Beware: Numerics can be tricky! It can be the case that intermediate results need far more precision than your final result. It's not difficult to construct mathematical problems whose results are 8-bit, but where double precision is required for the computations. Especially in science, where you're asking new questions, you don't know the answer and prudence suggests that using more intermediate precision will save you headaches. Similar reasong justifies not flushing denormal numbers (floating point numbers smaller than the "machine epsilon", or smallest "normally" representable number greater than zero) to zero, but keeping as many digits as possible. This is provided by full IEEE 754 compliance. So go for full IEEE 754 double precision; it may prevent another Ariane 5 catastophe!
Really, who cares if there are multiple versions installed? Well, I used to, and I still appreciate minimal elegance of a single library, but modularity cannot be won without some redundancy. Think of the library version problem: we need to get developers at disparate locations and levels of time commitment all to update to a specific level of library? E.g., what if one developer's app depends on a specific bugfix/update of a library, and another very busy developer's app does not? This sounds like a recipe for disaster, or, the nightmare that linux distro. developers have to confront all the time. We need more decentralization!
I think this issue of centralization is a key difficulty with linux package management. Thanks for bringing it up. Now that disk space is cheap, we can afford to install libraries redundantly, with the guarantee that each app has the right versions of libraries installed. What we need is a decentralized package manager. Let's not worry about point and click for now, just command line. I think LSB aimed at providing a basis of known libraries, so that installers could work. Frankly I think the BSD approach of developing a core system with known and supported libraries is nice. But it shouldn't be necessary for a distributor of apps to have their app sanctioned on / approved for the ports tree. They should be able to distribute it themselves from their own web site. Currently either they are forced to pick a distribution or they must provide a tar-ball of source. (Binary tar-balls are pretty risky as they have binary dependencies which aren't enforced.)
Stop whining? RMS not only instilled in us the ideal of free software but also provided gobs of it and with excellent quality yet!
Move on? You mean accept the situation. You mean define what _we_ desire in terms of other's interests? Especially if those interests can hurt us all? I say respect yourself more and value your freedom.
I saw on the listing of seminars/tutorials at WWDC a few weeks ago that Apple was presenting a session on OpenGL Shading Language, yet I didn't even know it was supported on OS X. Is it currently supported in Panther? Will it be supported in Tiger? Will there be an IDE for writing shaders, etc?
Isn't it possible to save money by running half as many video cards, each running 2 X servers? I often run both VGA ports off a single card but both on a single X server with Xinerama, but maybe there's some X limitation that prevents 2 X servers for the 2 ports on the same card. Is this possible?
Anyone got the shinny on whether there'll be drivers & compiler for the new shading language? I'd speculate that they'll do so to stay on the leading edge of open standards. Only 3dlabs has a compiler for windows & linux now, and apparently there is some buggy support for ati.
Slashdot Mirror
Beware: Numerics can be tricky! It can be the case that intermediate results need far more precision than your final result. It's not difficult to construct mathematical problems whose results are 8-bit, but where double precision is required for the computations. Especially in science, where you're asking new questions, you don't know the answer and prudence suggests that using more intermediate precision will save you headaches. Similar reasong justifies not flushing denormal numbers (floating point numbers smaller than the "machine epsilon", or smallest "normally" representable number greater than zero) to zero, but keeping as many digits as possible. This is provided by full IEEE 754 compliance. So go for full IEEE 754 double precision; it may prevent another Ariane 5 catastophe!
What are you running your renderer on now? Or is this power lust? You'll pay a heavy price, especially in your time. I regret doing this myself.
Really, who cares if there are multiple versions installed? Well, I used to, and I still appreciate minimal elegance of a single library, but modularity cannot be won without some redundancy. Think of the library version problem: we need to get developers at disparate locations and levels of time commitment all to update to a specific level of library? E.g., what if one developer's app depends on a specific bugfix/update of a library, and another very busy developer's app does not? This sounds like a recipe for disaster, or, the nightmare that linux distro. developers have to confront all the time. We need more decentralization!
I think this issue of centralization is a key difficulty with linux package management. Thanks for bringing it up. Now that disk space is cheap, we can afford to install libraries redundantly, with the guarantee that each app has the right versions of libraries installed. What we need is a decentralized package manager. Let's not worry about point and click for now, just command line. I think LSB aimed at providing a basis of known libraries, so that installers could work. Frankly I think the BSD approach of developing a core system with known and supported libraries is nice. But it shouldn't be necessary for a distributor of apps to have their app sanctioned on / approved for the ports tree. They should be able to distribute it themselves from their own web site. Currently either they are forced to pick a distribution or they must provide a tar-ball of source. (Binary tar-balls are pretty risky as they have binary dependencies which aren't enforced.)
Be Einstein!
Stop whining? RMS not only instilled in us the ideal of free software but also provided gobs of it and with excellent quality yet! Move on? You mean accept the situation. You mean define what _we_ desire in terms of other's interests? Especially if those interests can hurt us all? I say respect yourself more and value your freedom.
I saw on the listing of seminars/tutorials at WWDC a few weeks ago that Apple was presenting a session on OpenGL Shading Language, yet I didn't even know it was supported on OS X. Is it currently supported in Panther? Will it be supported in Tiger? Will there be an IDE for writing shaders, etc?
Isn't it possible to save money by running half as many video cards, each running 2 X servers? I often run both VGA ports off a single card but both on a single X server with Xinerama, but maybe there's some X limitation that prevents 2 X servers for the 2 ports on the same card. Is this possible?
Unfortunately, any two users don't always need the same 10 services.
Anyone got the shinny on whether there'll be drivers & compiler for the new shading language? I'd speculate that they'll do so to stay on the leading edge of open standards. Only 3dlabs has a compiler for windows & linux now, and apparently there is some buggy support for ati.