[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: LCC and blobs

Bruce Perens <bruce@perens.com> writes:

> Goswin von Brederlow wrote:
>>And that is without mentioning the non-free ness and license issues coming up after sarge. The firmware blobs and the like.
> The whole system has to be DFSG-free. Debian won't compromise on that.
> I have been thinking about the blob problem for a while. I propose to
> remove blobs from the driver, and store them as files in  initramfs
> (the kernel-internal filesystem created by the stuff in
> /usr/src/linux*/usr) or initrd.img. At boot time, the drivers would
> look for the blobs and load them if they are present, and fail
> gracefully or fall back if they are not. This gets around some GPL
> issues, because rather than be treated as code, the blobs become
> external files that are just sent to the device by the driver.
> Doing the above doesn't necessarily solve the problem for Debian,
> because we would still not want to distribute the blobs. but it allows
> the kernel to be GPL-compliant, which I don't feel it is while the
> blobs live in drivers.
>  An alternative is to make blobs their own loadable modules, but then
> we are treating them as code rather than as just a file that the
> kernel sends to some device, and we get GPL issues.
>     Thanks
>     Bruce

Blobs can be made binary files and stored in the initrd. Since no
linking is involved and the kernel-image without the blobs is still
fully functional (except for a very very few hardware constelations)
that should get the GPL issue settled.

As for distributing the blobs itself they can be relicensed under BSD
license or similar (if their aren't already) that doesn't have such a
problem with a char data[] = { 0x17, ... } source file, something
without the prefered source form clause. Even putting some in non-free
works fine.

But presonaly I think even distributing such blobs as GPL is fine as
they are the source "as recieved" from upstream. Only an author could
sue us for obfuscating the source but the author released the source
as is. No obfuscating is done on our part.

Where I see problems is firmware and sources that were not explicitly
released as GPL by the authors but somehow sneaked their way into the
kernel just the same. Many of the problem cases fall under this.

But enough said. It's all put off till sarge is out.


PS: Having the firmware as GPL would allow reverse engeneering to get
a more readable source format.

Reply to: