I also was equally annoyed at dselect. What a hard to use interface. It's supposed to be the easy way to install packages? The default sorting of "newest packages" at the top is pointless - when I'm installing a new system, there are certain things I want (say, xwindows), I don't even know if they're "new" or not. Also, the apt repository's large size is both it's biggest asset and biggest downfall. On one hand, there's lots and lots of stuff, and I've discovered programs in apt I've never heard of before. On the other hand, it makes dselect a nightmare to lose, because you end up scrolling through 1000 entries with only a vague idea of what you want.
The writers of the GPL anticipated your point and so the GPL specifies that the source must be presented in "the form preferred for making modifications" or something like that.
Well, from the article, it looks like this code modifies itself on a regular basis... so the object code is in fact (or, could be interpreted as by sufficiently well paid lawyers) "the form preferred for making modifications".
People once said computer chess programs could not beat the best human grandmaster...
I did say "yet":)
Whether you hire someone with no assenbler experience depend on what you hire him for. Writing programs for telecommunications or symbolic mathematics.
Very true. Assembler would not be the primary requirement for hiring someone to write a symbolic mathematics package, but I would still see it as a very big plus. All else (i.e. the primary skillset) being equal, the asm coder would be more aware of what the underlying machine would be doing and more capable of tuning the higher level code to use the hardware efficiently.
Efficient programming is not a factor of how low level your code is - it's a factor of how good your algorithm is.
Actually, it's both. Good algorithms give good big-O behaviour, and low-level knowledge gives tight control, and useful understanding, of machine behaviour.
One example: Quicksort is a good sort algorithm, yes ? I've seen developers use a call to quicksort() to sort a 4-element array inside a critical inner loop, which is a spectacularly dumb thing to do. The author of the code didn't understand what the compiler does with the source, what function call overhead costs, what the effect on the icache is, and all the other little details that affect performance.
The problem with O(n) notation and that whole worldview is that it is only meaningful for large values of n. People forget this.
A good algorithm in a high-level language will perform just as efficiently as the same algorithm implemented in assembly
Huh ? Try: "a good algorithm in a high-level language will perform no more efficiently than the same algorithm implemented in assembly"
I've not yet seen a compiler that can generate code as tight as a skilled human. Faster code generation - yes, but faster code - no.
and it will also take a tenth of the time to write and be a hundred times easier to maintain.
Agreed, but I run code more often than I write it.
High-level languages are there for describing the problem to a machine; the job of writing the solution should be left to compilers. Not only will they do a better job than most humans
Which is more of a reflection on the abilities of most humans than of the compilers...
Plus, even with a perfect compiler, all the layers of cruft that get automatically brought in when you use a high-level language bloat your code, nuke your icache, and soak up huge amounts of potential performance.
they can also do that better job for any system you may care to put them on. Whereas the x86 assembly language program I write today will be utterly worthless when we buy a PowerPC-based system tomorrow.
This, I will accept, is a valid argument. I tend to write a "reference implementation" for each module in C that gets built if no asm version is available for a specific platform.
It's not possible to write huge apps entirely in assembler, but you don't need to, just the critical 10%. My preference is for C for the bulk of my code, asm for the speed critical parts, and ML for high-level algorithm design and prototyping. I (personally) see no need for the plethora of middle-level languages.
Even if you don't choose to write assembler, understanding it lets you understand what the physical reality underneath all your wonderfully elegant algorithmic abstractions is actually doing - which is critical for writing high-performance, efficient code.
People who only know high-level languages feel to me like people who only get where they are going by using public transport: it's easier, they don't have to know about how vehicles work, but they have a limited choice of destination and no way of controlling the behaviour of the vehicle (and to stretch the analogy for amusement's sake: C programmers are like people who drive automatics, assembler programmers are like people who drive stick).
I guess nowadays it's "Oh, wait a minute... is my object model right?"
No, asshole, half the programming you do ISN'T object-oriented, OK?? And a significant piece of the other half shouldn't be written in Java, OK?? Oh, and by the way, that piece of SHIT code you wrote, multiplied by all the other pieces of SHIT code that you've written so far, adds up to ONE SLOW M-FING APPLICATION, in case you didn't notice.
Remember, we used to run pretty effective word processors in a 48K TPA on a 4mhz Z80. Think about that the next time you power up 37 megabytes worth of MS Word.
Preach on, brother !
High-level tools are for understanding the problem, not for describing the solution to a dumb machine.
Put it this way, if I don't see assembler on someone's resume, plus some evidence that they know how to make *efficient* use of the hardware they're given, I'm not going to be inviting them to the interview...
(Though, I'll admit, not every company is interested in making efficient use of hardware... which is great for me, because if they were, no-one would buy new stuff and I'd be out of a job...)
Someone mod the parent up all the way so that people get those wonderful words burnt into their brain.
That's true, but this S3 chipset is only moderately faster than the GeForce3-class GPU in the XBox
This is the latest one from S3 (Deltachrome), which with alpha drivers scores somewhere between ATI's 9500 and 9600. A little more than just "moderately faster" - your $300 system is coming with the equivalent of a $150 graphics card.
I did that in fact, and discovered that no amount of money will buy me more memory for my otherwise just fine computer. (ECC DIMMs or some such are required)
Try 1-800-4-memory. I got a whole bunch of ECC DIMMs for my quad (P2) Xeon box from them a while back. There's a cheapo clearance section too.
My understanding is that if the Linux kernel was being distributed under plain vanilla GPL, your drivers would be automatically brought under the GPL. However, IIRC (and you should spend some time with Google checking this yourself), Linus - as the copyright owner - has allowed a special exemption for binary drivers to be included in the kernel.
This is just from memory, I could be talking out of the wrong orifice, so I repeat: check this yourself.
Well, S3 was in at the start of the PC video card game. They probably hold enough 2D patents to block anyone who won't share with them from making a card that can even do 640x480 VGA.
Don't forget to return (empty or containing a nice message to the poor guy that is paid to process the replies) all prepaid business reply envelopes that get sent to you in junk mail.
Remember to staple the envelope closed. It jams the machines they use to automatically open and process them.
This is because every copy of a program that is made in the course of installing and running it (e.g. the installed copy on your hard-drive, the copy in memory, the bits of it in your swap, etc.) is ''actionable'' under copyright law, i.e. you can be sued for making those copies.
Capatial One just does note seem to get the hint I'm not interested in their dumb credit card. I have been known to get two applications in a single day. So I now take the envalopes, fill them with some gravel from the condo grounds, seal it, and send it back. Thay have to pay for the postage and it makes a mess when they open it in the processing centre.
I always thoroughly staple the envelope shut before sending it back to them. The staples jam the machine that opens the envelopes, causing the whole "production line" to stall.
However, until I meet a majority of Europeans that don't openly say they hate us, we're imperialistic, we deserved 9/11, we're shallow and ignorant, then I will have to think that the majority of Europeans truly do hate us. Consider it my own personal straw poll.
I'll bite. (Brit living in the US).
"they hate us"
Nope, never met an individual American that I've hated. I have mixed feelings about America as a as a political entity, but (a) that's a long way from hate, and (b) also true for any country you could name, including the one I was born in.
"we're imperialistic"
Not in the same way as colonial Europe was, certainly, though there's a definite tendency to what used to be termed "gunboat diplomacy" when practiced by the British. However, there's also a tendency in the US to talk a lot about the evils of European colonialism whilst conveniently ignoring what happened over here post-1776. At least the European powers eventually gave most (if not all) of the land back to the people they took it from, rather than making sure that there were not enough of the original owners left to complain, and both slavery and apartheid were abandoned in Europe long before the US. Maybe the reason the "imperialist" tag gets thrown at the US nowadays is that it is so often thrown by the US at (historical) Europe in such a hypocritical fashion.
we deserved 9/11
Nope, innocent people never deserve to die. I can *understand* (to some extent) why it happened - but that is a long way from saying it was right. I suspect (and this is just opinion based on my experience in the US) that there are more Americans that will say that the UK deserved the hundreds of terrorists acts committed by the IRA than Europeans that will say that the US deserved 9/11.
we're shallow and ignorant
I've met hundreds of Americans who are neither shallow nor ignorant, and very few who are either. The reason that the US is perceived as shallow and ignorant - apart from the current President's unique linguistic style - is that its popular culture is so much more widely exported. Popular cultures everywhere tend to the lowest common denominator.
So, in summary, (a) I don't feel that any of those statements apply to Americans (as individuals), but the picture is a little greyer (but only a little) when talking about the US as a political entity rather than a collection of people; and (b) I like the US well enough to choose to live here and have met many Americans I like and very few that I dislike. (Plus, your women just *love* English accents...)
Got to agree with you. It really seems heavily dumbed-down and flashed-up compared to how it used to be. This will be the last subscription I take out as well.
Slashdot Mirror
I also was equally annoyed at dselect. What a hard to use interface. It's supposed to be the easy way to install packages? The default sorting of "newest packages" at the top is pointless - when I'm installing a new system, there are certain things I want (say, xwindows), I don't even know if they're "new" or not. Also, the apt repository's large size is both it's biggest asset and biggest downfall. On one hand, there's lots and lots of stuff, and I've discovered programs in apt I've never heard of before. On the other hand, it makes dselect a nightmare to lose, because you end up scrolling through 1000 entries with only a vague idea of what you want.
Less painful than dselect:
aptitude
or
apt-cache search foo | less
HTH
Well, from the article, it looks like this code modifies itself on a regular basis ... so the object code is in fact (or, could be interpreted as by sufficiently well paid lawyers) "the form preferred for making modifications".
People once said computer chess programs could not beat the best human grandmaster ...
:)
I did say "yet"
Whether you hire someone with no assenbler experience depend on what you hire him for. Writing programs for telecommunications or symbolic mathematics.
Very true. Assembler would not be the primary requirement for hiring someone to write a symbolic mathematics package, but I would still see it as a very big plus. All else (i.e. the primary skillset) being equal, the asm coder would be more aware of what the underlying machine would be doing and more capable of tuning the higher level code to use the hardware efficiently.
You, sir, are an idiot^W^Wignorant^Wmisguided.
...
...
Efficient programming is not a factor of how low level your code is - it's a factor of how good your algorithm is.
Actually, it's both. Good algorithms give good big-O behaviour, and low-level knowledge gives tight control, and useful understanding, of machine behaviour.
One example: Quicksort is a good sort algorithm, yes ? I've seen developers use a call to quicksort() to sort a 4-element array inside a critical inner loop, which is a spectacularly dumb thing to do. The author of the code didn't understand what the compiler does with the source, what function call overhead costs, what the effect on the icache is, and all the other little details that affect performance.
The problem with O(n) notation and that whole worldview is that it is only meaningful for large values of n. People forget this.
A good algorithm in a high-level language will perform just as efficiently as the same algorithm implemented in assembly
Huh ? Try: "a good algorithm in a high-level language will perform no more efficiently than the same algorithm implemented in assembly"
I've not yet seen a compiler that can generate code as tight as a skilled human. Faster code generation - yes, but faster code - no.
and it will also take a tenth of the time to write and be a hundred times easier to maintain.
Agreed, but I run code more often than I write it.
High-level languages are there for describing the problem to a machine; the job of writing the solution should be left to compilers. Not only will they do a better job than most humans
Which is more of a reflection on the abilities of most humans than of the compilers
Plus, even with a perfect compiler, all the layers of cruft that get automatically brought in when you use a high-level language bloat your code, nuke your icache, and soak up huge amounts of potential performance.
they can also do that better job for any system you may care to put them on. Whereas the x86 assembly language program I write today will be utterly worthless when we buy a PowerPC-based system tomorrow.
This, I will accept, is a valid argument. I tend to write a "reference implementation" for each module in C that gets built if no asm version is available for a specific platform.
It's not possible to write huge apps entirely in assembler, but you don't need to, just the critical 10%. My preference is for C for the bulk of my code, asm for the speed critical parts, and ML for high-level algorithm design and prototyping. I (personally) see no need for the plethora of middle-level languages.
Even if you don't choose to write assembler, understanding it lets you understand what the physical reality underneath all your wonderfully elegant algorithmic abstractions is actually doing - which is critical for writing high-performance, efficient code.
People who only know high-level languages feel to me like people who only get where they are going by using public transport: it's easier, they don't have to know about how vehicles work, but they have a limited choice of destination and no way of controlling the behaviour of the vehicle (and to stretch the analogy for amusement's sake: C programmers are like people who drive automatics, assembler programmers are like people who drive stick).
Anyway, morning coffee finished, ramble over
I guess nowadays it's "Oh, wait a minute... is my object model right?"
... which is great for me, because if they were, no-one would buy new stuff and I'd be out of a job...)
No, asshole, half the programming you do ISN'T object-oriented, OK?? And a significant piece of the other half shouldn't be written in Java, OK?? Oh, and by the way, that piece of SHIT code you wrote, multiplied by all the other pieces of SHIT code that you've written so far, adds up to ONE SLOW M-FING APPLICATION, in case you didn't notice.
Remember, we used to run pretty effective word processors in a 48K TPA on a 4mhz Z80. Think about that the next time you power up 37 megabytes worth of MS Word.
Preach on, brother !
High-level tools are for understanding the problem, not for describing the solution to a dumb machine.
Put it this way, if I don't see assembler on someone's resume, plus some evidence that they know how to make *efficient* use of the hardware they're given, I'm not going to be inviting them to the interview...
(Though, I'll admit, not every company is interested in making efficient use of hardware
Someone mod the parent up all the way so that people get those wonderful words burnt into their brain.
By the way, it's very unlikely that you'd get $135K in the Valley now, unless you had a Masters degree and 10 years experience.
Heh, I wish. I work in the Valley, have a Masters and 10+ years experience and I don't make $135k. Just happy to be still working right now...
You forgot the classic: "it's not leaking oil, it's marking its territory".
Yes, I still love my MG...
This is the latest one from S3 (Deltachrome), which with alpha drivers scores somewhere between ATI's 9500 and 9600. A little more than just "moderately faster" - your $300 system is coming with the equivalent of a $150 graphics card.
I did that in fact, and discovered that no amount of money will buy me more memory for my otherwise just fine computer. (ECC DIMMs or some such are required)
Try 1-800-4-memory. I got a whole bunch of ECC DIMMs for my quad (P2) Xeon box from them a while back. There's a cheapo clearance section too.
My understanding is that if the Linux kernel was being distributed under plain vanilla GPL, your drivers would be automatically brought under the GPL. However, IIRC (and you should spend some time with Google checking this yourself), Linus - as the copyright owner - has allowed a special exemption for binary drivers to be included in the kernel.
This is just from memory, I could be talking out of the wrong orifice, so I repeat: check this yourself.
Or even:
Darth McBride: "Linux, I am your author"
Linux: "No. It can't be. That's not true. That's impossible!"
Darth McBride: "Search your CVS, you know it to be true"
There are people who actively share the contents of their entire hard disk on KaZaA.
A quick search on Gnutella for "doc" can be very very enlightening...
That hunk of paper can't be good for SCO's bottom line...
I would have said that applying it to their bottoms is the only thing that it is good for.
... at atmospheric pressure
Well, S3 was in at the start of the PC video card game. They probably hold enough 2D patents to block anyone who won't share with them from making a card that can even do 640x480 VGA.
I shut down every night to save energy! With the new CPU's generating all the heat its worth it.
I treat it as a tax-deductible way of heating my apartment...
Don't forget to return (empty or containing a nice message to the poor guy that is paid to process the replies) all prepaid business reply envelopes that get sent to you in junk mail.
Remember to staple the envelope closed. It jams the machines they use to automatically open and process them.
Given that they've claimed they own code that Linus himself wrote, I'd expect the lawsuit to be the other way round, if anything.
the Fed mandates a maximum liability of $50
I still don't understand this - why are you liable if it's not got your signature on it ? Anyone care to enlighten me ?
This is because every copy of a program that is made in the course of installing and running it (e.g. the installed copy on your hard-drive, the copy in memory, the bits of it in your swap, etc.) is ''actionable'' under copyright law, i.e. you can be sued for making those copies.
Really ?
I always thoroughly staple the envelope shut before sending it back to them. The staples jam the machine that opens the envelopes, causing the whole "production line" to stall.
However, until I meet a majority of Europeans that don't openly say they hate us, we're imperialistic, we deserved 9/11, we're shallow and ignorant, then I will have to think that the majority of Europeans truly do hate us. Consider it my own personal straw poll.
I'll bite. (Brit living in the US).
"they hate us"
Nope, never met an individual American that I've hated. I have mixed feelings about America as a as a political entity, but (a) that's a long way from hate, and (b) also true for any country you could name, including the one I was born in.
"we're imperialistic"
Not in the same way as colonial Europe was, certainly, though there's a definite tendency to what used to be termed "gunboat diplomacy" when practiced by the British. However, there's also a tendency in the US to talk a lot about the evils of European colonialism whilst conveniently ignoring what happened over here post-1776. At least the European powers eventually gave most (if not all) of the land back to the people they took it from, rather than making sure that there were not enough of the original owners left to complain, and both slavery and apartheid were abandoned in Europe long before the US. Maybe the reason the "imperialist" tag gets thrown at the US nowadays is that it is so often thrown by the US at (historical) Europe in such a hypocritical fashion.we deserved 9/11
Nope, innocent people never deserve to die. I can *understand* (to some extent) why it happened - but that is a long way from saying it was right. I suspect (and this is just opinion based on my experience in the US) that there are more Americans that will say that the UK deserved the hundreds of terrorists acts committed by the IRA than Europeans that will say that the US deserved 9/11.
we're shallow and ignorant
I've met hundreds of Americans who are neither shallow nor ignorant, and very few who are either. The reason that the US is perceived as shallow and ignorant - apart from the current President's unique linguistic style - is that its popular culture is so much more widely exported. Popular cultures everywhere tend to the lowest common denominator.
So, in summary, (a) I don't feel that any of those statements apply to Americans (as individuals), but the picture is a little greyer (but only a little) when talking about the US as a political entity rather than a collection of people; and (b) I like the US well enough to choose to live here and have met many Americans I like and very few that I dislike. (Plus, your women just *love* English accents...)
Interestingly enough, the process of learning this huge amount of geographical information causes verifiable physical changes in the brains of London's cabbies.
Pot, Kettle.
Got to agree with you. It really seems heavily dumbed-down and flashed-up compared to how it used to be. This will be the last subscription I take out as well.