On Wed, 2009-06-24 at 21:36 +0200, Ronald Vanschoren wrote: > > Note: Linux kernel modules need to be GPL, too - the only way to have > > non-GPL drivers in Linux is to have userspace drivers, which are quite > > limited in capabilities. > > > > This is not correct. GPL v2 talks about "derived work". It's well > accepted (by Linus himself and most of the community except the holier > then pope people) that binary kernel modules (so not releasing source > code) is acceptable IF the original driver was NOT written for Linux. If > it is merely ported to Linux, only the portation layer has to be > released. I think the NVIDIA drivers are an example. I know a company > who uses such binary drivers and was in contact with FSF who approved > the approach.
This is _very_ apt observation, and one I almost forgot myself. > To stay on topic, following this (FSF approved) interpretation of GPL > v2, FTD2XX is NOT a derived work of OpenOCD so it should not be released > under GPL. As (I think) you say here, the Linux kernel module scenario does not affect the legality of a loadable module in the same user-space process as OpenOCD. The module still violates the GPL when distributed, since it would be derived in part from the type and API definitions provided by the driver interface header files. Thus, binaries of the driver module could not be distributed. However, I have mentioned that modular drivers could allow a build-kit to only produce that single driver from source, though the "build" requirements still get out of control. The latest "link-kit" idea would be perfect for this though! Modular drivers are still a Good Thing. Cheers, Zach _______________________________________________ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
