How about a driver for this ATI All-In-Wonder 3D Rage II +DVD PCI card I can't find drivers for?
This has been part of XFree86 and XOrg for years now (no 3d though - that is just marketing. The first "real" 3d card (i.e. capable of basic GL) was Rage Pro).
From a practical point of view 1kw of power consumption uses up about $70/month if turned on 24x7 (assuming a high rate of $0.10/kWh). This is about the price of DSL. In energy units, this is the same as 20 gallons of gasoline (if you decide to provide it with a portable generator) or 100kg of coal. Though, of course, the environmentally conscious way is to use nuclear - in which case this is about 100g of uranium (assuming 10% efficiency).
So far I'm hearing "commercial company hasn't written Linux drivers for their card". That's a legitimate complaint, but if the OSS community's reaction is to whine about it on cheesy blogs rather just hack the hardware...?
No, it is not, but it is strictly for pleasure only.
You see, 3d cards are complicated. On top of that the hardware itself if often finicky with lockups to the point that they should really be considered bugs.
So, you can only start once your got the hardware in your hands (which means after release) and with lots of work, at best you will have something semi-working a year later. It will be at least another year before the drivers mature so everyone can use them mostly without lockups. In the meanwhile ATI will release a few more variations and, if you are aiming for comprehensive support, you are back to square one.
If ATI wants to be nice to Open Source it means releasing partial specifications (at the very least) before the card is ready so that all their cards work with 2d, Xv and multi-monitor/multi-card when they are in stores (or a couple of months later) and having full specifications no later than 6 months after release.
Anything else and we are back to scrounging for older well-supported cards - which also happen to be a good deal cheaper and have less of a margin for ATI.
The latest card I have is Radeon 1600 - and given a choice I would gladly go back to R300 (or better yet - Rage Pro) if only those cards supported the resolutions I need and PCI express.
Look at the BBC entrance exam comparison. They show us one question given on a Chinese university entrance exam, and another from a British first year university exam. We don't know anything about the percentage of students who correctly answer either question, so the comparison is meaningless.
They just don't want to alarm the readers too much. And I am only half-joking - you'd be surprised to find out how perfectly smart students with no background can screw up an exam.
I'm sorry, but mathematics is useless as a defense mechanism: if someone has a gun on you, you can't put a Riemannian Manifold or anything like that in the way of the bullet, can you?
If you can get them to threaten you by shooting bullets at the feet and know the make of the gun you would have advance knowledge of when their cartridge will run out..
Agreed.I've never been a big fan of USB. The concept is fantastic, a unified connector that links just about any device to any other and can charge them is a great idea. However I am still bitter the Firewire lost out. It has more bandwidth, has sturdier connectors, and can deliver far more power. Being able to just plug one cable to power and link a hard drive would be great, I have one of those external IDE enclosures, and having *another* power brick is just silly.
One thing the author of the article missed is the original purpose of parallel and joystick ports. They were called "parallel digital input/output port" and "scientific port". The parallel port was meant for general purpose digital I/O (so you control relays or tell whether a contact is closed or not) and the scientific port could be used to connect thermocouples for example.
These were "starter" ports that made the computer useful in the lab or in the factory without purchasing additional components. Nowadays a special expansion card or external device are necessary to get back the same capability. The cheapest are around $100, but this quickly ratchets up to $1000 and more for anything capable.
One thing to keep in mind is that the original IBM PC (which was before XT and before AT) had a fairly slow processor (6 or 8 Mhz ? - can't recall) and the audio-range sampling rates of parallel port and joystick port matched them well. Nowadays we have Ghz cpus, but anything that captures faster than 100 Mhz is expensive - with Ghz cards costing more than the cost of the computer.
In all fairness the "Big Dig" was not just the tunnel - but also a reconstruction of the in-city highway system to free up a chunk of land previously obstructed by the highway. Working inside a busy city did not come cheap..
I'm afraid the days of drivers being programs that tweaked registers is long gone, my friend. Some of them include advanced optimizing compilers. There's a reason the nvidia driver is nearly the same size as the kernel itself - it's an extremely complex piece of code, with many clever things being done in software.
Right, but all we need to write said complex piece of code are register specs to the hardware. There are plenty of examples of high quality software (gcc, kernel, X) being written once the specs are available. Besides, trust me - there is nothing magically difficult about complex code, it is just a bit more fun to write that's all.
Also, I, personally, will gladly give up 10% of efficiency (heck, even 50) in exchange for the source code. If I need that efficiency back due to some particular application one can just find out what the bottleneck is and write an alternative path just for this piece of code.
There may be less fact there than those of us not in the know believe(d). From what I've read, register based video cards are obsolete and have long since been replaced with cards with work in some other (mostly unspecified) way. It seems that modern cards are as much intricate firmware as they are hardware, so releasing specs is no longer a simple matter of specifying register values.
Actually, a lot of current video cards use packet based approach. I.e. you write pieces of data that are somewhat similar to VLIW instructions into a memory area and then submit that memory area for processing. Some of these instructions initialize registers,
some fire off an indirect buffer, etc.
With regard to complexity, keep in mind that the video cards (and drivers) are developed by engineers who have a schedule - there is absolutely zero incentive to make it artificially complicated. A lot of complexity is not due to hardware design per se, but either because of bugs that need to be worked around or hacks to speedup the processing.
Not going to happen. NVidia and ATi have stated they couldn't open up the drivers if they wanted to. There's just too much licensed IP they don't have the rights to open.
They lie. Or pretext - or whatever current euphemism is.
The fact of the matter is that all they had to do is release register specs and the drivers would have been written, be more stable and provide fertile ground for experimentation with desktop technologies.
The latest ATI cards do not even work in 2d - probably because someone decided that hoarding "IP" is useful before AMD merger.
Wow, a post from the future! What's it like in 2036?
Well, when there is a dupe on Slashdot we can post directly to the original ! And we have arguments about whether the 1Gbps data limit is due to MAE-West or the entangled packet link somewhere in New Jersey.
So basically, the two NaNs have subtle semantics (much like his nullity) and don't have a catchy name or reuse a symbol that already means the golden ratio, therefore they're broken.
I think the big difference is that IEEE numbers were designed for practical use (if you got x=NaN you do not want if(x=y) to work) while his definition is designed for ease of teaching - it is probably easier to explain the rule for 0/0 rather than tell the students that in this case you have think what to do.
His example with f(x)=sin(x)/x is the best illustration - his arithmetic happily produces f(0)=NULL while in practice you should never assume that a floating point number is exact and thus the best definition is where f(x) is continuous in 0 and f(0)=1 and if the code is missing this special case it should return an error.
On the other hand, I have never seen an equivalent of NaN or NULL in analytic computation, so it might be a convenient shorthand after all in the similar way how +infinity is so convenient in measure theory. Of course, one big reason for doing analytic computation is that one can use continuity arguments and since NULL or NaN has to be an isolated point this would likely just introduce a bunch of combinatorics into derivations and make everything more complex.
On the contrary, it was very illuminating. Novell started the process first and, on top of that, they have illusions of technical collaboration with MS.
So, either they are really desperate and needed cash ASAP, or are thinking of somehow double-crossing MS on the deal.
From what I can understand the backwards-in-time measurement requires communication from one entangled photon to the other. This would allow faster-than-light communication which is the first thing you think of when you hear about entanglement. I thought it was well established that this was impossible since measuring one photon destroys the entanglement and you can never tell if you sent the signal or received it.
Can anyone explain how this experiment is different, and would it also allow for ftl comms?
The way I find easy to understand this is that quantum entanglement provides a way to prepare (in advance) two (or more) synchronized noise sources that remain synchronized regardless of time or space separation.
I.e. if you have source X and I have source Y than one can assure that each measurement i has the property X_i+Y_i=0. However, they are still random.
This can still be useful, but does not pass actual information between X and Y.
What I'd like to know is what in the hell is going on with the Aero theme that it is so absurdly demanding on the hardware.
I guess I don't understand the intricacies of what's going on because I see no reason whatsoever for a GUI to be more damanding than any contemporary PC game. The only excuse I see is sloppy and inefficient programming. It really leaves me with the impression that one of the big goals of Vista is to promote hardware sales.
Well, the big difference is that a game has a (relatively) small number of windows displayed at the same time and it can know exactly what their graphic demands are. I.e. if an extra transparent overlay is too much programmers can turn it off - before the release or in the patch or options..
With OS the easy way is the bruteforce "have each window paint into its own area and then combine" which (obviously) places a lot of stress on the hardware. Now, I am sure Windows is smarter than this, but likely not by much - it would not be able to get window snapshots otherwise.
In other words, there is a reason that old-style GUIs use static opaque rectangular windows - this can be done efficiently even on 1 Mhz cpus.
Again, you still don't get it. That type of attitude is precisely why a lot folks still shy away from Linux. They don't want to be told to do X, and not understand why it is that they're doing it. Most people would prefer to figure out the answer themselves. For most regular users, that means playing around with the GUI and hoping that the right button clicks solve their problem.
As someone who specialized in Usability for my Master's, I can honestly say that it would help tremendously if all software engineers were forced to watch usability studies involving normal computer users interacting with software. Or better yet, participating in such studies and/or tutoring such folks. You will quickly realize how attitudes like yours need to change, lest you continue to alienate people even more and send them running to easier-to-use (but less secure/powerful/etc.) alternatives.
A very interesting point, but I strongly disagree with the conclusion.
You are saying that users want to understand what is going on, don't want to learn what is going on and prefer a gui that creates the illusion that they understand what is going on ? It is such users attitude that needs to change. And, of course, having shiny gui - closed source only encourages such behaviour.
There are, of course, good guis: for example plots, spreadsheets, diff viewers, hex editors. And we need more of these. Guis that show what is happenning and provide the ability to drill down to the lowest detail required instead of displaying an animation running in separate thread that conveys no information whatsoever.
I haven't had a machine with one of those in at least 5 years. I also don't have a 5 1/4" floppy drive anymore. Both turn a modern dual-core machine into an Apple ][e class machine.
In all seriousness, why is this even supported in 64bit Vista?
Memory is no longer a constraint in a 64bit system. If you can afford $450+ (widely leaked price) for the non-crippled Vista, you can afford the RAM. And if you're running a server, paging = death, even when using 15k RPM drives.
A few notes:
Paging actually happens even without the page file - with memory mapped files, in particular executables. The page file (or swap partition) exists for anonymous memory - one that is not tied to a particular existing data.
One application of a pagefile is when you have opened 20 firefox windows a week ago, minimized them and then opened another 20 now. The old windows will likely have their data sitting in the pagefile and it is as they ware not running - except that in another week you can open the old windows and they will come back in a second or two. (Not sure how realistic this is with firefox, but I had done this plenty of times with konqueror, don't use firefox much).
Swap is exceedingly slow only if you are doing random access on it. On the other hand there are reasonable uses if you are trying to access in a linear fashion more memory than you physically have. For example, suppose you have 8GB machine and a 10 GB file you want a plot of. You could try to write your own script to do it while reading from disc or in chunk. Or, you could take an existing program (like R or Octave) load the file there and make a plot. Sure it'll take half an hour or so, but you could be doing something useful on another machine nearby.
Also it is not that easy to get a machine with large amount of RAM. On most motherboards (with the exception of some newer Serverworks chipsets) one has at most 4 sockets per CPU and the largest RAM stick one can get is 2 GB (there are 4GB ones but they are very expensive and slower). So a single cpu system can have at most 8GB. A two cpu - only 16 and the price would be more than $8K.
Lastly, on Linux the swapfile is also used to save system state when hibernating.
I think fundamentally digital downloads are more consumer friendly because in effect you cut out the middle men (Distrobution Centers/Resellers) and are able to get product directly to the user. Unfortunatly some of the digital download sites are charging to much per song to entice a greater number of legitmate downloaders.
Correct me if I'm wrong, but I was always under the impression that the NT kernel and its later variants was actually put together quite well with the input of the ex-VMS team, and that everything Microsoft dumped on top of it was the primary weak point in the OS.
They decided to use the opportunity of porting to 64-bit to install a new component - the Holy of Holes. I am not certain yet whether they mean one or many.
But the main point is essentially correct: Stallman is trying to aggressively expand his "freedom empire" with the GPL 3, and it could just bite him on the ass.
I don't think "expand" is the right term here. "Preserve" would be much better.
Keep in mind that when GPL-2 was created there was no such thing as DMCA and software patents were a rather exotic idea. GPL-3 is the answer to new laws and aggressive interpretations of old ones.
"Global" and "Internet" are not just empty words. Combine a potential loophole (as one can use GPL source and lock down binaries with crypto key and DMCA) and millions of people and there *will* be a few unscrupulous ones that will spoil it for everybody (example: e-mail).
So, yes, we do need GPL-3, but the issue of how to deal with existing GPL-2 software is truly a hairy one.
Slashdot Mirror
From a practical point of view 1kw of power consumption uses up about $70/month if turned on 24x7 (assuming a high rate of $0.10/kWh). This is about the price of DSL. In energy units, this is the same as 20 gallons of gasoline (if you decide to provide it with a portable generator) or 100kg of coal. Though, of course, the environmentally conscious way is to use nuclear - in which case this is about 100g of uranium (assuming 10% efficiency).
You see, 3d cards are complicated. On top of that the hardware itself if often finicky with lockups to the point that they should really be considered bugs. So, you can only start once your got the hardware in your hands (which means after release) and with lots of work, at best you will have something semi-working a year later. It will be at least another year before the drivers mature so everyone can use them mostly without lockups. In the meanwhile ATI will release a few more variations and, if you are aiming for comprehensive support, you are back to square one.
If ATI wants to be nice to Open Source it means releasing partial specifications (at the very least) before the card is ready so that all their cards work with 2d, Xv and multi-monitor/multi-card when they are in stores (or a couple of months later) and having full specifications no later than 6 months after release.
Anything else and we are back to scrounging for older well-supported cards - which also happen to be a good deal cheaper and have less of a margin for ATI.
The latest card I have is Radeon 1600 - and given a choice I would gladly go back to R300 (or better yet - Rage Pro) if only those cards supported the resolutions I need and PCI express.
If you can get them to threaten you by shooting bullets at the feet and know the make of the gun you would have advance knowledge of when their cartridge will run out..
Obviously, your submission was too soon :)
One thing the author of the article missed is the original purpose of parallel and joystick ports. They were called "parallel digital input/output port" and "scientific port". The parallel port was meant for general purpose digital I/O (so you control relays or tell whether a contact is closed or not) and the scientific port could be used to connect thermocouples for example.
These were "starter" ports that made the computer useful in the lab or in the factory without purchasing additional components. Nowadays a special expansion card or external device are necessary to get back the same capability. The cheapest are around $100, but this quickly ratchets up to $1000 and more for anything capable.
One thing to keep in mind is that the original IBM PC (which was before XT and before AT) had a fairly slow processor (6 or 8 Mhz ? - can't recall) and the audio-range sampling rates of parallel port and joystick port matched them well. Nowadays we have Ghz cpus, but anything that captures faster than 100 Mhz is expensive - with Ghz cards costing more than the cost of the computer.
But yes, lots of things were done inefficiently.
Nonsense. In security discussions Vista means a particular kind of a very large hole.
You might find the following site useful: http://www.webscriptions.net/ - lots of e-books in HTML, at prices below hardcopy versions.
Right, but all we need to write said complex piece of code are register specs to the hardware. There are plenty of examples of high quality software (gcc, kernel, X) being written once the specs are available. Besides, trust me - there is nothing magically difficult about complex code, it is just a bit more fun to write that's all.
Also, I, personally, will gladly give up 10% of efficiency (heck, even 50) in exchange for the source code. If I need that efficiency back due to some particular application one can just find out what the bottleneck is and write an alternative path just for this piece of code.
Actually, a lot of current video cards use packet based approach. I.e. you write pieces of data that are somewhat similar to VLIW instructions into a memory area and then submit that memory area for processing. Some of these instructions initialize registers, some fire off an indirect buffer, etc.
With regard to complexity, keep in mind that the video cards (and drivers) are developed by engineers who have a schedule - there is absolutely zero incentive to make it artificially complicated. A lot of complexity is not due to hardware design per se, but either because of bugs that need to be worked around or hacks to speedup the processing.
They lie. Or pretext - or whatever current euphemism is.
The fact of the matter is that all they had to do is release register specs and the drivers would have been written, be more stable and provide fertile ground for experimentation with desktop technologies.
The latest ATI cards do not even work in 2d - probably because someone decided that hoarding "IP" is useful before AMD merger.
Well, when there is a dupe on Slashdot we can post directly to the original ! And we have arguments about whether the 1Gbps data limit is due to MAE-West or the entangled packet link somewhere in New Jersey.
I think the big difference is that IEEE numbers were designed for practical use (if you got x=NaN you do not want if(x=y) to work) while his definition is designed for ease of teaching - it is probably easier to explain the rule for 0/0 rather than tell the students that in this case you have think what to do.
His example with f(x)=sin(x)/x is the best illustration - his arithmetic happily produces f(0)=NULL while in practice you should never assume that a floating point number is exact and thus the best definition is where f(x) is continuous in 0 and f(0)=1 and if the code is missing this special case it should return an error.
On the other hand, I have never seen an equivalent of NaN or NULL in analytic computation, so it might be a convenient shorthand after all in the similar way how +infinity is so convenient in measure theory. Of course, one big reason for doing analytic computation is that one can use continuity arguments and since NULL or NaN has to be an isolated point this would likely just introduce a bunch of combinatorics into derivations and make everything more complex.
On the contrary, it was very illuminating. Novell started the process first and, on top of that, they have illusions of technical collaboration with MS. So, either they are really desperate and needed cash ASAP, or are thinking of somehow double-crossing MS on the deal.
The way I find easy to understand this is that quantum entanglement provides a way to prepare (in advance) two (or more) synchronized noise sources that remain synchronized regardless of time or space separation.
I.e. if you have source X and I have source Y than one can assure that each measurement i has the property X_i+Y_i=0. However, they are still random.
This can still be useful, but does not pass actual information between X and Y.
Well, the big difference is that a game has a (relatively) small number of windows displayed at the same time and it can know exactly what their graphic demands are. I.e. if an extra transparent overlay is too much programmers can turn it off - before the release or in the patch or options..
With OS the easy way is the bruteforce "have each window paint into its own area and then combine" which (obviously) places a lot of stress on the hardware. Now, I am sure Windows is smarter than this, but likely not by much - it would not be able to get window snapshots otherwise.
In other words, there is a reason that old-style GUIs use static opaque rectangular windows - this can be done efficiently even on 1 Mhz cpus.
A very interesting point, but I strongly disagree with the conclusion.
You are saying that users want to understand what is going on, don't want to learn what is going on and prefer a gui that creates the illusion that they understand what is going on ? It is such users attitude that needs to change. And, of course, having shiny gui - closed source only encourages such behaviour.
There are, of course, good guis: for example plots, spreadsheets, diff viewers, hex editors. And we need more of these. Guis that show what is happenning and provide the ability to drill down to the lowest detail required instead of displaying an animation running in separate thread that conveys no information whatsoever.
A few notes:
Record companies (and RIAA) are the middlemen
They decided to use the opportunity of porting to 64-bit to install a new component - the Holy of Holes. I am not certain yet whether they mean one or many.
Sorry, couldn't resist..
I don't think "expand" is the right term here. "Preserve" would be much better.
Keep in mind that when GPL-2 was created there was no such thing as DMCA and software patents were a rather exotic idea. GPL-3 is the answer to new laws and aggressive interpretations of old ones.
"Global" and "Internet" are not just empty words. Combine a potential loophole (as one can use GPL source and lock down binaries with crypto key and DMCA) and millions of people and there *will* be a few unscrupulous ones that will spoil it for everybody (example: e-mail).
So, yes, we do need GPL-3, but the issue of how to deal with existing GPL-2 software is truly a hairy one.
I believe you are mistaken, MS never sold any OS but Windows..