FreeBSD 5.0 RC2 Released
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
you fell for the same headline as slashdot. 5.0 is going to be released then..not the second release candidate.
ummm… it clearly says that RC2 is already released and that 5.0 (AKA the final release) will be released on the 17th.
foofie, listen to Jim… and from now on, read articles as you would normally read anything else, don’t stare at it too long and try to find a discrepancy
Why does everyone have to babble.. these are not comments. they are wastes of space.. shut all yer faces and comment on the article.. not on other people’s comments.
My comment on the “article”:
excellent.. freebsd 5 is coming soon.. i’ve been waiting long enuff
Free BSD doesnt have its place anymore. I mean, who would want to install it anyway?? We’ve got linux for everything, openBSD for the rest. I think it’s time they get it and stop developping an obsolete OS.
…make sure that FreeBSD got support for J2SE 1.4.1 and I’ll be happy to throw myself over it.
…make sure that FreeBSD got support for J2SE 1.4.1 and I’ll be happy to throw myself over it.
I am thinking to switch on *bsd, i am sick of linux, hard to say why, updating every weeks for bug or security reason, or enough with the insane linux troll ?
I am looking too for Java compatibility. Do you know where to find the information ?
>Free BSD doesnt have its place anymore. I mean, who would >want to install it anyway?? We’ve got linux for >everything, openBSD for the rest. I think it’s time they >get it and stop developping an obsolete OS.
FreeBSD is far from obsolete, especially 5.0, FreeBSD can run anything linux can, has an awesome package system. 5.0’s kernel is just as good as 2.6, if not better in some areas, Linux and FreeBSD have both their advantages. FreeBSD is not obsolete.
troll.
I’ll take the bait!
“Free BSD doesnt have its place anymore. I mean, who would want to install it anyway??
Many people I know use FreeBSD on their servers.
“We’ve got linux for everything, openBSD for the rest. I think it’s time they get it and stop developping an obsolete OS.”
Obsolete in what way? I’m guessing you can’t answer that. Let’s look at just one of the ways FreeBSD has clear and incontrovertible technical superiority to Linux: the kqueue mechanism.
Linux has no event notification mechanism. This makes it very difficult to cleanly implement things like file managers, because they must constantly poll the filesystem looking for changes, instead of being notified with changes occur. This also makes utilities like tail(1) a kludge (when doing tail -f) as they must enter a sleep/stat loop.
kqueues also provide a stateful multiplexing mechanism. Linux has no such mechanism, and consequently any transaction-based network servers that use a select()/poll() loop will crumple under load in comparison to the same service implemented with kqueues. For awhile people provided a /dev/poll patch for Linux to emulate the Solaris facility, but this was definately lacking in performance and doesn’t provide as elegant and powerful an interface as kqueues. Here are some quick performance numbers for select/poll and /dev/poll versus kqueues: (see http://www.kegel.com/dkftpbench/Poller_bench.html)
Numbers for Linux (in microseconds): (650MHz PIII)
pipes 100 1000 10000
select 28 – –
poll 23 890 11333
/dev/poll 19 146 4264
The same benchmark, on a slightly slower (600MHz PIII) FreeBSD system:
pipes 100 1000 10000 30000
select 54 – – –
poll 50 552 11559 35178
kqueue 8 8 8 8
Here the stateful, O(1) nature of kqueues makes itself very clear. poll requires that for each call a large array of file descriptors and flags be copied into kernel space. This is a very slow process, and often occurs in a loop with a relatively large and unchanging set of file descriptors.
I think it is clear from this that select/poll are unfit for use in high performance transaction based servers. From this I think the clear choice for any service that requires high performance and handling of a great number of connections is FreeBSD.
I guess you answered it alot better than me
“FreeBSD can run anything linux can”
FreeBSD runs on a great deal less platforms than Linux. It can’t really run on PPC, and there are several platforms that Linux runs on that FreeBSD isn’t even looking at. Yu might be thinking of NetBSD.
Tried 4.7 but I didn’t like the X server configuration tool.
He said FreeBSD can run anything Linux can.
He didn’t say FreeBSD can run on anything Linux can!
“FreeBSD runs on a great deal less platforms than Linux. It can’t really run on PPC, and there are several platforms that Linux runs on that FreeBSD isn’t even looking at. Yu might be thinking of NetBSD.”
I think he is talking about the software that FreeBSD can run. BTW, FreeBSD now runs on IA32, IA64, SPARC64, Alpha and PC98 architectures – you have to admit that they’re working on it.
As far as version 5 goes: this is a very exciting release for all *BSD users, and other *nix users should at least take note of it. Read the release notes to find out more.
Although I’m tempted to, I’m not going to react on that crazy ‘FreeBSD is obsolete’ troll.
“FreeBSD can run anything linux can”
FreeBSD runs on a great deal less platforms than Linux. It can’t really run on PPC, and there are several platforms that Linux runs on that FreeBSD isn’t even looking at. Yu might be thinking of NetBSD.
He said “run anythnig linux can” not “run on anything linux can. He’s referring to software, which is pretty much true. I’d say, just about anything.
He didn’t mean run hardware platforms, he meant software, if he had meant platforms he would have phrased it as:
“FreeBSD can run on anything linux can”
He’s referring to that most software for linux can compile on FreeBSD, and for most linux binaries, there’s linux “emulation”.
I would really like to see some tools for configuring kernel before the compilation. Linux has make menuconfig.
Many Linux users would appreciate such thing in FreeBSD.
What really keeps Linux users from FreeBSD are all those little annoying differences like midc instead of mc, uncomfortable csh instead of bash (bash has far better word completion),and totally crazy startup scripts.
Some say that Slackware has its scripts based on BSD. I don’s see much resemblance between the two. Slackware’s ones are a lot simplier and easier to maintain.
I’ll bet that FreeBSD users feel right the opposit about Linux.
FreeBSD Obsolete? Bwahahahaha!
Thanks – I needed a good laugh.
No, FreeBSD is not obsolete. It has an advantage over Linux that will directly affect its development; organisation/standardisation.
The FreeBSD community isn’t fragmented with 50 possible sourceforge choices for running their USB NICs, or have anywhere near the same problems as a Linux user who wants to buy a new USB Scanner first, and worry about support later…
And with “Ports,” there is no RPM (and the associated dependency-hell) and whatnot. Bliss!
The play that I just had on NetBSD (on a hacked I-Opener) was also dramatically smoother than Midori Linux on the same box! (And it recognised my USB NIC, too!)
Oh, btw, I’m an ex-FreeBSD-er (who is thinking about returning) who is currently using Gentoo, Debian & RedHat in various places. ie, I’m not an FRZ!
For those who complain about FBSD not running on lots of stuff compaired to others. Who cares. x86 is the only platform that really matters, ppc to some exstant but not that much and 64 bit x86 in the coming future. Most of the platforms linux can run on are things you would be hard pressed to find a great deal of people using. and those platforms probably all ready have an OS far superior to linux. I much rather see something run on one or two platforms really well than having people and it trying to run on everything.
To the person who was trolling wishing FBSD should just end. SHUT UP. How about having linux just end. I could care less about it. If some act of god happened and the world had to choice which one lives it would be FBSD. 1 os that is unified and orginized. No 50 bizzilion distros with who knows what kernel. An OS that a compainy could take and use commerialy, (apple). One that is not a religion. One that has a nice name instead of some horribly stupid/gross sounding name. Honestly, if you say the word linux to someone, someone who doens’t know about it, doesn’t it sound horrible weird and not good sounding.
The people who keap saying other os’s should end and everyone work on linux have to be one of the most anoying groups of people ever spawned. Why don’t you stop working on linux and work on something else like fbsd. I know I’m wasting my breath/fingers saying that. If you didn’t get it the first 378 million times someone said this to you your not going to get it now. Because you rather be anoying and stupid.
Things like FBSD’s licence will keap it going forever because those who matter, hardware and big software compaines will support it. Without full support an OS will not live forever, or if it lives it will not have a super grand life. Once companies like IBM and SUN notice the market for selling linux to some people who caught it as a buzz word is tapped out, they will toss their linux groups and go fully back to AIX and Solaris. MS and Apple will not do anything for linux as they have for fbsd. In the long run linux will burn out it’s fuel while fbsd ever marches along though slowly and surely.
What really keeps Linux users from FreeBSD are all those little annoying differences like midc instead of mc,
mc? Like Midnight Commander? There’s a port of that if you really want it.
I’d say some of the bigger differences are getting used to a platform with sysctl() as opposed to setting kernel variables through /proc (Linux has sysctl now, but obviously sysctl support is well integrated into BSD)
“uncomfortable csh instead of bash (bash has far better word completion)”
pkg_add -r bash
“and totally crazy startup scripts.”
I’d call SysV init pretty crazy myself. BSD is one of the sanest inits around.
As for the driver issue, while Linux has a broader range of supported hardware, many of those drivers are incredibly krufty (i.e. any which aren’t in the kernel mainline)
Also, some in the kernel mainline (i.e. eepro100) are rather poorly written in comparison to the FreeBSD drivers.
I love freeBSD
Just wanted to hear you guys talking passionately about it, very selfish I admit
You did well anyway. FreeBSD deserves that much. Sorry for all the turmoil 
)) Nah, not sorry, I loved every second of it hee hee hee.
Promise I won’t do it again.
Peace
Dee,
You were wrong and trolling, admit it with some dignity instead of crawling back like this.
Pathetic really …
Kevin
Been using FreeBSD on pretty much all of my computers (servers as well as my primary workstation and notebook) for quite a while now and it still amazes me!
>>>>Free BSD doesnt have its place anymore.<snip>
I’d like to punch in the face!
Because you are wrong or right? No.
Because you were given just enough language skills to be trollish. But the Critical Thinking line must have all dried up when you arrived…because you offer no insight and no facts.
foad.
I think 17-Jan might be too soon but … did they changed the dates with this release candidate ? (can’t remember the last scheduled date for 5.0 asa I write this).
I would prefer to see the 5.0 final released without bugs and exploits.
FreeBSD is a modern OS with an old feeling. People just don’t get it, the only Linux that comes closer is Debian current.
Linux is good but FreeBSD always had a more professional approach to problems. (I only use it since 4.0)
Can’t wait for 5.0 and, personaly, the new SMPng to try it on a dual PIII.
dee.
one last thing.
if you really do love freebsd…and you also have some linux skills.
do something useful and describe to me what situations might benefit the usage of one versus the other.
maybe post something useful for a change.
I’ve been trying DP2 and more recently the 5.0 tree through cvsup and that’s for desktop usage. All I can say is that it good. Some debugging bloat in the code still but it works nice and well. Sound loads nice through modules. Brooktree capture card not yet (also not when compiling into kernel) but havemn’t looked deeper yet also.
I use rox desktop currently. No probs. Devfs is great. rcng is fine (better than before) and then there’s stuff like geom, sceduler, or the new ata that I haven’t even looked at but well they appear to work ok. fBSD is NOT dead! Not at all.
Although some of you (like Brandon Sharit) might like Linux because its “cool” and you can run it on your Zaurus hand held P.C, you are forgetting something. FreeBSD does the job of making cheap hardware (ia32) perform well, REALLY well. If you WANT unix on your Playstation 2 or your Toaster, that is what NetBSD is for.
As for the linux troll, I’ve got a life 2 get back 2…
Cheers
FreeBSD is a beautiful thing…to me almost everything is more logical than under Linux.
And it runs most Linux binaries…well enough that you can run things like Oracle and the JVMs almost flawlessly…
And ports…after using ports it’s hard to go back to the Linux package managers…updates are really easy too thanks to portupgrade…i think that the kernel is very easy to configure, simply edit a text file…and all you need to install it is a pair of floppy disks and a fast network connection
hardware support is almost as good as linux…like others said software support is almost essentially the same as Linux
what’s not to love?
-bytes256
i use freebsd to power some homemade VPN/NAT/Firewall machines
wouldnt have it any other way
cant wait for the 5.0 release
Linux has no event notification mechanism. This makes it very difficult to cleanly implement things like file managers, because they must constantly poll the filesystem looking for changes, instead of being notified with changes occur.
Actually file managers can be notified when file changes occur very easily. If you need portability FAM can be used, otherwise there are special kernel functions (I don’t recall what they’re called) which can register callbacks for FS events. Nautilus for instance updates instantly, if you watch a download you can see the file size growing as the file is written to – no polling needed.
Dunno about the rest. Linux is OK for me.
First of all, I mainly use Win2K but dual boot to RH8 and FreeBSD 4.7 to get familiar with them before MS drops support for Win2K (the last useable Windows in my opinion)
I have a HP374 disk controller and all my HDs are connected to it (not as RAID), and I had a difficult time installing RH8 since 2.4.19 doesnt support it. Luckily I found a Red Hat specific driver from HighPoint and installed it. I like it
I think 2.4.20 supports it but I am not sure how to use a new kernel while installing Linux (may be using modprobe, but I’ll look into that later, any suggestions?). Too bad though, I really wanted to try out Mandrake 9.0.
As for FreeBSD, it recognised the HPT374 with no hassle. It installed the latest packages directly from ftp. (I wish I had only downloaded the floppy image, since I hardly used the CD). I liked it a lot as well. I don’t understand why Linux and *BSD people are bashing each other, it’s free software. If you don’t like it; don’t use it. Personally I’d like to thank the developers of both platforms and wish them a happy new year. Keep’em coming
Now something that puzzles me a lot:
How come before I use the drivers for my HPT374 (or any other disk controller), no hard disk is initially detected. But when they are installed (on the hard disk) and machine reboots; the OS can happily read them from the hard disk, and therefore recognise the hard disks. ???
Although I mostly use Linux, I really like FreeBSD (I really like OpenBSD and NetBSD as well). I ran FreeBSD on my Dell laptop for quite a long time and never had a single problem with it.
There have been a couple of things, however, that cause me to prefer Linux over FreeBSD, and I was wondering if v5.0 was going to address any of these things, or if solutions were already available that I just missed.
1) No kernel configuration tool such as Liunx’s menuconfig or xconfig. I can read and tweak a text file like anybody else, but I find doing so more prone to errors. Is there a decent configuration tool on FreeBSD that compares to menuconfig?
2) Somewhat poor Java support. What version of Java will be included with 5.0?
3) I can never seem to get the wheel on my mouse to work under FreeBSD. I am very familiar with setting up X, but for some reason I can’t get the wheel mouse to work with either the generic mouse driver (whatever it is called) nor by accessing the mouse directly. Any clues on this one would be greatly appreciated.
I think that is about it. I know people complain about the text-based installer, but my favorite Linux distros (Debian and Gentoo) are both text based and I simply don’t think the merits of an OS should be judged by something you only do for 30 minutes or so.
Anyway, I love FreeBSD and look forward to trying this release.
About Java support on FreeBSD: The Sun JDK for FreeBSD is currently at version 1.3.1, but 1.4 is being actively developed and already seems to work if you have some skills (initial developers release can be found here: http://www.eyesbeyond.com/freebsddom/java/jdk14.html).
1) No kernel configuration tool such as Liunx’s menuconfig or xconfig. I can read and tweak a text file like anybody else, but I find doing so more prone to errors. Is there a decent configuration tool on FreeBSD that compares to menuconfig?
One of the reasons I prefer FreeBSD to Linux is kernel configuration. I prefer NetBSD to FreeBSD because of _kernel configuration_. Kernel configuration is a complex task, but IMHO *BSD makes it as simple as possible through the use of a single, human readable/palatable, *and* documented, text file. YMMV, but as for me, I’ve never compiled a Linux kernel I was ‘happy’ with, or one that didn’t result in errors I couldn’t understand.
Bascule, you’re not quite up to speed as to what’s going on in the linux-world.
Davide Libenzi’s epoll syscall has solved the issue you’re bringing up. And yes, it’s very scalable. (search for epoll on lkml for the discussion surrounding this).
Then there’s the issue of threading. Linux now has a 1:1 implementation which has shown itself to be quite a performer (http://www.kerneltrap.org/node.php?id=422) whereas FreeBSD has chosen to go with a M:N model, which in theory is better, but in practice is a hell of a lot more complex, both to implement and to debug. Is it even a part of 5.0? The fact that Sun never quite managed to debug their M:N implementation and has now, as of Solaris 9, switched to a 1:1 implementation (which btw is both faster than the old implementation and so far has had no bug reports filed against it) should tell you something. (http://wwws.sun.com/software/whitepapers/solaris9/multithread.pdf)
And how about interactive desktop-performance? Quite a bit of work has been done and continues to be done on the Linux-side here. It’s already pretty good in 2.4 (though you have to apply a patch), whereas 2.6 promises to be even better, with both kernel preemption and tunable low latency scheduling. Having fairly recently installed and tested FreeBSD 4.6.2, I can personally say that Linux is way ahead here, though perhaps some tuning can be done on the FreeBSD-side. Has anything been done on 5.0 regarding this?
How about SMP? From what I’ve read (on FreeBSD mailing lists), the SMP in 5.0 is supposed to be on par with (‘or maybe even better than’) Linux 2.4, but 2.6 is good step up from 2.4 in that regard.
How about neat stuff like UML (http://www.kerneltrap.org/node.php?id=409)?
How about journaling file systems? XFS is not only reliable, it’s also _very_ fast. Then of course there is ReiserFS, JFS, and even ext3.
How about Logical Volume Management?
I’m sorry, but FreeBSD really doesn’t have much, if anything, on Linux kernelwise anymore, while Linux does have the edge in a number of areas. And it’s likely that this gap will widen rather than narrow, as there is simply a lot more people working on the Linux-side. However, if you’re comfortable with the BSD userland and know your way around ports (which I think is a very neat system) and the tools it provides to keep the system up to date, then there really is no good reason to not prefer it to Linux.
Atleast not at this stage.
The Linux kernel configuration is better documented than the FreeBSD configuration. Try either ‘make menuconfig’ and press ‘?’ at the option you’re wondering about, or/and take a look in the Documentation/ directory. And _of course_ you can edit your .config file directly. No problem.
I’d just like to address a few issues here.
1. The FreeBSD kernel config is quite easy. Yes, editing a text file might sound hard to you but for the most part it is just a matter of commenting and uncommenting lines. If you must add something of course there is chance for human error — but there is chance of human error whenever you do anything. Would I like to see a better kernel config tool? Yes, but is it necessary? No. I don’t know about linux but as far as I can remember a good kernel config tool was almost required.
2. FreeBSD doesn’t need a journalled file system. As of 5.0 FreeBSD will be using UFS2 + Softupdates + background fsck. I’m not going to argue softupdates vs journalling because they both have their strong points. You are never going to see a journalled fs from linux in FreeBSD because of the GPL. Don’t like it? Don’t license your code under a crappy license.
3. I’m writing this on a FreeBSD machine right now. It’s a decent little Desktop machine (with Gnome). FreeBSD is a bitch to setup as a desktop OS — or maybe I should say “gnome” and “xfree86” are a bitch to setup on FreeBSD.
4. FreeBSD 5.0 is NOT intended to be a finished product. It won’t be stable/useful until 5.2 or 5.3.
What does FreeBSD have over linux?
a) A mature codebase (4.x)
b) A business friendly license
c) A complete OS (not as patchy as linux)
On a technical level both are likely pretty much equal. For years linux had been trailing behind FreeBSD (and might still be in a couple areas) but has caught up a lot in the last couple years (and maybe even surpassed FreeBSD in a couple areas). Don’t forget that a lot of this has been because large companies have started to adopt & develop for Linux (IBM, etc). Companies like Redhat have also done a lot for Linux.
Linux has a lot or corporate backing right now. Apart from this it also has a pretty big community of rebel coders.
The BSD’s (and many BSD related technologies) have more acedemic backing (IMO) — more code developped as a result of research grants, etc (atleast traditionally this was how it was).
I like both linux & bsd but I prefer bsd because of the license. I know that might sound silly but I just prefer the additional freedom that the bsdl gives me — I do infact agree with many of Richard Stallman’s ideals but not the way he tries to impliment them. I think there are many people (in both the bsd & linux camps) who will agree with me on this.
is always fun. Trolling is a great way to get new information
So, the issues I broached with Linux were:
* Lack of a kernel based event monitoring mechanism.
* Lack of a stateful multiplexing mechanism.
Both of which are addressed with FreeBSD (and now OpenBSD and NetBSD’s) kqueues, through a single system call at that (kevent)
So, the replies I got were:
For event monitor, FAM and some currently unknown system calls
So, I looked into FAM. SGI has developed a lot of good stuff which they’re adding into the Linux kernel, and this looks like yet another one of those.
Information on FAM may be found at: http://oss.sgi.com/projects/fam/faq.html
Unfortunately, it requires a kernel patch (either IMon or DNotify) so for now this still appears to be a facility the Linux kernel is lacking (unless someone can point me to these alleged system calls)
Meanwhile, kqueues are present on every FreeBSD system since 4.0
Now, to the issue of a stateful multiplexing mechanism. I took a look into epoll. Unless I’m mistaken the project home page is at http://epoll.hackerdojo.com/
Once again, this is a patch that’s not in the kernel mainline. Furthermore, it’s not extensible… it requires you to set an arbitrary limit which it presumably uses when allocating resources. It doesn’t appear to be actively maintained (their latest patch is for Linux 2.4.14)
I ask you honestly… if a feature like this isn’t available in the kernel mainline, what developer is going to consider using it?
I highly suggest you read the whitepaper on kqueues. It’s available at http://people.freebsd.org/~jlemon/papers/kqueue.pdf
Looking at the API for epoll, I can’t say it’s anything but nasty. It looks like an overcomplicated knock off of Solaris’s /dev/poll. I see there’s been some work to make epoll into a system call, which I’d much prefer, see:
http://lwn.net/Articles/13264/
That article links to what is perhaps the new home page for epoll, but the site (possibly Linux powered
isn’t available at the time of this posting.
kqueues are elegant and mature. They are present on virtually every FreeBSD system and have made their way onto OpenBSD and NetBSD systems. These provide event notification for not just file descriptors, but for virtually any event imaginable including signals and asynchronous I/O events (Linux has no way besides polling to determine the state of asynchronous I/O events as it lacks POSIX realtime signal queues)
I certainly can’t respond to everything that’s been posted. Obviously there are areas where a great deal of work has been done in Linux where virtually nothing has been done in FreeBSD, and so you can accept any non-replies as a taciturn concession of Linux’s superiority
How about SMP? From what I’ve read (on FreeBSD mailing lists), the SMP in 5.0 is supposed to be on par with (‘or maybe even better than’) Linux 2.4, but 2.6 is good step up from 2.4 in that regard.
From everything I’ve read, FreeBSD 5.0/SMPng came out much more finely grained than Linux (2.6 included) Everything being said about Linux versus FreeBSD SMP performance at this point is just puffery as there are no hard numbers. Unfortunately this isn’t as simple an issue as benchmarking filesystem performance where a universally recognized benchmarking tool (bonnie) exists.
I’d be most grateful if someone could either point me to some benchmarks or benchmarking utilities for SMP that run under both FreeBSD and Linux. Perhaps then we could move beyond the “some guy on some mailing list somewhere said Linux’s SMP is gonna be way better than Linux’s” or vice versa.
How about journaling file systems? XFS is not only reliable, it’s also _very_ fast. Then of course there is ReiserFS, JFS, and even ext3.
This issue was addressed by someone else in an earlier post too. Nevertheless, I could just as easily point out how Linux has no filesystems with metadata caching. With filesystem consistency checking in the kernel, FreeBSD can reap the benefits of a journaled filesystem while also reaping the benefits of metadata caching (soft updates). With these two features, FreeBSD has one of the best all-around filesystems available.
I’m sorry, but FreeBSD really doesn’t have much, if anything, on Linux kernelwise anymore, while Linux does have the edge in a number of areas.
I’ll briefly summarize what I’ve mentioned here as far as the advantages of FreeBSD’s kernel over Linux:
* FreeBSD has an mature and elegant (and to a certain extent portable) event monitoring mechanism in the mainline kernel, which can monitor all sorts of events (filesystem, network, async I/O, signals) whereas Linux, in conjunction with a kernel patch, can monitor a limited set of filesystem events through a userspace daemon and library.
* FreeBSD has a mature and elegant (and to a certain extent portable) stateful multiplexing mechanism which can multiplex network socket, filesystem, async I/O, and signal events. With a patch there exists a proprietary stateful multiplexing mechanism for Linux.
Really, there’s no comparing these features found in rag tag kernel patches to the ones present in the FreeBSD mainline. Perhaps in a few years when/if these features are present on the majority of installed Linux systems and developers have begun to take advantage of them can a comparison be made.
I cannot understand this kernel configuration discussion. I can only say that the people raising these issues seem to be coming from the Linux world where it’s truly necessary to rebuild your kernel.
My question to all of you who are asking about the kernel configuration is: why are you rebuilding your kernel?
I don’t use the generic kernel, but I have my own reasons for doing so (I’m playing around with FreeBSD’s asynchronous I/O features which aren’t enabled per default due to security concerns)
Unless you have some issue such as this, it shouldn’t be necessary to use anything but the generic kernel. Most likely whatever it is your tweaking can easily be solved by adding a few lines to an appropriate text file (sysctl or loader config) instead of recompiling your kernel. Doing it the proper way means when you upgrade your system you won’t once again have to recompile your kernel when you upgrade.
The need NOT to recompile your kernel should be counted among one of the many advantages of FreeBSD over Linux
1. I never said anything about the difficulty of configuring and compiling a kernel under FreeBSD or Linux. I only brought up the issue of documentation of the available options. The Linux kernel options are well documented, the FreeBSD options less so.
2. I’m sure you, and in fact most people, don’t particularly need a journaling file system, but I’m not sure people deploying mission critical solutions are going to be enticed about soft updates and stop-gap measures like background fsck. And how about speed, vis-a-vis for instance XFS? Is there any data about that?
3. I’m sure it’s quite decent, but I ask again, has any work been done regarding interactive usage in 5.0?
Regarding your ‘what does FreeBSD have over Linux’ points.
A) In terms of age of the original codebase, FreeBSD is certainly more mature, but how about actual real-world deployment and eyeballs? And in the real world, how does this manifest itself? I’d like factual objective data here which can clearly be traced back to this claim, not fuzzy ‘this is how I feel about it’-stuff.
B) You know, it’s really quite comical how proponents of the BSD license keeps harping on about how business friendly it is, while out in the real world, Linux is getting all the contributions from companies like IBM, SGI, Veritas and so forth, almost all of it under a GPL license (some of it cross-licensed). I know that for instance FreeBSD-people inside of IBM has made a lot of effort trying to get FS released under a BSD license in addition to the GPL, but to no avail. The GPL allows IBM and other businesses to make open source contributions without letting anyone who doesn’t want to play ball grab it and run with it. It’s obvious that the GPL, not the BSD license, is the one that best looks after their interests.
That said, there are of course plenty of places where the BSD license makes perfect sense, for instance in reference implementations of technology that should be widely adopted.
C) Debian is a complete OS, Red Hat is a complete OS, Gentoo is a complete OS, and so on. No, they generally don’t maintain their own source tree for ‘base’ utilities like FreeBSD does, nor their own kerneltree (though they take care to integrate both the kernel and the tools into a coherent whole). And why should they? It’s a waste of effort.
And having several different organizations pushing their own distribution is a _great_ advantage. You can more easily pick a distribution which suits your particular taste (the three I’ve mentioned above plus Slackware all represents distinct flavors), plus you get a lot more people contributing and you get a lot more people pushing it. So you have for instance SuSE which has a strong presence in Germany, you have Conectiva who are big in Brazil, Red Hat who are big everywhere, and so on. FreeBSD on oth doesn’t have all these organizations trying to push it as if their livelihood depends on it.
Do you really think Linux would have been where it is today had Linus decided to adopt a BSD-model and built a complete OS from the beginning instead of just providing the kernel and letting third parties do the assembling? I doubt it.
Part of the problem with the BSD-community is their inability to make an _honest_ evalution of what their strengths and weaknesses are vis-a-vis the competition. Instead you get a stream of outright lies, half-truths, or simply grossly outdated information regarding the competition, while at the same time strongly downplaying their weaknesses (‘we don’t _need_ a journaling file system’). Your claim that Linux has been trailing FreeBSD for years, while only _maybe_ recently surpassed FreeBSD in a couple of areas is a perfect example of this.
You can keep repeating the ‘more mature codebase’- and ‘BSD is a business-friendly license’-mantras, but at the end of the day, where does that really get you?
The first issue, I don’t know.
The epoll syscall was included in Linux 2.5.45, iow, fairly recently.
Regarding the maturity and elegance if these features, I’m not qualified to comment.
I believe most Linux users do not compile their own kernel, as just about every distribution provides kernels with everything necessary as modules (usually providing more features and support for more devices than what is found in the mainline kernel). Personally, I’ve compiled my own custom kernel (as Debian relies on Linus’ kernel out of the box) to include two features not found in the mainline 2.4-kernel, the XFS file system and kernel preemption (backported from 2.5). It’s done in no time. Besides, it gives me a warm fuzzy feeling to do so
Do you really think Linux would have been where it is today had Linus decided to adopt a BSD-model and built a complete OS from the beginning instead of just providing the kernel and letting third parties do the assembling? I doubt it.
I dunno, I’d rather run an OS where the same people write the kernel and C library as opposed to an OS where the people who write the kernel argue with the people who write the C library
Uh. The glibc maintainer, Ulrich Drepper, is a Red Hat employee, and very much in-the-loop and responsive to what goes on in kernel-land.
Samb: You need to take some chill pills.
As for your claims I could spend all day arguing with you and achieve nothing. FFS + Softupdates + background fsck sound wordy but are a very elegant solution. It doesn’t sound like you’ve even read about it — so heres a link:
http://www.mckusick.com/softdep/index.html
Although not a perfect indication, I think netcraft’s top uptimes demonstrate the maturity of the *BSD codebase.
http://uptime.netcraft.com/up/today/top.avg.html
GPL vs BSDL: No comments, forget I even mentioned it.
You sound like you have already made up your mind, and I apologize if you are unsatisfied with my reply, but I don’t have time for essays. My time can be better allocated to more productive causes.
Cheers.
I’m sorry, but I just can’t agree with calling background fsck an elegant solution when there are in fact better solutions out there, regardless of whatever other merits the system might have. An adequate solution, perhaps. Judging by the FreeBSD mailing lists, people seem to be having mixed experiences with it though. I guess they’ll sort it out once it gets wide deployment in the 5.x releases.
As for the Netcraft link, feel free to take a look at the FAQ as well. Heck, I’ll save you the trouble and quote from it: “Additionally HP-UX, Linux, Solaris and recent releases of FreeBSD cycle back to zero after 497 days, exactly as if the machine had been rebooted at that precise point. Thus it is not possible to see a HP-UX, Linux or Solaris system with an uptime measurement above 497 days.”
If I came off a bit agitated on the license issue, it’s because I’m sick and tired of seeing a claim which is so strongly at odds with what actually occurs in the world of software, repeated again and again. One might have been excused for believing it back in ’97 or so, but now, as we’re moving into 2k3 and commercial open source support and contributions are so squarely on the GPL side, it’s just silly.
Have I made up my mind? I’d like to think I have an informed opinion, as opposed an opinion based on ignorance and/or a wish about how the world _should_ look.
Whether or not FreeBSD is lagging featurewise with Linux in some area or not, I have good reasons for prefering FreeBSD.
Simplicity of configuration, man 7 hier, license, documentation, professionalism, and the kick-arse features outlined by others (KSE, SMPng(a WIP, unlikely completed for 5.0), MAC, Softupdates, et al.)
Yes, FreeBSD is easier to configure than most Linuxen with basic knowledge of a text editor (the UNIX way of doing things) and http://www.freebsd.org/docs.html . I don’t get the (linux) posters who complain about not having a pointy-clicky interface for configuring an OS, the text editor/commandline way is a much more precise way of configuring a system.
BTW, my OS history is DOS->OS/2->Win9x->Linux->FreeBSD and I still use Win2k, so I use pointy-clicky too. It just doesn’t give me confidence that the system is doing what I asked of it.
As I’ve said, for a lot of people there are perfectly legitimate reasons for prefering FreeBSD to anything else, though mostly subjective ones.
Some of the things you mentioned can’t really be argued objectively, and I suspect are only thrown in ‘for good measure’.
I see the documentation-argument quite a bit though, and it might very well be valid compared to many, perhaps most, Linux distributions. I really wouldn’t know. Debian however, is thoroughly documented (http://www.debian.org/doc/), and everything must adhere to a rigid system-wide policy. If something doesn’t ‘fit’, it’s a bug and should be reported as such.
Two of the ‘kick-arse’ features you bring up (KSE and SMPng) aren’t even done yet, and even when they eventually get there, neither are promising that they will eclipse what Linux can deliver. Why even bring them up? Linux is ahead _today_ in those areas. Softupdates we’ve been over already.
While I’m not very familiar (not at all, actually) with the graphical tools provided by the different Linux distributions out there, I would be very surprised if they can’t be completely bypassed by those who don’t care for them. Care to mention any such tool which get in the way of text-file jockeys? I certainly know of no mandatory graphical tool in Debian.
Look at things this way. There are over 100 distro’s of linux. You do a minimal install of RH, suse, or mandrake with X and lets say afterstep. Lets do the same thing on Freebsd 4.7 (or 5.0. result…. Linux with a billion processes running.. prolly around 40-50. It runs semi-slow. Do the same thing with FreeBSD.. and run it with power hungry KDE or Gnome2.. guess what? You’ll have only 25-30 processes, most of the RAM free and a blazingly fast OS. (THe assumption is a AMD xp 2000, 512meg ram, geforce4). Also.. many of the ways to configure freebsd are the same as the old days in linux before these full of bugs GUI’s come into play that take longer to load and run than command line.
Sigh. Some people obviously put no thought whatsoever behind whatever ‘arguments’ they throw out.
Yes, technically speaking, there might be over 100 Linux distributions out there (nevermind the fact that most of them are put together for specific purposes, and not meant for the general desktop user). So what? What exactly are you trying to argue here? How does {set of all Linux distributions – X} affect people who are running distribution X? My distribution of choice is a self-contained OS, just as FreeBSD. If anything, it’s FreeBSD that struggles with outside dependencies, with a fair amount of people depending on stuff only usable via the Linux emulation layer.
As for the minimal install of the distributions you mention, I can’t comment on the truthfulness of it, having never installed them, but even if what you say is true (which I doubt), how does that affect those of us who run distributions which doesn’t do this? Why do you insist on comparing FreeBSD to distributions with ‘worst of breed’-behaviour in this area?
Your last ‘argument’, I’ve already gone over.
Samb, if you spend as much time arguing how much better linux is than BSD .. and actually spend as much time hole heartedly installing and running BSD– maybe you would see the point that everybody has been trying to make here . I’ve installed at least 35 different distro’s of linux and I keep coming back to freebsd. Aside from peanut linux, give me a distro that you can install completely and ready to go in under 10 mins, one that finishes the boot process in 10s, one that after installation runs only 19 processes while in X using windowmaker. I know you are hardcore linux man, but when it comes to fullfilling my needs (Not a company or corp), BSD goes above and beyond. Personally I wish Sun spent more time on their x86 platform, unfortunately they lack support for the hardware unless you are running a 2 year old system or older.
Just a comparison:
Linux is like a nice christmas tree with beautiful little shiny things on it. Good to see that all these little things makes its fans blindly love it.
BSD is an old oak. Big, robust, old (mature), and you know you are safe in its shadow… (And can see that all the newly grown up christmas trees in the fields fall down)
meaningless to discuss about OSs, if you just cannot acknowledge that some OSs do things better than other.
I appreciative that some things does linux better than FreeBSD, but:
What I love in FreeBSD (after years of experiences with various linux distros) is its robustness, it performs well under highest load (OK, I know, that the latest greatest linux 312.892 kernel makes that better, but I stability is more important in the commercial world than speed). Unbelievable easy to update ports and the whole OS. (No suck with “glibc X.X isn’t compatible with older X.X version binutils” or I don’t know, but it was a nightmare for me to update the system.)
And people at linux world: please try not to begin your programs with “#include <linux.h>”. That doesn’t help the opensource community at all; that’s a kind of windows-programmer thinking. Thanks.
mizu
I can’t help but notice that your pro-FreeBSD arguments centers around speed of installation, how fast it boots, and the number of processes the system starts up after boot (all compared to a subset of Linux distributions), but no actual arguments regarding use, which strongly suggests to me that you actually spend most of your time in a version of Microsoft Windows.
Keep chasing that holy grail, dude.
I’m not particularly interested in arguing the superiority of neither this nor that, I’m merely trying to counter the barrage of misinformation regarding my system of choice. If you love FreeBSD, great!
The only use for windows is to run games that can’t be run using wine. You’re funny samb trying to come across so smart to have guessed by my questioning that I use windows.? It doesnt take a genious to see I used an aol junk email address. Don’t try so hard to pretend to read between the lines =) Maybe you should try fortune telling as a profession too?? j/k j/k
If you are curious i triple boot redhat 8, bsd, and xp. You have valid arguments with linux. It offers more, but the arguments lie onto where the offers are geared to. Personally I liked Solaris a couple years back and used it quite often. Unfortunately it does not keep up with x86 hardware very well, so I had to ditch it. It just all depends how you look at it though. Its like a car.. what matters more to you? The looks? whats under the hood? How fast? Easy to Use? Ecnomoical? The answer and reasoning to each of those questions are going to give you different OS’s. May I ask you Samb, what is your main purpose for using linux over other OS’s? Personal? Job? I’m curious.
Hi all,
I really wonder about the patent infringement lawsuit threat to Linux. I’ve read some FUD about it…and wonder if it will come to pass or not — and what the effects will be on Linux and FreeBSD. Especially for those USING the OS in business settings.
I’ve only read stuff talking about Linux being vulnerable to patent infringement suits, but wonder what other operating systems (i.e. *BSD) might be affected too.
If Microsoft (or it’s allies) buys a patent portfolio (or has one already) they can use to start filing patent lawsuits against the Linux distro makers or *business users* of Linux, the uptake of Linux by business might evaporate quite quickly. Companies would likely overreact emotionally and swear off OSS operating systems for a good long while.
I’m NOT in favor of this — just worry about it happening.
I use both Linux and FreeBSD.
What think ye on the threat of patent infringement suits?
Thanks,
Louis