FreeBSD the most distributed Linux in the World(blog.milkingthegnu.org)
blog.milkingthegnu.org
FreeBSD the most distributed Linux in the World
http://blog.milkingthegnu.org/2008/05/freebsd-the-most-used-linux-in-the-world.html
4 comments
developers prefer the GPL
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.
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.
Take Apple out of the list. And, obviously, the Google SoC.
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.
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.
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.
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.
I'll be glad if you don't make me take the time to do it, but I could argue that many, many more companies have built businesses on Linux than on BSD. A huge number of products are built on Linux now. Linux is a more popular embedded BSP OS than BSD is.
Sure, in the case of Linux vs. FreeBSD, you're probably right (although there are certainly more factors involved than just licensing).
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.
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.
Very good point.
But it's not completely ludicrous either.
The "sponsored by" usually means that the developer is actually employed by the company (hence the sponsor) and that he contributed in this capacity (i.e. ebing paid working on BSD for instance).
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.
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.
meant "not to condone officially" ;)
BSD gets an edge over proprietary software because if you distribute changes then you don't have to contribute source. GPL gets an edge over proprietary software because if you distribute changes then you must contribute source. Both models have merit and motive. BSD code is distributed widely within proprietary devices and operating systems. GPL code is distributed in a more homogenous form. However, which licence has the largest sum of installed modules?
I'm not so sure about this guy's policework. For instance, NetApp ONTAP is decidedly not FreeBSD. Juniper uses FreeBSD as a control platform, but neither the core functions nor the user CLI are FreeBSD. I'm not sure why I care what IronPort uses.
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.
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.
> 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.
I've noticed that FreeBSD works more responsively in low memory environments. Perhaps that's the VM tuning.
I've stopped visiting milkingthegnu.org, as the inflammatory always outweighs the insightful...both are present, but I always have to fight my way through every post.
OS X isn't even really FreeBSD, parts of it are based on FreeBSD (but the kernel isn't).
That's not exactly true. xnu/bsd/kern, the "Unix half" of the kernel, is FreeBSD. xnu/osfmk, the "Mach half", isn't.
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.