Re: Binaries under GPL(2)

[Walter Landry <wlandry@ucsd.edu>]

Alexander Cherepanov <cherepan@mccme.ru> wrote:
> 15-Dec-03 07:39 Walter Landry wrote:
> > Alexander Cherepanov <cherepan@mccme.ru> wrote:
> >> 8-Dec-03 20:43 Walter Landry wrote:
> >> Thus, when distributing binaries compiled from sources, the
> >> compilation is under Section 2 and the distribution is under Section
> >> 3. That's part of the original confusion--requirement of source form
> >> in Section 2 applies only to distribution, not to modification.
> 
> > Compilation is not covered by the license, only distribution.
> > Distribution of modified binaries is covered by Sections 2 and 3.
> > Section 3 lets distribute binaries as long as you distribute source.
> > Section 2 tells you what you have to do if you modified the source.
> 
> Pardon? You agreed above that compilation is a kind of modification.
> And modification is clearly covered by Section 2.
> Or you mean that Section 3 implicitly contains a permission to
> compile? Or what?

I mean that Section 3 lets you distribute binaries as long as you
distribute source.  If you happen to have modified that source, then
you have to comply with Section 2 in how you've modified the source
(prominent notices etc.)

> >> > Distributing binaries under Section 2 probably means
> >> > editting the binaries with a hex editor.  You also need to have the
> >> > rights to distribute everything in the binary under the GPL.
> 
> BTW, that's also interesting. Can you distribute, under Section 3, a
> binary statically linked with a library from non-free compiler?
> Probably not. Section 3 clearly says that the binary must be
> distributed under the terms Sections 1 and 2. And 2b says that the
> whole thing must be under the GPL. If the library linked in is, for
> example, non-modifiable that's impossible. Though one can argue that
> this is the usual way of creating an executable and therefore Section
> 3 gives a permission to do just that...

Section 3 specifically lists the compiler as one of the things that
you don't have to distribute under the GPL.  Historically, many things
were compiled with a non-free compiler and distributed by the FSF
under the GPL.

> >> > With
> >> > non-free compilers, that may be a problem.  With gcc, that probably
> >> > means more hex editing to include the FSF, HP, SGI, etc. copyrights.
> >>
> >> The only difference in distribution under Section 2 and under Section
> >> 3 is the requirement for sources.
> 
> > True.  That would seem to imply that you have to preserve copyright
> > notices etc. in all of the modified files.  Otherwise it makes no
> > sense to refer to the terms of Section 1.
> 
> If it makes no sense to refer to it from Section 2 then it makes no
> sense to refer to it from Section 3 at all.
> 
> In fact, Section 1 contains the following requirements:
> 
>  (a)  you conspicuously and appropriately publish on each copy an
>       appropriate copyright notice and disclaimer of warranty;
> 
>  (b)  keep intact all the notices that refer to this License and to
>       the absence of any warranty; and
> 
>  (c)  give any other recipients of the Program a copy of this License
>       along with the Program.
> 
> You seems to talk only about (b) but (a) and (c) are quite reasonable
> in any situation (with "the Program" in (c) replaced by something
> else, of course).

Exactly.  But those are much easier to comply with.

> > But the usual way of
> > creating an executable is by running code through a compiler, which
> > removes most of those notices.  So you could argue that Section 3
> > gives you permission to do just that.
> 
> Now I get your idea. You say that one usually cannot distribute a
> binary under Section 2 because he cannot comply with some technical
> requirements from Section 1 but these requirements are overridden when
> distributing under Section 3. That would be an interesting solution.

Yup.

> >> The hole in the explicit wording seems to be so clear that I start
> >> doubting it is just an oversight. Maybe it's normal for sections of a
> >> license to trump each other?
> 
> > The hole is there, but exploiting it is hard.  People don't normally
> > modify machine code.
> 
> That's true for machine code but there are other kinds of binaries.
> Consider info for example. It usually contains copyright notices and
> it usually is not source. Is it ok to distribute .info without .texi?

As long as it complies with Section 2, it should be fine.  I don't
know if, in general, .info files will comply.  For example, does any
copyright from makeinfo leak into the .info files?

Regards,
Walter Landry
wlandry@ucsd.edu