So who holds the copyright on these man pages? Did MySQL get copyright assignments from all contributers before Oracle bought them? If not, how do they have the authority to relicense the content?
As I recall, MySQL had a firm requirement for copyright assignment for contributions to MySQL because they dual-licensed it, using both GPL and commercial licenses, and without assignment they wouldn't have the freedom to do that without owning the copyright.
So, yeah, I would expect that they own the copyrights on everything in the man pages and can release them exclusively under closed licenses if they choose.
If the maintainer (whether an individual, for-profit company or a not-for-profit foundation like the ASF or FSF) doesn't hold a copyright to all of the contributed code and documentation, they couldn't release it under an updated license (like the ASL2 or GPL3) or dual-license it for compatibility with other projects without getting clearance from each and every contributor.
If this seems "dodgy", bear in mind a couple of points:
1. You're not "giving up" the rights to your contribution, you are sharing the copyright with the organization. That is, your work is your work and you're still free to do whatever you want with it. You're just giving the organization the same rights, so that work that was once "copyright andyroid" is now (independently) "copyright andyroid" and "copyright whatever-project-you-contributed-to". You don't give up any rights, you're just giving some rights to others.
2. Re-licensing a project doesn't change the license for the already released work. I.e., if Oracle decides to only use a proprietary license for all future MySQL releases, that doesn't post-facto change the license on the previous releases. You (or anyone else) remain free to take the last GPL'ed version and run with it (under the terms of the license under which it was released). As crappy as that re-licensing decision might be, it really only impacts the stuff you use from them or contribute to them after the re-licensing occurs.
I'm not sure if it would hold in court, but it's fairly common for GPL projects to include "or later", specifically for that reason.
or dual-license it for compatibility with other projects without getting clearance from each and every contributor.
If you're releasing under a non-copyleft license, it shouldn't be needed, and if you're releasing under the GPL, that's (arguably) a feature, not a bug ;)
Sorry, but - as a general rule, this is false, at least in common law jurisdictions. Copyright is a transferable property right. If I assign copyright on a work to you, I don't have it any more. (That's a bit of a simplification, ignoring moral rights and glossing over more complex situations like if we're joint authors, but the gist is true).
Now, the FSF agreement (and most other CLAs) has a 'grant-back' clause: after you assign them your copyright, they grant to you a wide perpetual licence to use, copy, relicense, etc. etc. the software. So you can still do most[1] of what you could do before.
But that's a function of that clause, not something inherent in assigning copyright.
[1] Most -- but not all. There's one thing that having the copyright lets you do that a licence never can, and that's enforce it against third parties. You can't sue someone for breach of your copyright after you've signed it away, only the copyright owner can, and that ain't you any more.
> Sorry, but - as a general rule, this is false, at least in common law jurisdictions.
It can work either way: It depends on the particular Contributor License Agreement being used, and what it says.
Some actually transfer copyright, some just grant the project entity a perpetual non-exclusive license to all rights.
IMHO, this is a huge advantage: it guarantees that the maintainer of the project can't run off and so something crazy. The linux kernel allows every author to retain copyright, and there's some security in that: since it is basically impossible to get clearance from even the most significant subset of copyright holders in the kernel, it will forever be GPLv2.
Look at it this way: if Torvalds were evil, he could require everybody to assign him copyright, and then take everybody's work and let corporations pay him to allow them to use it closed-source enviorments. Or even un-GPL the whole thing and sell it to somebody. (IANAL, obviously - not sure about that last one)
Point is, there is security in nobody having controlling ownership of a project,
When I refer to "the documentation" I'm mostly talking about the web docs at dev.mysql.com, which are are far more important than the man pages.
If the man pages are covered by the GPL'd may be a thorny issue. The GPL is not generally applied to documentation.
With an MIT or BSD licensed work, you can add more restrictive conditions, e.g. make a derived work proprietary, or add a "noncommercial use only" clause to your derived work, or whatever else you want. But with a GPL-licensed work, you cannot add additional conditions to the derived work, i.e. you must give all downstream reusers the same permissions you received. That's what's sometimes referred to as the "viral" property.
Unless, as here, you are the owner of the copyrights, in which case you can apply any license you want to the work.
Not being able to relicense under a license not acceptable to the authors of the particular copyleft license at issue is a main element of copyleft, but copyleft licenses often allow redistributing under more restrictive licenses that remain acceptable to the copyleft license author (e.g., the FSF encourages the use of "or any later version" language in GPL licenses, and GPLv3 is in many respects more restrictive than GPLv2; further, GPLv3 allows relicensing with specified kinds of additional terms that may in some ways be more restrictive.)
If you consider what developers are allowed to do with a piece of software, the main element of copyleft is actually not being able to relicense under a less restrictive license on the spectrum of FOSS licenses. The least restrictive "license" of all is no license, no copyright, i.e. public domain (in the US).
Because you can convert MIT/BSD to a closed source license, but you cannot convert GPL to a closed source license, in this sense the GPL is more restrictive than MIT/BSD. You can convert MIT/BSD to GPL (less to more) but not GPL to MIT/BSD (more to less).
I understand that closed source licenses are the most restrictive of all, this is just an alternative perspective on software freedom with regard to developers.