Re: How long is it acceptable to leave *undistributable* files in the kernel package?
Humberto Massa wrote:
> @ 18/06/2004 10:39 : wrote Dave Howe :
>> At what point does the unpackager/installer become an
>> interdependency? most installers come in three forms 1) a archive
>> containing the product, and a uncompactor capable of extracting the
>> files from the archive, and correctly placing them (possibly under
>> the control of a script embedded in the archive) 2) an archive
>> containing the product *bonded with* the uncompactor as above 3) a
> I have been arguing that such bonding is "mere aggregation" (makes an
> anthology and not a derivative)
"Mere aggregation" has a very specific definition in the GPL:
> In addition, mere aggregation of another work not based on the Program
> with the Program (or with a work based on the Program) on a volume of
> a storage or distribution medium does not bring the other work under
> the scope of this License.
"work based on the Program" also has a specific definition:
> a "work based on the Program" means either the Program or any
> derivative work under copyright law: that is to say, a work
> containing the Program or a portion of it, either verbatim or with
> modifications and/or translated into another language.
When the two items are combined into one compiled program, that program
is "a work containing the Program or a portion of it". In addition, a
source file which contains the program as a string, or a kernel driver
which includes the firmware as a binary array, is also "a work
containing the Program or a portion of it". Therefore, this cannot
qualify as "another work not based on the Program with the Program".
Furthermore, a compiled program is no "a volume or a storage or
Finally, as Brian Thomas Sniffen said in <[🔎] email@example.com>:
> The GPL grants clear license to make and distribute several types of
> derivative work. It also grants clear license to make and distribute
> mere aggregations. It grants no other rights regarding incorporation
> into collective works.
> If the kernel+firmware bundle is a collective work and not a
> derivative work of the cpu-running kernel, as you say, then there is
> no license to distribute it.
So either the combined work is covered by the GPL, in which case it
must follow the rules of the GPL, or it is not covered by the GPL, in
which case it is non-distributable. My argument above shows that it is
not "mere aggregation", and all other forms of combined works covered by
the GPL are required to be distributed under the GPL.
- Josh Triplett
 I'm intentionally using "combined" rather than any specific term.