On Sun, Mar 30, 2008 at 08:45:31AM -0400, Alan Cox wrote: > I've been slowly working this way for a long time for quality, and other > reasons like paging. All the raving anti binary nutters have succeeded in > doing so far is making everyone hate them and setting the work back at > least two years. > > But it goes on - very slowly, as nowdays I have to do them one at a time > after long discussion with authors of the code having waiting until they > forgot the last stupid flamewar that put them against it. Indeed. Nuking code is not a productive solution. Users will just choose something that works, especially if they can't even install the OS due to missing drivers. Discriminating against vendors with blobs (vs. ones that don't) isn't very productive either. The driver with the blob may have a vendor that fully supports their driver, fix bugs in both firmware and driver. The one without any blobs may be the result of reverse-engineering (vendor totally un-cooperative) and the hardware has bugs that are unfixable (that would be fixable if the thing had uploadable firmware). Productive ways to make the world better - Make sure the libre alternatives work better than the blob ones (the "device contains flash you can upgrade" seldom do, and there's nothing you can do anyway) - Update drivers to use request_firmware - Come up with infrastructure so you can PXE-install with NICs that use firmware blobs (tg3, bnx2). More or less the driver will have to have the md5sum of the blob that it needs. (Or else you have all kinds of funny bugs). Well, RH used to have a "RMS Linux" product (without netscape etc.) for a short time. I suppose a similar distro based on Fedora could be done as long as it's not forced upon anyone who doesn't care. I don't see why the same infrastructure couldn't be used either. -- Pekka Pietikainen -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
