FreeBSD the most distributed Linux in the World
blog.milkingthegnu.org
blog.milkingthegnu.org
I think the reason it's used in a lot of hardware and OS X is because of the licensing. They don't need to release the source code after they've used other people's work for their own product.
I suppose that's OK since the developer agreed to the license but the rapid explosion of Linux is in my opinion indicative that developers prefer the GPL even with its flaws.
Speaking as a FreeBSD developer: I prefer the BSD license. :-)
In all seriousness, while Apple hasn't set a great example for giving code back to FreeBSD, they have been very good at sending bug reports back -- and really, a report of "it looks like you're doing bazzle wrong on line 234 of foo.c" is often more useful than a patch which fixes the bug but doesn't explain why.
Looking at FreeBSD commit mail for the past 4 years, I see commits which were marked as "Sponsored by:" ActiveState, Apple, Chelsio, DARPA, Dell, Ensure Technology Ltd, Ethon Technologies, Google SoC, IronPort Systems, Isilon Systems, Juniper Networks, McAfee Research, Myricom, NetApp, nCircle Network Security, Nokia, Pair Networks, Sandvine, Seccuris, Secure Computing, Sippy Software, Soekris, Sophos, SPARTA, SPAWAR, Timing Solutions, and Yahoo (and I know committers don't always record who sponsored their work in commit messages, so there's certain to be more than just these companies) -- so the notion that the BSD license somehow means that companies don't "give back" to the community is quite ludicrous.
However the code itself often amounts to patches and the company itself (Apple in this article) has chosen not to condemn officially FreeBSD by making a direct financial donation to the foundation.
Now strike DARPA, SPAWAR, Secure Computing Corporation and McAfee Research as well --- that's TrustedBSD, a government-funded project that was required to give its code back (the "McAfee Research" there refers to TIS Labs).
Now strike Chelsio, Dell, Myricom, probably Pair Networks, and Timing Solutions, which is driver code.
Now strike Ethon Technologies, Ensure Technology, and Seccuris, companies nobody's ever heard of with no IP to donate to the project.
What do we have left? ActiveState, IronPort, Isilon, Juniper, NetApp, nCircle, Nokia, Sandvine, and Yahoo.
* What did ActiveState contribute?
* Did IronPort contribute back its custom filesystem and I/O scheduling code?
* Did Isilon contribute back its kernel filesystem clustering code?
* Did Juniper contribute back its networking code? Wait, I know the answer to that one.
* Did NetApp contribute back its WAFL filesystem?
* Does nCircle even have anything to contribute?
* What did Nokia contribute? I know their Checkpoint appliances run FreeBSD.
I sound snarkier than I really mean to here, but I think the list is a bit misleading. Of course, more companies contribute back to Linux than contribute back to FreeBSD; they're obligated to.
I don't think it's as simple as that. The license has a major influence on the ecosystem around the project, and it is that ecosystem that determines the contributions that flow back to the open source code. The GPL encourages a certain type of ecosystem, and BSD another; that the GPL will automatically yield a healthier ecosystem than BSD is not at all obvious to me.
One of the major differences is that with the BSD license, companies are free to build proprietary products on top of the open source code, which opens up a lot of business models that aren't feasible with the GPL. Take Postgres, for example -- there are many companies that have used Postgres to build a proprietary database of some description. Some of these companies choose to contribute changes back, or to employ Postgres developers full-time, or to hire them as consultants -- mandating that all changes be contributed back would probably mean the companies wouldn't exist in the first place, or wouldn't be using Postgres, which would probably be a net loss for the project. And if a company chooses not to contribute code back? Well, personally that's fine with me, but I can understand why others might disagree.
My point is that requiring code to be given back doesn't necessarily yield a healthier ecosystem, or even more code contributions, than a more laissez-faire attitude.
From several years commercial reversing experience, my guess is that even if you lump NetBSD in with FreeBSD (NetBSD is the preferred BSD OS for non-X86 BSPs), you still see more Linux than BSD in embedded and appliance apps.
From several years shipping appliance products, I'll also suggest that while there may be apps that FreeBSD excels at (due to a better tuned filesystem and VM), for things where raw I/O and compute are the bottleneck, Linux tends to win.
I've noticed that FreeBSD works more responsively in low memory environments. Perhaps that's the VM tuning.