Re: Summary: binary firmware in the kernel
Zefram <zefram@fysh.org> writes:
> Adrian Bunk wrote:
> >1. a 300 kB binary firmware is DFSG-free
> >
> >2. binary firmware must go to non-free
> >this implies that the installer will no longer be able to install Debian
> >on several computers
>
> There was also a range of opinions of the nature "binary firmware *might*
> be DFSG-free, on a case-by-case basis". It was suggested that if a
> hardware manufacturer provided a "char firmware[] = { ..." file with a GPL
> header, and claimed that that was the preferred form for modification,
> then the firmware was free. There were other opinions expressed that
> varied on what the presumption should be if the file was not explicitly
> stated to be source.
>
> -zefram
Some Arguments (on irc) were wether the claim "that that was the
preferred form for modification" is believable. If the file is
compiled it should be possible to disassemble it and show certain
patterns compilers generate, to fingerprint the binary to a certain
compiler.
Nowadays I think anyone would be hard pressed arguing that hex code is
the prefered form for modification. But since the code is GPL anyone
with intrest and knowledge can pick it apart and convert it into a
more readable form. It would be hard work but it would be legal
everywhere to pick the binary code apart and to replace parts of it
with 'real' source.
Having "char firmware[] = { ..." is certainly not nice and I hope any
firm doing that gets pressed into releasing the real source or gets
the code forked, disassembled and picked apart until its recognisable.
As a sidenote a similar problem was just discussed for certain
bootloaders for m68k/mac. The problem is that the original bootblock
code is non-free so it can't be used in the bootloader. A "char
bootcode[] = { ... }" under GPL from Apple would perfectly solve the
problem and the code would be picked apart, disassembled and properly
commented within hours.
MfG
Goswin
Reply to: