Joshua Kinard posted on Tue, 17 Feb 2015 12:46:12 -0500 as excerpted: > On 02/16/2015 13:01, Rich Freeman wrote: >> On Mon, Feb 16, 2015 at 11:02 AM, Joshua Kinard <[email protected]> >> wrote: >>> On 02/16/2015 09:04, Rich Freeman wrote: >>> >>>> Maybe another approach is to just ditch >>>> per-file copyrights entirely (which a random perusal suggests is how >>>> Linux does things), but that would STILL require stripping the >>>> copyright out of these files with all the issues that entails, and >>>> limit our ability to borrow license-compatible code. >>> >>> [M]y understanding is the kernel retains per-file copyrights. This >>> is why the kernel is permanently wedded to GPLv2, because some of the >>> contributors owning those copyrights have died and thus can no longer >>> consent to changing to the GPLv3[.] >> >> Perhaps I should have worded that better. >> >> s/per-file copyrights/per-file copyright notices/ >> >> Obviously the content of individual files will always be copyrighted >> absent a release into the public domain. The Linux kernel just doesn't >> stick notices on individual files that attempt to identify who owns the >> copyright on what. Presumably that also means that if they borrow a >> file from somewhere else they don't care to change the copyright notice >> that was already there (or somehow they manage to avoid the euthusiasm >> stirred up by removing said notices). > > Well, I just sent a patch upstream that adds a new RTC driver to the > kernel, and I added copyright to myself and the guy that created the > original driver that I based off of to the top of the source file (and > its header). So that practice is still used, and akpm recently added it > to -mm with no comment on any of the copyright bits, so I must've gotten > part that right. > > It's probably left to the person writing the specific source file(s) on > how they want to do copyright, as long as they stick to recognized norms > and GPLv2.
The kernel's relatively relaxed per-file copyright and license policy is in the context of git and its record of a rather strong explicit per-commit signed-off-by policy. As a result of the strong per-commit signed-off-by policy, they can be and are relatively more relaxed on a per-file policy, since the sign-off policy requires legal responsibility and the authority to grant default-gpl2-only permissions on anything committed in the first place. As such, any file without an explicit license CAN BE ASSUMED to have the GPLv2 license, and copyright CAN BE ASSUMED to remain with the original author (company in the case of a work- for-hire unless otherwise stated), because that's part of the conditions that are agreed to by the explicit signed-off-by. Since gentoo lacks this sort of formal signed-off policy and in fact has yet to move to git where it could be most easily tracked and enforced (let alone such a policy created and formally agreed in the first place), the extent to which the kernel's relatively relaxed per-file policies could apply to gentoo in its current cvs and policy state is rather limited. IOW, the kernel's policy doesn't apply here, except to the extent that we use it as a goal/model to increase the urgency of the switch to git, and once having done so, creating and adopting a similarly strict per-commit- sign-off basic policy context in which to apply a similarly relaxed per- file policy. -- Duncan - List replies preferred. No HTML msgs. "Every nonfree program has a lord, a master -- and if you use the program, he is your master." Richard Stallman
