-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Frank Barknecht escribió: > Hallo, Jonny Stutters hat gesagt: // Jonny Stutters wrote: > >> The solution as I see it (and so this isn't a /completely/ >> negative rant ;) ) is more advocacy => more users => noticable >> difference to the bottom line of companies who don't supply >> quality (open-source where legally possible) Linux drivers. > > An important part of the solution also is to boycott vendors, who > don't provide specifications for the development of open source > drivers for their hardware. > > As long as people still buy NVidia chips, why should they change > their current position? > > Ciao I'll play devil's advocate here for a second, to what extent does a driver include such information of the hardware it "drives" that it would be considered a trade secret to release either its source or specifications of the hardware to build them? How much third party technology is involved in the drivers in the first place, that would violate *any* commercial agreements the hardware vendor may have with other companies in terms of licensed technologies? If removed these components from the drivers, how good their performance would be when compared to the closed ones? I'm not sure if you've noticed, but the video card market is one heck of a vicious business, with not only rapid development cycles, but also a lot (and I mean a *LOT*) of cross licensed technology not to mention a billion-dollar industry. This makes it way too complex (IMO) and extremely difficult to have fully open source drivers, unless it is for those considered low-end or entry-level class of cards/chips (where numbers level in the amount of units). I'm not sure what would be the panacea to this "problem", I for one have no problems using closed drivers (if they will get the job done, and will allow me to use the best hardware for the job). I do realize that if a high-end vendor opens up their drivers, Linux users will be more than happy to use them and there would be no problems with the GPL and other licensing issues, but this would also mean that these drivers wouldn't be allowed to include third-licensed technology in them. The most obvious situation with cross-license issues and open source video drivers has got to be the texture compression algorithms, especially S3TC. S3TC is not free, you have to license it to be able to implement it, and at the same time the hardware has to be compliant with it. I some years ago had a discussion about this very issue, and asked myself why wouldn't open source video drivers use an open texture compression algorithm... First I thought there was no such TC algorithm, but then I remembered 3dfx and the VSA-100 architecture (Voodoo 4, 5). They indeed developed a TC algorithm that they made open before they were bought up by nVidia, FXTC (not to be confused with DXTC, which is DirectX's texture compression, a derivative of S3TC). FXTC proved to have superior quality than S3TC, but it also required hardware support for it (as far as I remember, prove me wrong, please!), even if the algorithm and specification were open for anyone to implement, no one actually did (or did they?), so my allegation was why not use FXTC if it is open and pretty much (as much as I can remember) anyone could implement it for open source drivers as an alternative to S3TC? The answer is astonishingly simple: S3TC is now the de facto standard and the default format for a LOT of textures used in many applications and games, this would involve a massive move, which in turn would die (as it did) because (as MS-Office) some else made it not so nice, but bearably nice, cheaper and most of all, first (Ok, MSO wasn't there first, but it was cheaper). Trying to incorporate such technology at this point would be incredibly hard, I wonder, though if it would be harder than trying to make S2TC (and DXTC by extension) Open in turn? At the current rate, seems like VIA and Intel are the best options for anyone to have in terms of open source driver support, even if you lose in performance/features, why can't we have the three: features, performance and FREEDOM? Well, with current business models, it is simply not going to happen. Fortunately more player are willing to give us them three, but at baby-ant' steps. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (GNU/Linux) Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org iD8DBQFEmU6tXM+XOp70dwoRAqhiAJ0eZUe/QlQAa8qKgCjkdiCzwoVlpQCcDgNt 9gMkP8ypcOomCg9ypFqsGLM= =gL0I -----END PGP SIGNATURE-----
