[ale] Topic for discussion

[Joe Knapka]

Christopher Fowler <cfowler at outpostsentinel.com> writes:

> On Thu, Oct 30, 2003 at 08:16:08PM -0500, Doug McNash wrote:
> > I guess it depends on how you define linking.
> > 
> > If "linking" is defined as symbol resolution then it 
> > occurs at compile time (ld) or at runtime (dlopen()).  I 
> > believe that there is another level of indirection on the 
> > dynamic linking whereas the static linking modifies the 
> > object.  
> 
> One good example would be in programs that allow plug-ins.  If
> dlopen() was considered "linking" then one can not write a 
> plug-in for GIMP unless it was under the GPL like Gimp.  I'm
> not sure what license GIMP is under.
> 
> 
> Maybe there are other ways to this. Using IPC to somehow communicate
> to the symbol to execute code in that function but not using any
> form of linking.

IMO these are all essentially the same thing. "Linking" is just "a way
to get different assemblies of code to talk to each other"; likewise
"int 0x80", dlopen(), CORBA, DCOM, RPC, RMI, XMLRPC, SOAP, Python's
"import", etc. These are all specific instances of the same
overarching concept, and I strongly suspect the FSF is going to have a
hard time justifying "linkage" as a special case if the GPL should
ever see the inside of a courtroom. (Linus said long ago that for
Linux's purposes, "int 0x80" doesn't constitute "linkage", even though
RMS wanted it to.)  I also think that if they ever *do* manage to make
some specific definition of "linkage" stick to the GPL, then very soon
thereafter everyone who cares will start doing something like Chris
suggests - invoking code in libraries in the simplest way that
circumvents the GPL definition. So to make the GPL do what they want,
the FSF is going to have to define "linkage" absurdly broadly, and
that will make the GPL hard to take seriously. (On the other hand, I
think the definition of "derived works" as alterations to source is
much more sensible, so the LGPL seems much more reasonable to me.)

Then again, perhaps I'm foolish for expecting any of this legal
claptrap to make any sense whatsoever.

Cheers,

-- Joe

> 
> 
> > 
> > On Thu, 30 Oct 2003 19:58:38 -0500
> >  Christopher Fowler <cfowler at outpostsentinel.com> wrote:
> > >On Thu, Oct 30, 2003 at 07:50:09PM -0500, Doug McNash 
> > >wrote:
> > >>>Does use of dlopen() == Linking?
> > >
> > >To me, linking happens at compile time.  When the program
> > >is actually build.  An executable that uses dlopen() can 
> > >still
> > >function without the symbol it is requesting.  It can be 
> > >programmed
> > >to terminate or to do something else.
> > >
> > --
> > Doug McNash <dmcnash at yahoo.com>
> > _______________________________________________
> > Ale mailing list
> > Ale at ale.org
> > http://www.ale.org/mailman/listinfo/ale
> _______________________________________________
> Ale mailing list
> Ale at ale.org
> http://www.ale.org/mailman/listinfo/ale