Re: Let's stop feeding the NVidia cuckoo
Kalle Kivimaa <killer@debian.org> writes:
> Goswin von Brederlow <brederlo@informatik.uni-tuebingen.de> writes:
>> Source code is source code. Obfuscated or not does not change that. It
>> fullfills at least the letter of DFSG#2. For it to violate DFSG#2 you
>> would have to show that it is not source and the gcc already prooves
>> you wrong there. If you use 'obfuscation' or in other words
>> 'readability' as measurement what source is then a lot of perl code
>> would not qualify in my eyes.
>
> Source code is commonly defined as "the author's most preferred source
> of making modifications". If the author actually works with the
> obfuscated files, then fine, that is source. But if the author works
> with the unobfuscated version, then the obfuscated files are something
> like Java class files, not source.
Which in itself is near impossible to proove. We have to believe that
what is released as source by the author is source (as defined by its
license mainly) unless it becomes too unbelievable. e.g. a char foo =
{ 0x12, ... } that shows a gcc footprint when disassembled would be
absolutely unbelievable.
Having a "#define REG1 0x4653" or just "0x4653" wherever REG1 is
used certainly isn't out of the believable.
> Or do you claim that any precompiled Java class package should be able
> to go into main because you can always use gcj to compile those files
> into native code? That would help Debian Java support, though :)
Java class files are bytecode binaries. And while gcj can cross
compile from bytecode binary to native code they clearly are not
source. Their main inteded purpose is to be run on a java VM and not
being input for gcj.
But where (obfuscated) source ends and binary begins is very unclear
and probably overlapping too. I couldn't write down any consistent
rule for it. I can only go by gut feeling there.
MfG
Goswin
Reply to: