| Paolo Bonzini 2005-09-20, 7:03 pm |
|
> You are thinking of a "program" as a bunch of source code, I was intending
> the word to mean "what the user runs". So if I deliver an application as a
> shell script (say) that starts GST with a .im file into which my code is
> already loaded, then that .im file will also contain GPLed code.
That's FUD. The image file won't contain GPLed code. This applies to
other Smalltalks as much as to GNU Smalltalk, because I am careful that
every bit of general purpose code is in LGPL'd files, and not in GPL
files.
> first place). It seems (to me) that my program would at least as much a
> "derivate work" as if it had made use of a GPLed DLL -- perhaps even more so
> since the "copying" involved in loading both my code and the GPLed code into
> the same VM produces a more intimate mix than that involved in copying a DLL
> into a process's address space.
There's no difference between these two cases as far as the GPL is
concerned.
> The switch to LGPL for the core classes /may/ remove these problems. I,
> personally, think it possible, but the bottom line is that I do not trust the
> LGPL -- it requires too much reading and re-reading and its applicability to
> languages like Smalltalk (or Java) is not clear enough for me to feel any
> confidence.
On the other hand, this is not FUD. I'm not happy with the language of
the LGPL as well, but after talking (and it is not easy to sustain the
conversation...) with Stallman and also with FSF lawyer Eben Moglen, we
came to the conclusion that you can apply it to Smalltalk or Java, and
that's reflected in the manual section that Bruce has been citing.
Smalltalk is remarkably similar to a statically linked C program. Java
is remarkably similar to a dynamically linked C program.
Paolo
|