No, that's not the case. By buying Sun, Oracle owns all the Open Solaris code.
Sun of course could not have open sourced code they didn't own the rights to open source in the first place. It was a long effort within Sun to scrub all the code to make sure. (This is why open sourcing proprietary code is a lot of work and most companies don't bother. Thankfully Sun did.) Some parts had to be distributed as binary blobs because Sun didn't have the ownership (over time there were separate efforts to rewrite these to reduce dependency on binary blobs).
> I don't see how you the copyright holder would not have the right to revoke a FOSS license over it.
Can you provide any examples where this has happened?
According to Section 203 (grants made after 1978), it looks like you have to wait 35 years, and then in a five year window after that, you can terminate.
So there is a way, but the copyright grantees have over three decades' worth of protection.
This law is for specific grantees; like Bob assigns the copyright to Alice (other than as a work for hire done for Alice). Alice is now the copyright holder. 35 years later, Bob has the right to say Alice is no longer the copyright holder.
Bob has five years to do it. Bob must write Alice some kind of letter. The termination date must be no sooner than 2 years, and no later than 10 years, from that letter.
A free software license isn't a copyright grant so the termination business, and its limitations, may be inapplicable. The users of software aren't copyright grantees, and those who are currently non-users aren't anything at all.
Now about the Open Solaris code; while some of it has copyright notices mentioning only AT&T and/or Sun Microsystems, there is a boatload of hits for "Regents of the University of California".
> Even if there were a doctrine that existing redistributors were grandfathered forever, so they could continue making copies and giving it away, the license would not actually be available to the recipients, who are not existing licensees.
Well, that doctrine is there: if you grant someone a copyright license and they take you up on it, they have 35 years before you can revoke it, in the United States.
It's not clear that third parties continue to have the same privilege after you indicate that the license is no longer available to new licensees.
I doubt such a case exists, but if it does I'd love to learn its story.
At best they could take back the 100% Oracle-owned parts. The community would have to build to fill in the gaps.
They couldn't revoke any files that have, say, a BSD license, even if Oracle (or Sun, before them) had modified them and added their own copyright notice, simply because they would need to contact the original authors and act in concert with them; those authors would likely refuse.
I'm mainly talking about situations when the revoking party is actually 100% copyright holder to the majority of the files implementing the payload functionality, without which there is no software. I don't see how you the copyright holder would not have the right to revoke a FOSS license over it.
Based on what? None of the existing licensees have a signed contract of any kind, nor has any exchange of value taken place; allowing free redistribution and use was an uncompensated, unilateral action.
Even if there were a doctrine that existing redistributors were grandfathered forever, so they could continue making copies and giving it away, the license would not actually be available to the recipients, who are not existing licensees.
The U.S. Copyright Act indeed has the concept of termination:
https://cdas.com/copyright-termination-prime/
There are limitations to it for various obvious reasons, like that you can't terminate the copyright assignment implicit in a work for hire. (Because, doh, you were paid.)