That's why Stallman prefers distributed code copyright: if contributors own their contributions, then dual-licensing won't work. On the other hand, this prevented a project like the linux kernel from moving to gplv3 since the distributed corpus of contributors could not all agree to switch.
Historically the FSF required copyright assignment for FSF works. This was to avoid having any potential problems where a contributor thought they were OK to contribute, but the employer potentially had a "work for hire" claim. (The ease with which such claims can exist varies by local law, and many are not aware of the fact that they have such a liability.)
Historically the FSF required copyright assignment for FSF works. This was to avoid having any potential problems where a contributor thought they were OK to contribute, but the employer potentially had a "work for hire" claim.
How would that work? If you don't have the legal standing to contribute to the project, you certainly can't reassign the copyright to another party.
The copyright assignment is so that the FSF can sue if someone break the project's license. They also require a disclaimer that the work isn't for-hire, but that's a different issue.
The process of getting the employer disclaimer mentioned in your link exists exactly to catch cases where the contributor was mistaken about their ability to legally contribute.
There is a third path, like that of the OpenStreetMap project. Originally it was released under a CC-BY-SA licence, but in 2012 was changed. As many users as possible were contacted and asked to agree or decline the new licence (ODbL). Eventually some data had to be deleted.
In response there is a Contributor Terms for all people wishing to add to OSM, which allows the OSM Foundation the right to relicence to a new free and open licence that 2/3 of active contributors agree to.
This means the copyright is owned by many people, not one organisation/company, but that there is a way to relicence without requring 100% acceptance.
> On the other hand, this prevented a project like the linux kernel from moving to gplv3 since the distributed corpus of contributors could not all agree to switch.
That's not enitrely true.
If I remember correctly Linus himself does not like the GPLv3 and opposes its use for the Linux kernel.
Basically his argument is that the changes and new restrictions imposed by GPLv3 in practice makes it a new kind of license, but that the naming seems designed to "lure" people who think it's a regular license upgrade into adopting a license they would otherwise not consider using.
I'm not sure if this is in writing anywhere, but I recall him talking about this on the same Debconf where he admitted that he had never managed to install Debian Linux.
So moving the Linux kernel to GPLv3 may not be entirely trivial, but that's not the reason it hasn't been done. Linus simply didn't want to.
Yes, the GPL comes in two flavors. If copyright is assigned to one holder, than that holder has special rights. Linux does things differently, so much that Linux can't possibly be dual licensed.
MySQL and Linux have both been big successes as well as Apache Hadoop, which is under an Apache license. There are a lot of choices that can work out fine for open source.
