasibly fast algorithm exists) are a major topic.
>Or they must be at least
>writing a compiler for it...
Steve Johnson (who wrote yacc and pcc) works
there, so they probably have that one in hand.
Note: back at Bell Labs, Steve was Dave Ditzel's
boss for a while when Ditzel and Rae Maclellan
were designing the Crisp microprocessor, whose
branch-folding scheme smacks of the sort of
code-rewriting that the most interesting
Transmeta rumors feature.
Another former Labbie at Transmeta is Pat
Parseghian.