Gentoo/FreeBSD On Hold Due To Licensing Issues

Alan Trick writes "Flameeyes (a Gentoo/FreeBSD developer) recently came up with some serious problems among the various *BSD projects who use BSD-4 licensed code (which is all of them). Even other projects like Open Darwin may be affected.

The saga started when he discovered the license problems with libkvm and start-stop-daemon. "libkvm is a userspace interface to FreeBSD kernel, and it's licensed under the original BSD license, BSD-4 if you want, the one with the nasty advertising clause." start-stop-daemon links to libkvm, but it's licensed under the GPL which is incompatible with the advertising clause. The good new is that the University of California/Berkley has given people permission to drop the advertising clause. The bad news is that libkvm has code from many other sources and each of them needs to give their permission for the license to be changed.

At the moment, development on the Gentoo/FreeBSD is on hold and the downloads have been removed from the Gentoo mirrors."

Someone didn't read his next email...

[Zaurus]

Um, hello? Whoever submitted this basically took the original email that Flameeyes sent, but ignored the next one that came a few minutes later:

On Sunday 07 January 2007 02:47, Diego 'Flameeyes' Pettenò wrote:
> This is a very sad blog by my side, although I hope this can be cleared up
> soon so that I don't have to be this sad anymore in the future.
Edit: Timothy (drizzt) found us the escape route. Applying
ftp://ftp.cs.berkeley.edu/pub/4bsd/READM E.Impt.License.Change we can legally
drop the clause 3 of 4-clause BSD license, and be done with it. I'm writing
in this moment the code to do this, but it might require a new stage to come
out. Anyway, the problem is solved, and I think I'll mail FSF for them to
actually put that note somewhere, as it doesn't seem to be that documented
around here.

This *should* cover our asses about the problem, although I'm still looking if
there are sources that are redistributed under 4-clause BSD license, and for
which the license change is not effective (i.e.: they are not under UCB
copyright).

Re:hmmm

[First Person]

For those who need the (admittedly weak) joke explained, try this [everything2.com].

apple phasing out libkvm support

[oudzeeman]

OS X for Intel shipped with /dev/kmem disabled by default, which breaks libkvm (they kept the libkvm header and object files around though, and /dev/kmem can be re-enabled through a kernel boot argument). It is expected future versions will drop the support completely. It also appears that OS X on PPC based on Darwin 8.8.1 or newer also has /dev/kmem disabled by default.

I just had to remove all dependencies on libkvm for a project I work on, since we recently had our first users try to use it on OS X x86. It is software used on HPC clusters and SMPs, so there hadn't been much interest in OS X x86 until the Xeon XServes. I had been trying to get a hold of an x86 system to test on for months, and then this problem hit us.

Obviously this could affect OS X/Darwin until they completely phase this out and remove libkvm objects and headers from the software distribution.

FUD

[brass1]

FUD, plain and simple.

1. The clause that's being referred to is clause three which states:

3. All advertising materials mentioning features or use of this software must display the following acknowledgement:

The operative phrase being, "mentioning features or use of this software." Somehow I doubt there's so much with mentioning the features or use of libkvm no matter what the actual meaning of the word advertising is.

2. I've gone through all 15 of the .c files in my FreeBSD tree, exactly 2 of them have what *may* be a non-waived clause three: kvm_arm.c, and kvm_powerpc.c. The rest of the files are either copyright the Regents, don't have clause three, or use the CMU license.

The two files are copyright Wolfgang Solfrank and TooLs GmbH. I would submit that there is probably a clause three waiver from these folks; it's just that we haven't found it yet. Also, removing the two effected files would have no effect on functionality. Neither the ARM or PPC ports are functional.

The FUD here may not have been intentional, but it is FUD none the less.

Re:But wait a minute...

[The_Paya]

Actually, start-stop-daemon is Public Domain, the problem lies on the 4-Clause BSD license of the whole thing, RTFA and you'll see that there are about a hundred different "mentions" to make regarding advertisement, creating a whole mess for *any* derivative work of FreeBSD, and, perhaps, even for FreeBSD itself.

More information at: http://en.wikipedia.org/wiki/BSD_Licenses [wikipedia.org]
And at: http://www.gnu.org/philosophy/bsd.html [gnu.org]

Re:Someone didn't read his next email...

[multipartmixed]

You're fucking kidding me, right?

I'm still using the same OS I did five years ago.. but I've upgraded through SEVERAL versions of gcc...

> Besides which, a recompile every month or so is good for your system.

What, is that the Gentoo equivalent of rebooting a Windows box? *!@*!*@!!

Re:Trouble in paradise

[JoshJ]

As far as I'm concerned, use of public domain material in a program puts that program into the public domain also.

The lawyers would disagree. This is why we have copyleft in the form of the GNU GPL.

Re:But wait a minute...

[drmerope]

Just so. The anti-4 clause movement began with the FSF. Let me stake out the position that they are not entirely objective on this point. The imfamous clause 3 problem was and has always been a canard.

What's amazing is that people cite to the FSF propoganda and conclude they've prove their point.

Well here is the truth of the matter: Clause 3 relates particularly to advertising that discusses the features implemented by the code given in clause 3. What this means is you want to brag about softupdates and softupdates were covered by this imfamous third clause, you would have to say 'as implemented by Kirk...'

Anyways, this only applies to advertising with sufficient specificity to implicate particular code. Basically if you can trace a feature to 100s of contributors the clause is self-invalidating. No one contribution was responsible for the feature discussed in the advertising, therefore no mention is required.

The whole topic has been FUD for twenty years. That said, it has been such good FUD that people have actually taken extensive effort to purge the clause from the standard license. Only a few small files retain it today.

I think DragonflyBSD which is forked from FreeBSD 4.x is 4-clause free.

Re:But wait a minute...

[arkanes]

It's not FUD - you seem to be claiming that it's a minor requirement and no big deal, which (as far as I know) nobody at the FSF has ever disputed. Not a big deal doesn't mean it's GPL compatible, though.

But the GPL specifies *no* additional restrictions, advertising clause is an additional restriction, end of story. That's all there is to it - 4 clause BSD license is not GPL compatible.

The PyDev extensions for Eclipse are distributed under a free license that includes the requirement that you take a deep breath and relax. That's a GPL-incompatible clause. Legal nitpicking is how you keep your house clean.

Re:Goofy

[Anonymous Coward]

The problem is that many developers change the reference to Berkley to themselves or their company, which results in the need for many permutations of the sentence. And the more serious problem is that that licensing restriction conflicts linking with code under the GPL.

And the answer is...

[ByTor-2112]

http://lists.freebsd.org/pipermail/cvs-src/2007-Ja nuary/073415.html [freebsd.org]

All this painful discussion over what is probably a non-issue? Don't you just love this brave new world of 30 blogs linking to each other creating an artificial buzz/panic? Is this a case of premature eblogulation?

"clean interface"???

[r00t]

Dude, you're dumpster diving in KERNEL MEMORY without any locking whatsoever. The data structures are changing as you examine them. A location holding a pointer to a process structure can suddenly change to hold a pointer to the read port of a hardware device FIFO, with any read you do being destructive by stealing the data. Even a page of memory can get remapped at any moment.

Also, this ties you to a specific version of the kernel. It ties you to a specific patch level. If ever Apple changes the layout of a kernel data structure, you're screwed.

You call this clean??? Uh...

In your mind, what exactly does it take for an interface to be dirty? I'd love to see a few examples.