> I'm sorry, but you fail to see the bigger picture. > > > > One thing that is needed is open standard. Writing an operating system > > should not be about writing drivers for hundreds of devices. Standards > > should exists for drivers. Like OSKit or I/O Kit. Hince there should not > > have to be Linux drivers, FreeBSD drivers or Windows drivers, just Standard > > Drivers. > > I agree that having a standardised driver framework would probably not > be a bad thing, but there's another catch - DRIVERS FOR FREE OPERATING > SYSTEMS WOULD STILL NEED TO BE FREE!
That's a whole other issue, but yes. > > The optimal solution would be if compiled kernel modules could be loaded by > > any kernel that implements the standard, making them as portable as elf > > executables. > > Absolutely no. There's a bunch of damn good reasons why the majority of > the free software developers strongly oppose blobs (see > http://en.wikipedia.org/wiki/Binary_blob and > http://en.wikipedia.org/wiki/Graphics_hardware_and_FOSS#Problems_with_binary_drivers) > - for example even if you don't care about freedom (which is, by the way, > what gives you the ability to either fix the problematic module yourself or > get someone competent to fix it) at all, using blobs means all the security > and robustness of a vanilla kernel goes instantly out of the window (just try > fglrx), which means you can simply stick with Windowns and literally save > yourself all the trouble. Blobs don't belong in free systems, period. Keep in > mind that if you're not having any problems with them doesn't mean you never > will and no one else does and that blob having a bug tracker doesn't mean > someone actually cares about you. The most optimal solution would be if > hardware manufacturers finally realised how to properly cooperate with free > software community and either started writing BSD/GPL/MIT/whatever code > themselves and submitting it for inclusion in the next release of the > component they wish to support or at least providing complete NDA-free > documentation and paying some of that component's developers to write that > code for them (because documentation availability alone unfortunately doesn't > guarantee someone's really gonna bother). > > Read these articles to get a better idea: > > * http://airlied.livejournal.com/73115.html > * http://airlied.livejournal.com/73337.html I dont see how you get from portable binary drivers to binary blobs. With some hacks binary blobs can be made portable without binary portability, the fglrx drivers is proof of that. Of course its only portable to platforms their limited framework support, but still its being done. But that's still a whole different problem. With my suggestion you would be able to for example use the GPL drivers from Linux on FreeBSD or Solaris without recompiling them. If we also would be able to agree on standard for packages and repositories you would also be able to just install the drivers from a common repository. Currently i have machines running forks of the same OSS drivers on Linux and FreeBSD. Sometimes forks can be healthy but not in this case, the only reason for the forks is kernel compatibility. _________________________________________________________________ Håll skräpposten borta med nya Hotmail. Klicka här! http://explore.live.com/windows-live-hotmail -- Microsoft has a majority market share https://bugs.launchpad.net/bugs/1 You received this bug notification because you are a member of Ubuntu Bugs, which is a direct subscriber. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
