On Sun, 2008-06-08 at 18:39 +0200, Hans de Goede wrote:
> Alex now says it isn't good enough, and is actually
counter-productive.
If I get Alex correctly he is saying that, to his goal, which is 100% Free
software everywhere (including in his toothbrush), this is counterproductive,
as it may make it easier to distribute binary firmware along with the kernel,
as it now could be put in a seperate tarbal removing GPL worries etc.
If it's a GPL violation to ship the non-GPL'd firmware in the source
tarball, it's also a GPL violation to distribute it as part of a
vmlinux.
So yes, putting it in a separate source tarball removes the GPL worries
there -- but if you're then going to use the CONFIG_BUILTIN_FIRMWARE
feature and build non-GPL'd firmware into your vmlinux, that brings
exactlly those same worries right back again.
If you build non-GPL'd firmware into your kernel, you may not distribute
that kernel image (although currently we do just that; we shouldn't).
If the firmware truely gets put in a different tarbal (and thus
eventually in a
different srpm), then it will be feasible to do a no blobs included Fedora spin
like gnewsense, which would be great.
Fedora already uses 'doctored' tarballs for stuff like openssh; I think
it would be reasonable enough to do that with a firmware-less kernel
source tarball too. It's easy enough to build the firmware package
separately (although first, I'm going to make the kernel srpm spit out a
'kernel-firmware' subpackage; we can talk about splitting it later).
Now Alex worries about someone still slipping in some firmware into
the kernel
itself instead of into the firmware package, well as with the current situation
with a completely seperate kernel, audits will still be necessary.
That's nothing special -- it's just the same as we should always be
vigilant for someone slipping non-GPL'd _code_ into the kernel.
Should we just give up, just because people slip up occasionally?
--
dwmw2