On Tue, 2006-06-20 at 18:52, jdow wrote: > From: "Les Mikesell" <lesmikesell@xxxxxxxxx> > > On Tue, 2006-06-20 at 17:46 +0200, Ingemar Nilsson wrote: > > >> And if it depends on GPL-covered > >> code, it has to be licensed under the GPL as well. Simple as that. > > > > It's not simple at all. The program may run only when linked to > > another person's own copy of the library in question. That's > > the way copyright and every other license works. Note that there > > is no restriction against writing and using your own copy of this > > work, or hiring someone to write it specifically for you. The claim > > is that this original work can't be distributed because it won't run > > without the GPL'd part - as though running is some kind of requirement > > for code distribution or that copyright is somehow dependent on code > > working correctly... > > And it's not even that simple, Les. The GPL faq apparently declares > that if you static link or even dynamic link then your code must be > under GPL but if you can link over a socket then your code is clear. > There seems to be an attempt to draw a line here where there really > should be no line. Either I can use their code or I cannot. There are no restrictions on use and there can't be because you don't have to agree to a license before obtaining your copy of the works. The restrictions apply to distribution of more copies. > The > manner in which I use their code should not matter in any logical > world. They can only restrict what copyright law would restrict if the license does not allow it. Thus someone has to imagine that the result is a derived work for any restriction to apply. There is an obvious case when covered content is actually being copied as would be the case for in-line inclusion or statically linked libraries, and it is pretty obviously not a derived work when a totally separate process is running in its own protected memory space even if it does use some sort of interprocess communication. It's pretty fuzzy for the dynamic linkage case where no covered code is copied and really far out when the application can optionally dlopen() libraries (or not) at runtime that don't need to be known at compile time. The FSF's interpretation seems to be that anything that is loaded into the same process memory becomes a derived work even if none of the authors involved knew that might happen. -- Les Mikesell lesmikesell@xxxxxxxxx
