FreeBSD 9.0 Released

An anonymous reader writes "FreeBSD 9.0 has been released. A few highlights include: A new installer, bsdinstall(8) has been added and is the installer used by the ISO images provided as part of this release, The Fast Filesystem now supports softupdates journaling, and Kernel support for Capsicum Capability Mode, an experimental set of features for sandboxing support."

Comments

Clang/LLVM in FreeBSD

[bonch]

As noted in the release notes, FreeBSD 9.0 includes Clang/LLVM, the goal is to be rid of all GPL dependencies by version 10.0. At the 2011 LLVM Developers' meeting, Brooks Davis covered the effort in bringing in LLVM for 9.0 and the work remaining for 10.0 to replace GCC. The move was originally intended for 9.0, but there wasn't enough time to get it all done, particularly due to the thousands of pieces of software in the ports tree that still require work. GPLv3 is cited as the catalyst for all this, for preventing cooperation between free and proprietary software sectors.

Re:Clang/LLVM in FreeBSD

[halfaperson]

Yeah, that's why large IT companies like IBM, Intel and even Microsoft are contributing to the BSD:s and not to Linux. Oh, wait..

Re:Clang/LLVM in FreeBSD

[iggymanz]

just a case of "different", not wrong. GPL can't be used in some cases where BSD licensed code can, for example one can distribute modified BSD code without providing the source code as long as its done the way the BSD copyright mandates.

Re:Clang/LLVM in FreeBSD

[halfaperson]

Oh, you mean sell hardware and support to companies that wants to use Linux on their hardware? But I was just told that's companies didn't want that due to the GPL?

Re:Clang/LLVM in FreeBSD

[Qzukk]

People with entitlement complexes and inability to understand simple instructions get surprised when someone gets angry after they take code that says "You can use this as long as you make the result GPL" and use it without making the result GPL.

They could have written their own or taken someone else's code with a more permissive license like BSD, but suggesting this causes them to react like the guy who defends his use of TPB for his movie watching by declaring he has some sort of right to do whatever he wants, whenever he wants, without having to pay for it.

You want it, you follow the rules to get it buddy. If you don't like the rules, nobody's forcing you to get it.

Re:Clang/LLVM in FreeBSD

[MrHanky]

How would the BSD license have saved LG and others for signing patent licenses with Microsoft? Let me rephrase that: The BSD license would not have helped LG at all.

Also, why the fuck should FOSS users care about what Apple does for their own closed-source OS? Before you say Darwin, consider the fact that not a single soul uses Darwin as his main OS. Why? Because it's shit.

Re:Clang/LLVM in FreeBSD

[Eivind Eklund]

What's wrong with GPL?

It annoys the minority of businesses who feel entitled to the free labor of strangers and don't want to give anything back.

You misspelled "everything" as "anything".

If they want to give 90% back - which is common behavior for proprietary derivatives of BSD licensed codebases - they can't. They have to give back 100%, or stay out.

I mean hey, launching a commercial product with most of the work already done for you, for free, is a nice racket if you can get it. But if the developers intend to allow this, they wouldn't use GPL, they would use a BSD-type license. For reasonable people, this is not a problem. Reasonable people think either "hey, this code is available for free and we have no problem complying with the license, so we can enjoy all the effort that has already been done for us and build on that", or they think "the terms of that license aren't compatible with our business model, or we're afraid of how a court may interpret them, so we can't use that code, oh well, this has not harmed us in any way so we really have no complaint".

If you can't see how the GPL dominating a market can do harm to groups of people that can't use the GPLed version but needs customisations - then you've not thought hard enough about the problem. Think about monopolies and "embrace, extend, extinguish", just performed with software given away gratis with restrictions.

Re:Clang/LLVM in FreeBSD

[causality]

If they want to give 90% back - which is common behavior for proprietary derivatives of BSD licensed codebases - they can't. They have to give back 100%, or stay out.

And if that's not to their liking, the only thing they miss out on is the gratis skilled labor of strangers. They are still free to write their own code under any license they want. I just don't see the problem, unless of course there is a sense of entitlement to something no one actually owes them. That's the only explanation for why anyone would experience any distress over this.

If you can't see how the GPL dominating a market can do harm to groups of people that can't use the GPLed version but needs customisations - then you've not thought hard enough about the problem. Think about monopolies and "embrace, extend, extinguish", just performed with software given away gratis with restrictions.

How does GPL "dominate" a market? By that I mean: what's stopping these hypothetical groups from hiring their own programmers to write their own software that is licensed any way they like? A patent could definitely do that, but the GNU Public License is not a patent. If I am a developer who uses the GPL, how am I "doing harm" to you by not giving you my work for free? Again, only a false belief that you are entitled to my labor would make you feel "harmed" in any way.

What non-patented feature can you name for me in a GPL'ed project that an independent commercial project could not also implement? They would have to write their own code, sure, but if you really believe that constitutes "embrace, extend, extinguish" then you don't really understand what that term means. "Embrace, extend, extinguish" is not possible without closed source and/or patents.

That's too bad (for them only) some people feel offended that they can't just copy-and-paste someone else's code into their project, but nothing is stopping them from using their own original code to match every feature found in any non-patented GPL'ed project.

Re:Clang/LLVM in FreeBSD

[Anonymous Coward]

No, actually.

It annoys and horrifies the thousands of developers that develop for platforms where the GPL is incompatible with libraries they must use or is not allowed by the platform rights holder.

There are many projects that used to be GPL/LGPL that are heavily used in the game developer community that are now BSD/MIT/zlib licensed, and they see even more contributions than they did before because more developers are able to use them for projects. (See Ogre3D, SDL for just of many two well-known examples.)

The GPL may be an appropriate license for some developer communities, but in others, it actually *reduces* the number of contributions and users of a project.

Like most things, one size does not fit all.

Re:Clang/LLVM in FreeBSD

[preaction]

FreeBSD (and NetBSD and OpenBSD) have been around roughly as long as Linux has, since the early 1990s. How do you explain the fact that *BSD is a niche OS most users have never heard of, while usage of Linux skyrocketed and it became something that most Joe Sixpacks have at least heard of if not something they actually use as a Windows alternative?

BSD had patent/copyright concerns from System V that were not fully addressed at the time Linux rose to prominence. This is why you hear "This is the year of the Linux desktop" instead of "This is the year of the BSD desktop". This is basic *nix history here, folks.

It would appear that the GPL is superior in terms of attracting developers and establishing a userbase on standard PC hardware in a Windows-dominated world.

Correlation is not causation.

Re:Clang/LLVM in FreeBSD

[tomhudson]

Apple hired several of the FreeBSD devs, and lots of the code was, and continues to be, given back to the FreeBSD project. Also, in case you haven't noticed it, that Apple-sponsored code includes the rewrite to WebKit that all you Chrome users like so much. You might want to check out the other stuff here and here. Also, if you use CUPS to print anything, thank Apple - they bought the source code from the original developer. Use zeroconf for networking? Thank Apple for their open standard (as opposed to Microsoft's closed one, which wasn't adopted by the IETF).

Re:Clang/LLVM in FreeBSD

[tomhudson]

Why would android devices need file system compatibility with non-android devices? If you want to transfer data, you don't open up the device and remove the memory chip - you do it over wifi or whatever networking floats your boat. Next you'll be demanding that we keep fat12 compatibility for transferring data using floppies.

This is the 21st century. Please get with the program.

Re:Clang/LLVM in FreeBSD

[causality]

There are reasons not to use GPL not having to do with modifying code, but simply running the code. E.g. GPL'd libraries.

Libraries are generally licensed under the LGPL. The LGPL is specifically designed to avoid the imaginary problems you bring up. From that link (emphasis mine):

The LGPL places copyleft restrictions on the program governed under it but does not apply these restrictions to other software that merely link with the program.

If you're going to be childish and call me names like "moron" and "zealot", you should least demonstrate a basic familiarity with the facts. If you feel a need to deal with things that way, it is a sure sign you are reacting emotionally and not proactively evaluating anything reasonably. Against anyone who remains reasonable, you are going to make yourself look foolish. Just for your future reference.

Re:Clang/LLVM in FreeBSD

[icebraining]

You wouldn't even be on the net today if it weren't for BSDs networking stack, which both linux and microsoft use.

That's ridiculous. I'm all for acknowledging BSD's contributions, but you can't possibly claim nobody would've implemented a stack if the BSD project hadn't. It's as ridiculous as saying the FreeBSD project wouldn't have existed until today without GCC.

Of course we would be on the net, someone else would've written a networking stack for Linux and Microsoft would have either written their own or bought one of the companies which sold third party stacks for Windows.

Re:Clang/LLVM in FreeBSD

[icebraining]

So what do e.g. Oracle, SAP and Google if not proprietary software? Cheese?

Of course they use Linux, because they're not ignorant and know they can run proprietary software on it without having to touch their licenses. The GPL only affects derivative works, which userland applications running on the Linux kernel aren't considered to be.

Re:Clang/LLVM in FreeBSD

[hhw]

If they want to give 90% back - which is common behavior for proprietary derivatives of BSD licensed codebases - they can't. They have to give back 100%, or stay out.

And if that's not to their liking, the only thing they miss out on is the gratis skilled labor of strangers. They are still free to write their own code under any license they want. I just don't see the problem, unless of course there is a sense of entitlement to something no one actually owes them. That's the only explanation for why anyone would experience any distress over this.

Or maybe, because the rest of us lose out on the 90% they would have given back?

Re:Clang/LLVM in FreeBSD

[causality]

As an example TPB . It is an equalizing factor to the copyright rule that has been extended by stepping on everyone's rights so they will enforce rule number 3.

A regular individual guy who happens to have some programming talent, and decides to give me the fruits of his skilled labor at no charge, and says I may use it as much as I want and do anything I want with it except for a few reasonable restrictions ... that is a person I respect. He is not asking much. He is in fact giving to me more than he is asking from me. I have no problem respecting his wishes. They are quite reasonable. This person is dealing with me as an equal and doing so with equitable terms.

The RIAA and the MPAA lost this kind of respectability a long, long time ago if they ever had it to begin with. What they want for themselves is not reasonable. What they already take for themselves is never, ever enough. They have this insatiable need for more and more but are not themselves willing to give more and more. They do not want to deal as equals. They want to dominate. The terms they want are extremely one-sided in their favor only and continue to become worse as time passes.

Friend, these two are not in the same boat and do not deserve to be treated according to the same standard. A reasonable person could indeed agree that what you wrote in your post can, should, and often does apply to the *AAs of the world. But I just can't justify treating a generous, reasonable programmer the same way. I have no problem honoring that which is honorable, nor would I refuse to respect that which is respectable.

Re:Clang/LLVM in FreeBSD

[Fri13]

Nope... The operating system in OS X is XNU. Darwin is just XNU + compilation settings and tools. XNU is Server-Client architecture operating system, instead Monolithic like Linux is.

Any system program, library, application program etc, does not work without operating system. Not even your development programs, text editors or others work without operating system.

You can download XNU operating system from Apple Open Source site. You can study code and even FSF has accepted Apple license what is used as Free Software license (still incompatible with GPL). From here http://opensource.apple.com/source/xnu/xnu-1699.24.8/ you can download whole source code for XNU operating system. You get Mach microkernel, I/O Kit, FreeBSD parts of network protocols, filesystems and so on. The microkernel in XNU operating system is 3.0 by version.

Android is just one distribution for Linux operating system. But unlike XNU, Linux is monolithic operating system. And Open Handset Alliance (what develops Android) is currently bringing Linux back to mainline. The Linux what is used in Android has been hosted as well on Linux GIT but it has just been own branch, now it is going to be joined back.

One reason why Android has been so successful, is that people don't know that they are using Linux operating system. Same thing is with Ubuntu, where people has false believes that they use "Ubuntu operating system" while they really use Linux Operating System and distribution called "Ubuntu". If they would know the truth, they would cry and have sleepiness nights wondering what they are going to do because "Linux" is so 'difficult to use'.

It is just sad that most people are victims of marketing propaganda and they don't have a clue what technology is being used and how it works.

Re:Clang/LLVM in FreeBSD

[smash]

I just don't see the problem, unless of course there is a sense of entitlement to something no one actually owes them

Here's one: the GPL is hostile to the development/promotion of standards. If the original reference TCP/IP stack was GPLed, you wouldn't be posting here via TCP/IP.

Re:Clang/LLVM in FreeBSD

[jimi1x]

Now, I've yet to see aviation and medical mission-critical software run on Linux.

I think you may want to revisit your statement. I used to work in a hospital. We had medical equipment that ran embedded Linux. I currently work for a bank. We have ATMs that run embedded Linux. Our CCTV system runs embeded Linux in the cameras. Did you know that a lot of banking mainframes run on Linux? I'd argue completely against your statement that no mission critical software runs on Linux.

Re:Clang/LLVM in FreeBSD

[icebraining]

Sure, except the Linux stack and the Microsoft stack probably wouldn't have been compatible, meaning fragmented networks.

Why? There were multiple compatible stacks already before MS implemented theirs from BSD.

I don't see why exactly would MS or Linux develop a stack which wouldn't be compatible with the networks that already existed and were standards in many places. It's not like the Internet appeared after Windows 95.

And we in fact have plenty of protocols which have different but compatible implementations. Why would TCP/IP be any different?

Re:Clang/LLVM in FreeBSD

[icebraining]

TCP/IP was already a standard in many places before MS got around to implement their stack, and it had different but compatible implementations from the start (according to documents, Stanford, University College of London and BBN all had their own).

It seems to me TCP/IP was a standard because DARPA pushed for it, not because of BSD.

Re:Clang/LLVM in FreeBSD

[evilviper]

The problem is really with GPLv3, which has absolutely onerous restrictions, trying to prevent Tivoization of GPL'd code, or rather remake the world into RMS' dystopian fantasy land. But don't take my word for it... Read Linus' opinion on GPLv3

In a broader context, most companies hate GPL and other copyleft licenses in general (LGPL is usually minimally okay), because it's like a worm... Someone can bring it in without you knowing about it, and all of a sudden, you're subject to terms which may make you go out of business all-together (if one large piece of software is your main product). No such problem with freer licenses. And interestingly enough, there is much less of a problem with proprietary code.

You see, with proprietary code, perhaps you already have a license for if. Perhaps you need to re-negotiate the license to include this new usage. Or perhaps it was completely illegal, and now you have to either go to company X with hat in-hand and negotiate a license for your former illegal use, and continued use going forward, or perhaps you'll continue to use it, and hope you don't get caught.

In any of those (proprietary code) cases, it's just a question of money. Maybe it'll be a lot, maybe it'll be a little, but the company wants your money, and will probably work something out with you, unless you're direct competitors...

With GPL'd code, this doesn't work. If it's a small, one-man project, you can try negotiating a license, and probably get one. But if that one-man is an RMS-esque extremist, or if it's the work of multiple people, too many to possibly track down... No amount of money is enough to allow you to keep your copyright on your own code, likely with many millions of dollars invested in it...

THAT is why the GPL scares companies.
Remember Windows' source code leaking out onto the internet years ago? Open source developers were afraid it was done intentionally, so Microsoft could make the case that other projects stole their publicly available source code, used it without permission, and demand exorbitant, insane license fees. These two situations really are surprisingly similar from a software companies' perspective.

Re:Clang/LLVM in FreeBSD

[evilviper]

How do you explain the fact that *BSD is a niche OS most users have never heard of, while usage of Linux skyrocketed and it became something that most Joe Sixpacks have at least heard of if not something they actually use as a Windows alternative?

Most experts explain it as being because of the AT&T vs BSD lawsuit. Until that was decided, FreeBSD was in murky waters few were willing to go along with. And it happened at exactly the wrong time, when i386 systems were growing in popularity, and people wanted some Unix-like OS to run on it, and really wanted it for free.

Network effects kicked-in at that point. Linux got more developers because it was getting more press (and a lone student writing an OS is a better story than Berkley's largess), and it got more press because it was getting more developers, and it got more press because it got more press.

And the definitive counter-point to GPL supporters, is network services... Anyone can name a million and one network services that became defacto standards. BIND *is* DNS. Sendmail *is* SMTP. The BSD TCP/IP stack *is* the internet protocol, and it's bugs and limitations have become the standard.

The most recent example is OpenSSH. It wasn't FreSSH that gained 98% market share in a few years... Nope. And until OpenSSH, crypto was massively overdue, yet none of the alternatives caught-on... Licensing had a hell of a lot to do with that... always does.

NFS was in the same boat... Sun released NFS with an open license (not GPL'd), and it became the standard. NFSv3 was massively crufty and overdue for replacement, yet the dozens of GPL'd network file systems with modern features ever caught on... NFSv4 finally came out, with the main implementation under a free license, that finally made progress.

Re:Clang/LLVM in FreeBSD

[FooBarWidget]

I believe his point was that BSD'ed code promotes standards by allowing anybody to easily incorporate the code. Suppose that the original TCP/IP stack was GPL'ed. Others could still have written their own TCP/IP stack, but would they have bothered to do that or would they just have invented their own proprietary standards instead of bothering with TCP/IP at all? BSD'ing the code makes TCP/IP the path of least resistance.

I don't necessarily agree with this point of view but I can see the reasoning.

Dennis Ritchie

[ShanghaiBill]

The FreeBSD Project dedicates the FreeBSD 9.0-RELEASE to the memory of Dennis M. Ritchie, one of the founding fathers of the UNIX operating system. It is on the foundation laid by the work of visionaries like Dennis that software like the FreeBSD operating system came to be. The fact that his work of so many years ago continues to influence new design decisions to this very day speaks for the brilliant engineer that he was.

May he rest in peace.

Re:Dennis Ritchie

[muon-catalyzed]

more importantly, the creator of C language

Memory Requirements

[AmberBlackCat]

Last week, I downloaded Fedora Core 16 and found that, for the first time, I was not able to update Linux on my Inspiron 8200. Because it has 512 megs of RAM and that install required more. Not sure why an installer requires 768 megabytes. So anyway, maybe that's a sign I should look at BSD.

Re:Memory Requirements

[halfaperson]

Maybe it's a sign that you should consider a new computer?

Re:Memory Requirements

[smash]

Thats not really bloat, It uses ram/cpu for checksums/de-dup/compression and cache. Its by design. If you don't require/want those features, dont enable them/live wth slower i/o or run ufs.

Re:Memory Requirements

[Anonymous Coward]

2GB RAM per TB HDD is only if you're running dedupe. I'm running 10TB with 4GB ram w/o dedup, but with compression on. No issues...fast and reliable.

EC2 AMIs available

[cperciva]

FreeBSD 9.0-RELEASE machine images for Amazon EC2 are available for m1.large and larger instance types: http://www.daemonology.net/freebsd-on-ec2/

Re:woohoo

[Anonymous Coward]

I used NetBSD and/or FreeBSD between 1995 and 2005 and Linux between 1996 and today. By around 2000 Linux was far from a basement project of amateur code, being built primarily by full time developers. The stability of the more mature distributions (go Debian!) matched or exceeded the BSDs, the latter fast losing any remaining technical advantages.

As to "no comments or documentation", you've just revealed that you haven't tried writing in kernel space for either. Linux has been superbly documented for those who, say, wish to write a device driver, while last I gave up on the BSDs it was still a matter of "copy existing code". This works excellently as long as you've decided to throw all engineering principles out of the window and don't understand the difference between stable interface and implementation dependence. Like I said, the BSDs have remained deliberately cliquish, like some stupid nerdish club: to contribute effectively you have to catch the eye of and be guided by existing team members, who will fill in the details for you.

Whenever Stallman irritates me, I remind myself of what freedom's really about: the particular license wording is only an implementation detail, and what is really required in principle is people who are prepared to be open and to share. The BSDs simply don't have this.

Re:woohoo

[Brian Feldman]

You had problems developing BSD kernel code and not Linux? That's amazing. What kind of driver or system call did you work on? I've never heard of anyone saying the Linux kernel APIs are more coherent. Ever.

FreeBSD & ZFS

[ThorGod]

What? A new FreeBSD release and no body talks about the ZFS features in the release? I don't memorize version numbers, but I know the ZFS system has updated significantly between 8.2 and 9.0. Deduplication is in there, now, for instance.

Granted, the new installer is one of the bigger changes. sysinstall...I'm happy to see you go!

Re:FreeBSD & ZFS

[Anonymous Coward]

Again: before ANYONE considers using dedup or compression on FreeBSD, please see this post of mine the last time this came up (a week ago), as it contains references to and admittance of the problems:

http://ask.slashdot.org/comments.pl?sid=2604202&cid=38589558

Additionally, manual "memory tuning" (specifically arc size maximum) on FreeBSD is still required, and it becomes more important to tune some other kernel variables if you run a mixed environment (ZFS + MySQL + shell machine, for example).

The Slashdot community can expect me to appear and comment every single time "ZFS on FreeBSD" is mentioned. Users need to be aware of the shortcomings so they can make an appropriate choices/decisions (staying with UFS, using another OS, etc.).

Re:FreeBSD & ZFS

[Maglos]

I never noticed any of this /w dedup and compress on. it chugged along and responded just fine. Manual "memory tuning" is not required, my 5tb file server w/ 48gb ram has no problem addressing memory whenever it wants /wo any tuning. ZFS is a beast, regardless of os or hardware but that is the point. You don't get bit loss protection, ram caching, compression or de-dup hash tables for free.

Re:Gnome version 2.32.1

[geek]

In the future Gnome3 will require SystemD which is Linux only.

ZFS v28

[Maglos]

ZFS v28 not a highlight? I just finished testing a 5tb Freebsd 9.0rc2 Supermicro server. ZFS v28 adds de-duplication and a removes rather nasty failure when an intent log device is removed. It also had built in support for the LSI HBA controller card I used, which made installation much easier. We'll save at least %40 with compression and de-dup but it does half write speeds with our xeon 5600(200MB/s down to 80MB/s) .

Re:ZFS v28

[bill_mcgonigle]

For many scenarios, ZFS v28 is the minimal 'usable' version number, which has previously limited FreeBSD's ZFS adoption. Now it's a real contender, and I congratulate the team.

Re: deduplication. Be sure you have enough RAM or you're going to be in for a heck of a surprise. 2GB of dedicated RAM per TB of disk usage is recommended as a rule of thumb. I found this out the hard way when it was new.

Re:But, what can I do with it?

[Galactic Dominator]

Yes, yes, yes, and yes.

http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/x11-wm.html
http://www.freebsd.org/doc/handbook/desktop-browsers.html
http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/virtualization-host.html

PC-BSD 9

[future assassin]

http://www.pcbsd.org/ will be announced today hopefully. Looking forward to giving it a spin and hopefully might change my mind about Linux Mint and become my main OS. Didn't have hardware luck with it in the early days.