Fedora Freedom and linux-libre

[Les Mikesell]

Alexandre Oliva wrote:
>
>> No, it _is_ the point.
> 
> No, you're focusing only on the firmware.  I'm focusing on the whole
> work distributed under the name Linux and allegedly (and falsely)
> under GPLv2, in spite of containing these firmwares so intertwined
> that you can't just rip off the "pages" of the book corresponding to
> the firmwares and expect the rest of make sense (to a compiler).
> Although a number of the firmwares are indeed printed in "separate
> pages" (files), some are actually interspersed with GPLed code. 

What I don't understand is why you think there is some requirement in 
terms of page boundaries or storage mechanism involved in copyright 
separation.  I've never seen any such thing.

> Even
> if you were to believe the theory that separate files bound together
> say at link are covered by the "mere aggregation in the same media"
> exception, I don't see how it could possibly fly when it's hard even
> to pinpoint the separate works within a single file, and that
> modifying that file so as to remove the offending portions might even
> be argued as a violation of the license of the offending portions.
> This all *screams* SINGLE WORK, not separate works.

No, it screams common storage, particularly when you know that it runs 
on separate hardware elements and can't possibly be a single work when 
it executes.

> To be separate
> works, you'd have to start out by being able to point at and obtain
> both/all the works separately.

Where is that requirement stated?  I've never seen anything explicitly 
stating technical details of how separation must appear or be maintained.

>> The GPL'd part doesn't depend any more or less on it whether it is
>> loaded on the fly or already there in ROM
> 
> Please tell me you don't honestly believe you wouldn't have to modify
> the GPLed part to be able to load the tg3 firmwares on the fly or use
> them straight from ROM.

I believe there there wouldn't be any difference if the identical 
content were already in ROM - or some non-GPL'd pre-execute loader 
installed it instead of the kernel.

>> You do have access to all of the parts and the freedom to split them
>> up any way you want.  What's stopping you?
> 
> Nothing.  In fact, I'm doing just that.
> 
> However, this doesn't mean I didn't have to modify any of the
> allegedly separate works in order to accomplish that.  I did.  So how
> are they separate works in the first place?

It is unrelated content that generally executes in an unrelated context. 
   It could have been in ROM already.  It could have been loaded by 
something else.  Loading those contents some other way doesn't make them 
more or less separate, just less convenient.  This might be made more 
clear if the loader is generalized and the contents separated so that it 
is obvious that changes in firmware contents do not need modifications 
in the GPL'd mechanism - and in fact are not related to the loader at 
all. But that seems obvious anyway, like the code that plays music is 
separate from the music content, or an ebook text is separate from the 
code that displays it.  That would not be changed even if you combined 
them in the same medium or file for convenient storage.

> If you have to rephrase a chapter of a book for it to make sense if
> you print the book after removing another chapter, could you honestly
> defend the claim that the chapters were separate works in the first
> place?

Yes, and I think copyright law supports that even if it is a bit fuzzy.

>> Exactly.  If such aggregations weren't permitted and perfectly normal
>> when you meet the terms of each component separately, we probably
>> wouldn't have had any anthologies to read for homework in school.
> 
> Read again what you wrote.  "when you meet the terms of each component
> separately".  Think about it.

I did think about it.  It is an aggregation, explicitly permitted by the 
GPL, so only the terms of the loaded component could be a problem with 
the combination.  And I don't see any more reason to question the 
compliance of terms on included firmware than on included sources.

> Now, consider this: you got a copy this fictitious kernel with source
> code and firmware embedded in it, but you don't have permission to
> modify the kernel at all,   even though you have source code and
> permission to study it, recompile it and distribute it.  So you look
> at the sources and you notice that you don't want this firmware in
> your kernel image, because you don't have the device that requires it.

I don't think that's a sufficient reason for me to want to remove 
something, but OK.

> So you remove the firwmare and notice that the kernel won't compile
> any more.  What now?

Ignore it, consider it a feature in case you get one of the devices in 
question, or ask someone with permission to modify to make the change 
for you.

> Do you get permission to modify the kernel
> (creating a derived work thereof) just because it won't compile after
> you rip off a "separate work"?

If it mattered that much, I'd try to figure out why the removal affected 
compilation and supply a compiler-satisfying substitute.

> Is it really still a separate work if
> you have to do that?

Yes, unless you can establish that it is impossible to provide a 
compiler-satisfying alternative.

-- 
   Les Mikesell
     lesmikesell at gmail.com