Re: GTK Netscape is impossible

[Carl Thompson <carlt cc64317-a hwrd1 md home com>]

On Tue, 10 Mar 1998, Gregor Hoffleit wrote:

...

> I hope your reply ("Please go back and read it.") didn't apply to my
> posting.  I didn't write a word about the consequences of the LGPL.

Your right.  Sorry, I misinterpreted your reply.  I apologize.

> Still, I'm quite sure that  your interpretetation (the last sentences)
> of the LGPL is not correct: Netscape doesn't distribute libc or any other
> (L)GPLed library. They deliver a Communicator binary that can be linked
> dynamically on your machine by ld.so with e.g. Linux libc. Whoever supplied
> you with the libc on your system is responsible for providing you with the
> source (according to the LGPL), and with that source, you're perfectly able
> to fix that LGPLed libraries. AFAIK: As long as Netscape doesn't include
> LGPLed code in the distribution (which they don't apart from the stubs,
> which is fine with the LGPL), they don't have to provide you with the
> source!
...

I agree that it should be as you describe, however I do not think Richard
Stallman and the free software foundation agree.  Your say that stubs are
fine with the LGPL, but I can't find any mention of dynamically linked
executables (or their stubs) in the LGPL.  Also remember that in order to
use these LGPLed libraries, you must include LGPLed header files directly
in your code.

Here is a quote from Mr. Stallman:

  There is no fundamental difference between static and dynamic linking,
  as regards the GNU GPL or the GNU LGPL.  Our position is that if you
  link a GNU program into a combined program, no matter what kind of
  linking you use, you have to obey the terms of the GNU program in
  distributing the combination.

I pointed out that to me that did not make sense because one does not
have to link at run time with the LGPLed library, one could run with any
library that provides the same API as is the case with Motif/Lesstif.  He
did not wish to address that issue however:

  You are raising a scenario in which (1) there is an API-compatible
  proprietary library that might be used in the same executable, and (2)
  no header files from the library are used.

  I would have to think more to come up with a position about that
  scenario; but it does not arise in practice as far as I know, so I
  don't need to think about it now.

While I believe in the free software concept and have a great deal of
respect for his accomplishments, personally I believe that Mr. Stallman
may be taking it a bit too far in this case.  I had my e-mail discussion
with him on this issue on the MICO mailing list, so if you are interested
on his position you may wish to check the archives of that list.

Hope this clarifies things,
Carl Thompson