Matthew Garrett: How-To Drive Developers From OS X To Linux

[Del_]

Well yes the BSDs have been marginalized, but this basically has to do with historical reasons, as opposed to the license in and of itself:

Linux came about in 1991, whereas the BSDs took until 1993 to really become public, this has two major consequences.

I have seen this argument on occasions. I do believe it fails to take in the wider picture. The wider picture being BSD and GPL living side by side since 1984, while BSD code was distributed from 1977, adding on software components continuously since then. Hence I believe you are doing extreme cherry picking focusing only on the kernel. Gnu was a successful project before the kernel came about. Moreover, Linus is not the only one to embrace GPL, I believe you will find quite a lot of us being inspired by the license. Hence, for many contributors GPL is a concious choice. Canonical's choices are tightly linked to Debian, Mark being an old Debian developer. Debian itself very consciously uses GPL for its own code, while allowing BSD code into their repos without resistance. It seems you regard this as a throw of dice, while for me it is very much one approach being superior to the other almost across the board with three decades of abundant empirical evidence to prove it. Actually, I have a very hard time seeing how anybody can believe it was all by chance. Maybe because you don't believe in the necessity of copy-left yourself? Then you should consider the possibility of being a minority, and what that means in terms of developer mind share.

NetBSD also has ZFS since 2007 and it doesn't seem to be helping it's share any. So I'm rather inclined to think that it has nothing to do with ZFS.

My thoughts here was on FreeNAS, which I believe would be based on Debian rather than FreeBSD today if it hadn't been for ZFS. It is the only noteworthy usage I recall of FreeBSD in the open world. I have never seen NetBSD do anything noteworthy. With btrfs maturing and owncloud entering the scene, I believe FreeBSDs role on file serving has a grim future. Then what is the purpose of it? To keep Sony in the game against Steambox? If that is what it all boils down to, how meaningful is that? Certainly nothing I will spend my precious days on this planet doing.

[Luke_Wolf]

I have seen this argument on occasions. I do believe it fails to take in the wider picture. The wider picture being BSD and GPL living side by side since 1984, while BSD code was distributed from 1977, adding on software components continuously since then. Hence I believe you are doing extreme cherry picking focusing only on the kernel. Gnu was a successful project before the kernel came about.

Sure BSD and GNU were doing their own thing all the way back then, both being successful projects however it's disingenuous to remark that the BSDs are marginalized, which I responded to was for historical rather than license reasons, and then claim I'm the one focusing on the kernel, when my original comment was that the BSDs are a perfect example of proprietary, open source, and forks of open source all living together peacefully under a permissive license.

Further are you really proposing that even if the BSD had been publicly available in 1991 (which means Linus Torvalds by his own admission would have gone with it) instead that either GNU Hurd would have succeeded or someone would have inevitably made a different GPL licensed kernel that actually took off? I fail to see enough people caring about the license as opposed to something that is "Good Enough" from a technology perspective.

Moreover, Linus is not the only one to embrace GPL, I believe you will find quite a lot of us being inspired by the license. Hence, for many contributors GPL is a concious choice.

Enough to matter in the face of those who would see BSD as "good enough" and really just wanted an OSS kernel?

Canonical's choices are tightly linked to Debian, Mark being an old Debian developer. Debian itself very consciously uses GPL for its own code, while allowing BSD code into their repos without resistance.

And why wouldn't the debian guys have been okay with a BSD licensed kernel? They don't seem to be having any problems with it right now. Debian/kFreeBSD is a thing. Do you really believe that they wouldn't have just said good enough?

It seems you regard this as a throw of dice, while for me it is very much one approach being superior to the other almost across the board with three decades of abundant empirical evidence to prove it.

On the contrary I fail to see any evidence to the idea that the copyleft approach beats the permissive approach. Technical superiority is generally the determining factor, although the superiority needs to defeat the "Good Enough" factor of the entity that was first to market.

Linux for example was first to market, and by the time the BSDs were publicly available, "Good Enough" and the legal uncertainty of the BSDs at the time played to Linux's advantage.

Actually, I have a very hard time seeing how anybody can believe it was all by chance.

unless you wish to propose that some sort of God set up the perfect storm for Linux to succeed, yeah Linux's success was chance and primarily relied upon being first to market.

Maybe because you don't believe in the necessity of copy-left yourself?

Now what have I been saying this whole thread? No I don't believe it's necessary. The whole idea that it is necessary is under the basis that you need to beat code out of people in order for OSS software to work as a system. Instead I don't think you'll magically find more developers willing to contribute meaningfully by cutting out the people who make their living off of software licensing as a business model. You will however likely gain some by allowing those people to use it and contribute, because it's a royal pain to maintain patches downstream. For example the various things like AVX support that sony contributed upstream to FreeBSD.

Then you should consider the possibility of being a minority, and what that means in terms of developer mind share.

Actually I'm pretty sure most developers are like Linus, not RMS. Pragmatists not religious fanatics.

My thoughts here was on FreeNAS, which I believe would be based on Debian rather than FreeBSD today if it hadn't been for ZFS. It is the only noteworthy usage I recall of FreeBSD in the open world. I have never seen NetBSD do anything noteworthy. With btrfs maturing and owncloud entering the scene, I believe FreeBSDs role on file serving has a grim future. Then what is the purpose of it? To keep Sony in the game against Steambox? If that is what it all boils down to, how meaningful is that? Certainly nothing I will spend my precious days on this planet doing.

The BSDs currently provide the only actual working competitor to Linux, and provide us the benefit of the research they do, by being groups of people concentrated on certain areas under a license anyone can use. For example, without the existence of OpenBSD the people who go obsessive on security wouldn't have been concentrated into a large group, and OpenSSL would have been left to continuing being a horrible mess instead of being forked into LibreSSL. DragonFly is a bunch of people who are focused on doing kernel design "Right", and as a result have done some very interesting concurrency work, HAMMER and HAMMER2 basically have the goal of being a BSD licensed version of ZFS done right, and they're transitioning the BSD kernel into a microkernel design. FreeBSD is focused on being fast and general purpose which basically gives it the same goal as Linux while providing a far greater resemblance to UNIX than Linux has (for obvious reasons). It provides the benefit of allowing companies that would have otherwise developed proprietary OSes from the ground up to instead adopt an open core model, don't ask me why said companies bother because I don't know.

[Del_]

And why wouldn't the debian guys have been okay with a BSD licensed kernel?

They are, you already know that Debian gladly supports BSD as a kernel. I already told you that in the previous post, they take inn BSD code in their repos without resistence, although they do choose to go with GPL on their own code. It is the other way around, it is the BSDs that try to avoid copy-left, even giving its users an inferior product over it. Now who was the fanatic did you say? BSD still has every chance to prove itself, regardless of Linus. You put way too much emphasis on linux I believe. For instance, Debian is a democratic movement, and I believe it would have chosen GPL regardless of linux.

The BSDs currently provide the only actual working competitor to Linux,.

Not really, linux has won. Microsoft and Apple pose the only challenge, and they are losing. The gap between BSD and linux is widening everyday. Linux is lightyears ahead, regardless of what metric you use.

and provide us the benefit of the research they do, by being groups of people concentrated on certain areas under a license anyone can use. For example, without the existence of OpenBSD the people who go obsessive on security wouldn't have been concentrated into a large group,

Not true, they could easily have made their own linux distro. Just like all us network freaks gather around openwrt. Actually, I believe the OpenBSD crowd would have come much further if they embraced copy-left, but chances of that happening is zero I guess.

[Guest]

The BSDs currently provide the only actual working competitor to Linux, and provide us the benefit of the research they do, by being groups of people concentrated on certain areas under a license anyone can use.

BSD isn't competition to Linux. It was in early 2000, but not anymore.

DragonFly is a bunch of people who are focused on doing kernel design "Right", and as a result have done some very interesting concurrency work,

Yeah, they copied the Linux way, so it's nice that you agree Linux is designed "right". However, DragonFly is an only sane BSD operating system. The one that sucks less. It is what FreeBSD should be.

For example, without the existence of OpenBSD the people who go obsessive on security wouldn't have been concentrated into a large group, and OpenSSL would have been left to continuing being a horrible mess instead of being forked into LibreSSL.

OpenSSL is OpenBSD mess, so I bet it will be the same with LibreSSL.

HAMMER and HAMMER2 basically have the goal of being a BSD licensed version of ZFS done right, and they're transitioning the BSD kernel into a microkernel design. FreeBSD is focused on being fast and general purpose which basically gives it the same goal as Linux while providing a far greater resemblance to UNIX than Linux has (for obvious reasons). It provides the benefit of allowing companies that would have otherwise developed proprietary OSes from the ground up to instead adopt an open core model, don't ask me why said companies bother because I don't know.

Last time I heard FreeBSD doesn't aim to be fast (and benchmarks show it's not). If it has the same goals as Linux it doesn't help - it's years behind it. It's not possible for BSD to be more successful than Linux, because Linux is GPL licensed which gives much more freedom to users and Open Source software world. The BSD license is anti-progressive one. It exist to support proprietary software and make some fools to work for MS and apple for free.

[TeamBlackFox]

<snip>

First off, OpenSSL is not a product of OpenBSD. Despite the "Open" prefix, OpenBSD never was in control of the project. Also as you know OpenSSH is used on every UNIX and UNIX clone out there, including Linux. The alternatives are proprietary.

Linux is not a UNIX and never will be, its a clone, but its userbase, functionality and dependency on GNU makes it its own standalone. Its kernel is not based on a BSD, Research UNIX or System III/V kernel, but was designed in response to MINIX, another UNIX clone. I'm sure you probably know this.

Also FreeBSD is not associated with Apple nor Microsoft officially, I work for Microsoft as a contractor - and I know this. Stop spouting lies. You know nothing of FreeBSD and it sure as hell shows.

Dragonfly is very good, but unfortunately it lags in a lot of fashions, plus HAMMER will have to mature some more before I consider using it in a production machine. FreeBSD is the most practical BSD out there right now for a desktop BSD, and it makes some robust servers too.

I used Linux for 7 years prior to moving to BSD, and thats because the entire system feels held together with duct tape, and GNU is fanaticist. Thats why I frown at GNU/kFreeBSD, but hey the BSD license is a CYA version of public domain, so I'm all for it. Anything but the GPL.

GNU/Linux owes its success to the fact that, at the time of BSD, AT&T wasn't very pleased to hear that BSD was going to be released open source. And the Linux kernel was developed while the GNU Hurd was at a standstill. Meanwhile, AT&T, like SCO did in the 2000s, they fought BSD for what they saw as theft of intellectual property. This mitigated BSD for almost 3 years, so of course GNU/Linux has won that race. However, the BSDs will win out in the long run as GNU/Linux is making itself less and less unique and the GPL is causing the entire open source software community to teeter in a very precise balance. The GPL has long term implications on businesses ability to make money using a GPL product. Plus BSD has had a surge of interest in the userbase, so its not by any means a dying OS family.

[Luke_Wolf]

They are, you already know that Debian gladly supports BSD as a kernel. I already told you that in the previous post, they take inn BSD code in their repos without resistence, although they do choose to go with GPL on their own code.

Exactly and as a result if BSD had been first to market instead of Linux we'd all be using a BSD instead, particularly given the reputation and history of the Berkley System Distribution at that point in time. The need for a UNIX clone would have been filled and any competing kernel would be worse off than BSD was when it ran up against Linux, due to BSD being more complete and stable at that point.

It is the other way around, it is the BSDs that try to avoid copy-left, even giving its users an inferior product over it. Now who was the fanatic did you say?

The BSDs avoid copy-left for a pragmatic reason as opposed to a religious one. The BSDs all share a common goal of a "No Strings Attached" operating system, and the GPL violates this goal, so while the software may be technically better from a software standpoint the permissively licensed alternative is technically better for the BSDs from a project design goal standpoint (It doesn't matter if software X is the best software in the world if it doesn't match your project goals). However up to a point the BSDs are willing to live with copy-left in their base system if no suitable permissively licensed alternative is available. See them using the GCC (albeit they refused to shift past GCC 4.2) until LLVM/CLang was ready to take over.

BSD still has every chance to prove itself, regardless of Linus.

Correct, and while unlikely it could be that 10 years in the future a BSD will be eating Linux's lunch (the only one I see even potentially pulling that off is DragonFly).

You put way too much emphasis on linux I believe. For instance, Debian is a democratic movement, and I believe it would have chosen GPL regardless of linux.

What license the Debian project would prefer is irrelevant to the discussion of whether the BSDs would have won had they been first. The fact is as we can see from the debian project itself and as you yourself admit the BSD kernel would have been adopted.

The simple fact is that the market that Linux won was not looking specifically for a GPLed kernel (other than the FSF), instead they were looking for an open source UNIX-Clone. The fact that Linux was first to market and had years to establish itself and lock in it's market and then Red Hat rocketed them to the top, ensured that the BSDs didn't have a chance as Linux was "Good Enough".

Not really, linux has won. Microsoft and Apple pose the only challenge, and they are losing. The gap between BSD and linux is widening everyday. Linux is lightyears ahead, regardless of what metric you use.

Let me put this another way for you, the BSDs are the only open source alternative to linux that you can actually use in any real fashion, as opposed to only really being able to be run in a VM, and while they are behind on most significant metrics, you can still use BSD in all of the same roles as Linux just in a degraded fashion. This can't be said for any of the alternative OSS OSes

Not true, they could easily have made their own linux distro. Just like all us network freaks gather around openwrt. Actually, I believe the OpenBSD crowd would have come much further if they embraced copy-left, but chances of that happening is zero I guess.

Not exactly, the kinds of people who are like Theo wouldn't be happy with being just a distribution of linux they would have made their own secure fork of linux and probably would have ripped out and rewrote half of linux.

And again with this presumption of somehow gaining more meaningful contributions through copy-left which no evidence has actually supported. Meanwhile the fact is that proprietary forks do contribute code back upstream to permissively licensed projects, and there are still companies like iX Systems which operate in a similar manner to Red Hat. Which serves as evidence against the need for copy-left.

[doom_Oo7]

Meanwhile the fact is that proprietary forks do contribute code back upstream to permissively licensed projects,

So when can I play PS4 games on FreeBSD ?

[Luke_Wolf]

So when can I play PS4 games on FreeBSD ?

So when can I play Android games on a regular Linux distro?

[Luke_Wolf]

However, the BSDs will win out in the long run as GNU/Linux is making itself less and less unique and the GPL is causing the entire open source software community to teeter in a very precise balance. The GPL has long term implications on businesses ability to make money using a GPL product. Plus BSD has had a surge of interest in the userbase, so its not by any means a dying OS family.

Eh.. If there's going to be any sort of switchover it's going to occur on a technical basis not a 'uniqueness' basis. Which is why DragonFly has a chance that I don't really see the others as having. Further the number of people who need to do kernel level plumbing on a proprietary project are limited at best. So AFAICT this only really effects companies like Sony who want to make massive changes across the board to turn FreeBSD into a gaming OS.

That said I do expect that systemd (from the conservative faction that doesn't like it) + the BSDs now supporting Radeon (albeit they still need to backport power management support) is driving an increased interest. Personally I find the idea of source and packages being first class citizens an appealing concept, but I'm waiting on evaluating PC-BSD in more than a virtual machine until the BSDs support dynamic power management on the radeon driver.

[doom_Oo7]

So when can I play Android games on a regular Linux distro?

Android, you mean the OS with a GPL-licensed kernel, but a BSD-licensed userland specially crafted in order to prevent this ? (this is why people should *really* say GNU/Linux...).

And you don't really run the games on a Linux distro but a VM it hosts, which is being ported to BSD's : https://gitorious.org/freebroid