The FreeBSD project announces the availability of the second preview for their upcoming 5.0 release. This major release will be based on the GCC 3.2 branch, and bring us an extensible Mandatory Access Control framework (named TrustedBSD), the new UFS2 format with support for larger filesystems, support for Firewire devices, on-disk encryption, SMPng (next generation SMP support), KSE for multiple kernel-level threads, and much more. The Early Adopter's Guide might be a good read for those interested in getting started with the new 5.0 codebase.
The release announcement follows.
From: bmah@FreeBSD.org (Bruce A. Mah)
To: freebsd-announce@FreeBSD.org
Subject: FreeBSD 5.0 Developer Preview #2 Now Available
Date: Monday, 18 Nov 2002 16:39:13 -0800
The FreeBSD Project is proud to announce the availability of the second Developer Preview snapshot of FreeBSD 5.0 (5.0-DP2). This snapshot, intended for widespread testing purposes, is the latest milestone towards the eventual release of FreeBSD 5.0-RELEASE, currently scheduled for mid-December 2002.
Since the release of 5.0-DP1, a number of new features have been added on the CURRENT development branch. A few of these include:
* The GEOM disk geometry module and the GBDE on-disk encryption system.
* A compiler toolchain based on GCC 3.2.
* A new extensible Mandatory Access Control framework, the TrustedBSD MAC Framework.
* The new UFS2 on-disk format, with support for larger filesystems and extended file attributes.
* Support for Firewire devices.
* Experimental support for the RAIDframe software RAID disk driver.
Much more information can be found in the release documentation, as described below.
WARNING
This snapshot is not a supported release, and has not undergone any of the usual quality assurance checking that is a part of normal FreeBSD releases. It may include serious software bugs. Do not install this software on a machine where important data may be put at risk.
The purpose of this snapshot is to get wider exposure to FreeBSD 5.0's new features, in advance of the release. As such, we expect that testers of 5.0-DP2 will follow the technical discussions on the freebsd-current@ mailing list.
AVAILABILITY
FreeBSD 5.0-DP2 is being released for the alpha, ia64, i386, pc98, and sparc64 architectures. As of this writing, 5.0-DP2/i386 is available, with other architectures to follow soon.
The checksums for the two i386 ISO images are as follows:
MD5 (5.0-DP2-disc1.iso) = 997ef9ed5aa3e0721678f5482d2fc664
MD5 (5.0-DP2-disc2.iso) = 425718dbf1b771c8317556b7a13842c6
Before trying the central FTP site, we strongly recommend that you check FTP mirror sites in your country or region, such as:
ftp://ftp..FreeBSD.org/pub/FreeBSD/
Any additional mirror sites will be named ftp2, ftp3, and so forth. Appendix A of the FreeBSD Handbook has additional information about FreeBSD mirror sites; it is available on-line at:
http://www.FreeBSD.org/doc/en_US.ISO8859-1/books/handbook/mirrors-ftp.html
Some of the mirror sites known to carry 5.0-DP2 (in alphabetical order by country code) are:
* ftp://ftp.FreeBSD.org/pub/FreeBSD/
* ftp://ftp2.FreeBSD.org/pub/FreeBSD/
* ftp://ftp2.at.FreeBSD.org/pub/FreeBSD/
* ftp://ftp.au.FreeBSD.org/pub/FreeBSD/
* ftp://ftp4.de.FreeBSD.org/pub/FreeBSD/
* ftp://ftp.dk.FreeBSD.org/pub/FreeBSD/
Release documentation is available in the distributions, as well as on the FreeBSD Web site:
* Release notes: http://www.FreeBSD.org/releases/5.0R/DP2/relnotes.html
* Hardware notes: http://www.FreeBSD.org/releases/5.0R/DP2/hardware.html
* Errata: http://www.FreeBSD.org/releases/5.0R/DP2/errata.html
* Early Adopter's Guide: http://www.FreeBSD.org/releases/5.0R/DP2/early-adopter.html
LATE-BREAKING NEWS
Certain parts of the KDE and GNOME desktop environments did not make it to the package sets in the i386 ISO images.
The sparc64 and ia64 snapshots may or may not have package sets associated with them.
ACKNOWLEDGEMENTS
The FreeBSD Project wishes to thank the companies, developers, and users who continue to make FreeBSD releases possible.
Many companies donated equipment, network access, or person-hours to finance our ongoing release engineering activities, including The FreeBSD Mall, Compaq, Yahoo!, Sentex Communications, NTT/Verio, and Packet Design. We greatly appreciate their contributions.
The release engineering team for 5.0-DP2 includes:
Murray Stokely Release Engineering Team Lead
John Baldwin Release Engineering, alpha and sparc64 Builds
Scott Long Developer Communications
Bruce A. Mah Release Documentation, i386 Build
Robert Watson Release Engineering, Security
Marcel Moolenaar ia64 Build
TAKAHASHI Yoshihiro pc98 Build
Kris Kennaway i386 and alpha Package Builds
FreeBSD 5.0-DP2 is, as with every release, a cooperative effort of all of the FreeBSD committers , as well as FreeBSD users all over the world who have submitted new features, bug fixes, and suggestions. Please join us in thanking everyone for their hard work in polishing and improving the state of the FreeBSD -CURRENT development branch.
We'd like to emphasize once again that this snapshot represents a work-in-progress. Please help us by testing it now, so that we can make FreeBSD 5.0-RELEASE as stable and useful as possible.
Thanks!
Bruce A. Mah
(For the FreeBSD Release Engineering Team)
You've gotta love it!
FreeBSD 5.0 sounds great! I tried FreeBSD a while back and really liked it, except it didn't play nice with my machine....it turns out that nothing plays nice with this machine any more, so I don't think it was a BSD problem. The only thing that keeps be from moving my server to BSD is that I'm more comfortable with SysV init than I am with BSD init.
Way to go to the FreeBSD folks, it looks like you've done your usual wonderful job.
Check out rc_ng ...
Part of FreeBSD 5 is the adoption of the NetBSD RC system. Now, we have a system like SysV init in the form of /etc/rc.d/. Now things like "/etc/rc.d/sshd restart" and the like work very similiar to what you expect from SysV. Of course the system is far more sane in it's operation (dependancies).
Yay!
I like the sounds of that. I may just have to give FreeBSD 5 another shot once it's released.
Speed
/* Disclaimer: I like *BSD. This is not a troll. */
With all the speedups and performance boosts that we're getting with Linux 2.6/3.0, is *BSD going to be competitive in future? FreeBSD 5.0 is their first release that really tries to perform well on [large] SMP boxes, and according to the developers (last I asked.. quite a while back) it will be at the Linux 2.4 level or so. And what about [desktop] latency? Do they even think this is important? Maybe they're only pushing for server use? I used to be a FreeBSD user, but I moved to Linux when I saw (and felt) how fast 2.6/3.0 would be on the desktop.
hmm
Really? As I understood it, FreeBSD's smp level was supposed to surpass 2.4's!
In fact, it might even be around Linux 2.5/2.6 level.
Well, maybe I am wrong. is there anyone around here who knows for sure? I'd REALLY like to know.
I love FreeBSD and Linux so I hope the new releases of both, kick ass. Nothing like a little competition to motivate some development I think. :)
hmm
i'm going to have to agree, i also love them both. Competative motivation definitly helps too, as long as we dont waste too much time flaming eachother ^_^ (which is such a time waster). Too bad most BSD and linux users didn't see things that way, instead of the usual "well blah blah's better then your blah blah". It's about time we get those teamwork game faces on, and do the 'world domination' thing together ;)
absolutely
i'm going to have to agree, i also love them both. Competative motivation definitly helps too, as long as we dont waste too much time flaming eachother ^_^ (which is such a time waster). Too bad most BSD and linux users didn't see things that way, instead of the usual "well blah blah's better then your blah blah". It's about time we get those teamwork game faces on, and do the 'world domination' thing together ;)
I absolutely agree with you, I am so sick of the flame wars, it should be friendly competition! If FreeBSD is the best, we all win, if Linux is the best, we all win. We still have freedom, do we not? And really, they're both unix(-like), so it's not like there's all that much left to flame over considering all the recent improvements. I think they will overall be equal (each better in different areas) and if not, it's still good for everyone. Variety is the spice of life anyhow.
As for flame wars, how many r
As for flame wars, how many really develop between the core developers of Linux and BSD? I seem to recall one of the BSD VM hackers reviewing the Linux VM, comparing it point for point to the more mature (at the time) BSD VM. The luser Linux zealots may have taken it defensively, but the Linux VM hackers ate it right up and learned from it. For the developers, it was a positive experience on both sides of the fence.
There's been similar sharing of brain cycles comparing soft updates to journaling to phase trees to whatever in the filesystem space.
Sure, each side will thump its chest from time to time, no big deal. I don't see either side being openly antagonistic in any real deep way. If anything, it's the zealtrous users that are the problem. Fortunately, the developers largely ignore them. :-)
No
FreeBSD developers admit SMP scalability is a couple of years behind Linux. That said, when FreeBSD 5 does eventually release it could be ahead of 2.4. It is ahead of Linux in a lot of other areas, sure.
The bulk of the work is really identifying and fixing SMP hot spots. In Linux 2.5 development for example, a few developers spent months and months testing and profiling on a multi million dollar 32 processor POWER4 machine. I just don't think FreeBSD has those resources or focus yet.
For example, a 24 way 1.3GHz POWER4, running plain 2.5.40 can compile a kernel in 6 seconds flat. I don't think even FreeBSD 5 would be up to that.
evidence?
not that I doubt what you claim but I'd love to read on this further. Do you have any links on the topic?
Thnx!
~Christopher
Sure
http://www.samba.org/~anton/linux/2.5.40/configuration
and
http://www.samba.org/~anton/linux/2.5.40/kernel_compile/
The tests are I think mostly to aid kernel development and
not really for public consumption so don't bug anyone about
them please!
Re: Sure
Those boxes are very rare. The only really interesting test is between common SMP setups dualbooting FreeBSD 5/Linux 2.5.. I don't think any benchmarks have been done yet however.
Not the only one
No I think that was a really interesting test actually, yes it doesn't say anything about FreeBSD's performance, but if FreeBSD could run on that or say ran on a similar Alpha then it couldn't hope to get those sorts of numbers.
Nick
and FreeBSD developers
I'm sorry, no links for that but I have read it here and there. I don't think it is a disputed point.
5.0 scalability
I'm a freebsd developer and 5.0 will be an early adopter release. It won't be stable until 5.2 or so. So in this sense, it's like 2.5.
5.0 has a lot of capabilities towards SMP scalability but too much is not done yet for there to be any performance advantage, even over 4.x. For instance, no drivers are Giant-free yet (GEOM may be close). 5.0 provides the infrastructure to scale but that has to be utilized by all the drivers for it to make a difference. And then there is still testing and tuning to be done. Linux IS ahead but there are plenty of other places BSD is ahead.
In short, don't install 5.0 looking for performance. Instead, install it for the new features and to provide valuable testing input since as the other person said, we are dependent on user testing for bug fixes. Performance will come in time.
slash
Even if 2.6/3.0 will have much better SMP than FreeBSD, so what? There are plenty of areas where FreeBSD will blow Linux away and vice versa. Also, for 90% of systems, FreeBSD at the 4.x level is more than good enough. For about 98% of the systems, FreeBSD 5 will be more than good enough. For the remaining systems, people can go with Linux at the lower end to Mainframes on the high end.
Why don't you try to use each OS for their strengths. Consider using Linux 2.6/3.0 for Oracle 10i on a 64 cpu system. Consider using FreeBSD 5.0 on a Router/Firewall/Traffic shaping server or even low end databases with postgres SQL on eight cpu's. I hate it when people try to push one OS to do everything. It reminds me so much of Microsoft mentality of thought.
Personally, I use FreeBSD for everything FreeBSD does better than Linux and everything it does about the same as Linux. FreeBSD development process is much more mature, changes to how things are done is much less frequent, and it has a much more integrated feeling to it. Again, this is my own personal thought. However, I am not afraid to use Linux, or *GASP* even Windows, when FreeBSD is not able to do it as well, such as SMP DB systems or for Java applications.
Re: slash
SMP was just an example. Linux 2.6/3.0 is going to be faster than 2.4 in nearly every area (small->large->huge servers, desktop/workstation use) by the time they finish tuning it (they have only *started* tuning recently, since the feature freeze).
[`faster' can be read as smoother/more responsive in the desktop/workstation case]
Actually, that is not true.
Actually, that is not true. Linux 2.4 is already about as good as your going to get on the low end system. Linux 2.6/3.0 focus is on high end machine and it is going to respond extremely well for the high end SMP systems. It might even be comparable to Solaris. But in terms of any performance improvements for most intel systems (1-4 cpus), it will not be noticeable. This is what Andrew Morton, a kernel hacker, had to say about it: "For the uniprocessors and small servers, there will be significant gains in some corner cases. And some losses. [...] Generally, 2.6 should be nicer to use on the desktop. But not appreciably faster." http://www.kerneltrap.org/node.php?id=10
slight disagreement
I think due to the i/o improvements alone even low-end machines will feel more responsive - as in fewer to now mouse stutters. Sure, for the most part it will be the same but do not discount the importance of i/o.
Integrated compared to what?
Integrated compared to what? If you want a tightly integrated and consistent Linux-system, there is Debian GNU/Linux. I recently did a FreeBSD install (4.6.2 I think), and while the install was fairly straightforward (though the FreeBSD partitioning scheme puzzled me a little at first), and the ports-system seems to be everything it's cracked up to be, I didn't really see any tangible benefits compared to Debian. In fact, the desktop, which is what really matters to me, definately felt like a regression compared to Linux 2.4 with preempt- and low latency-patches applied.
I agree that generally Linux
I agree that generally Linux is light years ahead in the GUI department, but the integration I was talking about was on the command prompt. Linux has a tendency of adding more commands/extensions to solve problems that do not exist. On the other hand, FreeBSD uses traditional unix commands and makes you pipe commands (thus giving it a very integrated feeling.) They don't throw everything and the kitchen sink at you. If you play around with RedHat, you would know what I mean (even on the GUI end, RedHat's GUI feels cluttered and unintegrated.) Also, if you use RedHat, you would also definately see the benefits of BSD's ports and packaging system.
As with Debian's packaging system, I have heard very good things about it. I can't comment on it since I haven't tried it, but supposedly it is every bit as good as BSD. Hence, that is probably the reason why you are not impressed with BSD.
The reason why I never bothered myself with Debian is that, until recently, they were really behind in terms of the kernel they use. I don't believe that everyone should jump at the first sight of latest and greatest, but Debian really dragged their feet in this matter (and the 2.4 kernel was really a big improvement over the 2.2 kernel in almost every way). I can't justify myself to use Debian when they seem to consistently drag their feets on new technology even when the new technology is much more stable/secure than the older technology.
If you wanted to run Debian w
If you wanted to run Debian with 2.4.x kernels, there were many options for that. For instance there is the XFS boot cd which contains the 2.4 kernel patched for XFS. It is the exact Debian install cd except for the kernel.
There are other setups like this as well. You just have to have a look around and find the proper cd to install with.
FreeBSD requires tuning...
to be a decent desktop in my experience. By default it has some very conservative settings, knowing what to change and what to turn on helps tremendously in my experience. However even after my tuning I still experienced mouse-stutters more often under stress then with Linux - although disk i/o was noticably better but networking performance (on a cable modem) was a wash - both do great.
so what?
Well, seriously, there aren't a lot of things that distinguish FreeBSD and Linux. On the desktop for example, the advanced features and kernel behaviour of either is pretty much irrelevant (and don't tell me how much smoother FreeBSD is when running 7 make worlds on your 386 with 1 MB ram)...
Linux (2.4) has caught up a great deal with respect to stability and networking features and VM behaviour.
One area that is very visible and really easy to measure is performance. A lot of people would put the importance of a couple of % performance above other factors - its hard not to as it is the easiest thing to quantify. The 8xCPU server you mention - that is really a very high end SMP machine for FreeBSD - 5.0 will have trouble scaling that high mark my words.
I'm not trying to rubbish FreeBSD here, its a great system, its a fact of life that it doesn't have good SMP scaling yet. Unfortunately it is a glaring point. Most comparisons really focus on performance. So that.
Well I can't really comment
Well I can't really comment on how well FreeBSD 5.0 will scale on an 8 x CPU system since I haven't tried it nor have I seen benchmarks on this. But judging on rumors, which say that FreeBSD 5.0 will be somewhere between Linux 2.4 and 2.6/3.0, I find it difficult to believe that it won't scale to an 8 x CPU system. Linux 2.4 practical scaling limit was 8 CPU's. Hence, I find it difficult to believe that FreeBSD 5.0 will be better than Linux 2.4 in SMP yet not be able to scale as high as 2.4.
We'll see won't we
Obviously it _either_ has beetter scalability than 2.4 _or_ it won't be able to scale as high as 2.4. One of the two statements will be incorrect. Scalability is what you mean when you say better in SMP, right? Anyway we'll see.