[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Eclipse 3.0 Running ILLEGALY on Kaffe



Walter Landry <wlandry@ucsd.edu> writes:

> Måns Rullgård <mru@inprovide.com> wrote:
>> Walter Landry <wlandry@ucsd.edu> writes:
>> > What if there was a package wget++ that communicated with openssl
>> > entirely through system() or exec() calls?  It would construct
>> > appropriate input and parse openssl's output.  Would that constitute
>> > linking?  It ends up using all of the same code as the directly linked
>> > version.
>> 
>> The Gnus newsreader, running under Emacs, does precisely this, and has
>> done so for many years.
>> 
>> > If it is not linking, why couldn't you do this with all GPL'd
>> > libraries?  You could write a GPL'd wrapper around a library, and just
>> > use the wrapper with exec().
>> 
>> A while back, the general agreement seemed to be that this would be
>> allowed.
>
> I don't doubt you, but when was that?  It would be good to see
> surrounding discussion.

Sorry, I didn't have time to dig the archives.  I was thinking of the
thread starting at
http://lists.debian.org/debian-legal/2004/11/msg00008.html, though I'm
sure it's been discussed more than once.

>> > In essence, why does using exec() suddenly break the chain, while a
>> > linker or classloader does not?
>> 
>> I don't see an obvious difference, but the GPL FAQ does mention this
>> distinction.
>
> I think this is really the point of contention.  There is no good
> reason to say that exec() is a magic boundary.

The GPL FAQ also says that exec() isn't always the boundary, it
depends on what the programs communicate with each other.

> That is why I don't see the difference between linking in the C
> sense and calling libraries through an interpreter.

Interpreters are a different issue from the exec() situation.  The
program being interpreted generally does not communicate with the
interpreter at all.

-- 
Måns Rullgård
mru@inprovide.com



Reply to: