The FreeBSD status reports are back again with the 2003 year-end edition. Many new projects are starting up and gaining momentum, including XFS, MIPS, PowerPC, and networking locking and mutlithreading. The end of 2003 also saw the release of FreeBSD 4.9, the first stable release to have greater than 4GB support for the ia32 platform. Work on FreeBSD 5.2 also finished up and was released early in January of 2004 (our review), while we learned FreeBSD 5.2.1-RELEASE is expected sometime next week with some critical fixes in it.
FreeBSD: October-December 2003 Status Report
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
The real question about 5.2.1-RELEASE is whether it will make the ULE scheduler the default, and will KSE threading be enabled per default.
Until ULE and KSE get proper deployment and testing, 5.x will not begin to approach production readyness. Users who are accustomed to FreeBSD 5.x’s apparent stability now seem unaware that they’re still using the antequated 4BSD scheduler and userspace threading.
FreeBSD 5.x will definitely take a stability hit once ULE scheduling and KSE threads are enabled per default. But when will this be? Will we see it in 5.2.1-RELEASE, or do we have to wait until July for 5.3-RELEASE? If so, where does that put 5-STABLE? January 2005?
As for me, I’m quite content with Linux 2.6.1 and the NPTL. Together they provide (or exceed) the performance levels that the FreeBSD team has been promising with ULE and KSE, but already stable and production quality. I have what I want now, no need to wait 6 months to a year for FreeBSD.
I really like FreeBSD, it’s a solid well thought out OS – ports is a great package management system. But when it comes to desktop deployment I see no effort to work with Robert Love on Project Utopia fx. which at some point would require work on the FreeBSD kernel.
I can’t wait for 5.x to hit -STABLE (I really hope that 5.3 will make it, but then again it wouldn’t be the first time it has been pushed back)
I wouldn’t expect any desktop thoughts from the BSD team itself, however, I wouldn’t be surprised if some distro maker would start thinking in terms “Okey we have 1000 Linux distros and that market is way to filled and no money is over for us, whatabout BSD?” and suddenly we see a BSD distro which runs with BSD 5.x + a lot of smaller apps which makes configs a lot easier (which I myself would gladly pay 100$ for).
gfx util for Ports/packages wouldn’t be to hard, and because of BSD’s nature (structured), it’s definitely so that making a distro would be A LOT easier than with GNU/Linux.
As for me, I’m quite content with Linux 2.6.1 and the NPTL. Together they provide (or exceed) the performance levels that the FreeBSD team has been promising with ULE and KSE, but already stable and production quality. I have what I want now, no need to wait 6 months to a year for FreeBSD.
Since BSD systems runs for several years, the rush for upgrades are not the same as for Linux. Price on hardware really isn’t that expensive neither so if performance differs just a little, other issues would be crucial for decision but not this one. Issues that come to mind is stability/security/lowrated administration/keepingup2date etc…
I’m confident BSD as of now and as of the future do very well in all those areas and can compete with any Linux distro seen on the market today…
Perhaps you are thinking of something like Barry (KDE app):
http://www.student.uni-oldenburg.de/frerich.raabe/barry/
More stuff like that would be needed and packaged to a distro which makes this pop up as standard in default gnome install etc..
It’s about doing all the stuff for the user frmo the beginning so they don’t have to do it themselves…
A graphical installer is another thing which comes to mind.
Automatically set up X without interference from user etc..
The real question about 5.2.1-RELEASE is whether it will make the ULE scheduler the default, and will KSE threading be enabled per default.
No, it probably won’t. You obviously don’t know how FreeBSD development works (it’s different from, say, the Linux kernel) – if you have some knowledge of the FreeBSD development process, you know why. Both haven’t been the default in -CURRENT, therefore they haven’t received widespread testing yet, and will not be in a release.
ULE has just been made the default in -CURRENT (just after the 5.2-RELEASE, as planned). This allows all the people who run -CURRENT to start testing ULE, so that it’s stable in time for the 5.3-RELEASE. Same for libkse, which will be renamed to libpthread in -CURRENT soon.
They should definitely not make them the default in 5.2.1, because they haven’t been tested enough in -CURRENT yet – they weren’t the default.
And KSE has just been made default in -CURRENT too. So expect the 5.3 release to have both ULE and KSE as the default… I think this was Mullighan’s point though: it would now seem like a runshed step if 5.3 were to be made stable being the first release with these new things.
“I’m confident BSD as of now and as of the future do very well in all those areas and can compete with any Linux distro seen on the market today…”
Technically it can compete in many areas, that’s without doubt, but it simply cannot compare with the likes of RHEL or similar distros. Firstly, commercial support is much broader, and secondly, RHEL offer 5 years of fixes for each release.
To you or I that may not be a big deal, but in business it matters. A lot. Large companies don’t want to spend time upgrading their servers and/or workstations every 12 months, re-testing their applications and making sure everything works OK.
So to say FreeBSD can directly compete with ANY distro is well off the mark. This may change in time, but at the moment it’s simply not the case.
Technically it can compete in many areas, that’s without doubt, but it simply cannot compare with the likes of RHEL or similar distros. Firstly, commercial support is much broader, and secondly, RHEL offer 5 years of fixes for each release.
Large companies don’t want to spend time upgrading their servers and/or workstations every 12 months, re-testing their applications and making sure everything works OK.
Having this 5 year perspective on things make the decision process even more complex for a bunch of reasons
a) the SCO discussion (This is NOT meant as flamebait, no one knows how that will end, it’s US Law, and money is involved, it’s a lottery)
b) Will Red Hat survive 5 years???
c) Fixes etc, sure… but is Linux really that backwards compatible? It’s not a unified system and I get the feeling that apps from 3 years back don’t work that well on Linux systems released today.
BSD is open, has a long history and is not likely to fade away with it’s gigantic lojal userbase. Wouldn’t that due to above issues make it a choice just as good or maybe even better now that I think of it?
Besides, if business stability is on a 5 year perspective, Sun is the way to go…
I think it would be a disaster if FreeBSD attempted to emulate the desktop-friendliness of the various Linux distributions.
I think that Linux is currently suffering from MS-syndrome in the sense that they are trying to make one operating system fit the needs of everyone from big-iron down to people more worried about skinning their desktop.
FreeBSD’s strength is that it does not pander to these needs and concentrates on keeping things sane, simple and stable.
Well Redhat has over a billion in the bank so i don’t see them going out of business just let, then again i was never good at business studies
As for apps that are three years old, well RHEL will keep abi compatability for 5years, so you write and app for it today it will run in 5years. What you got to remember even though Linux isn’t quite a unified system that companies like Redhat, just don’t grab the packages off the net and release them, they tend to do a fair bit of testing to make sure it fits into the system. Hell their kernel is quite heavily tweaked from the standard, helps having Alan Cox and Dave Jones as employees that’s for sure
XBe:
A) I know what you mean, but whatever happens, Linux will continue to progress. In the 1000/1 chance that SCO wins, the few offending files will be replaced in a matter of days. This isn’t a problem.
B) Red Hat almost certainly will be around. But that’s not the point — even if RH collapsed tomorrow, someone like Progeny would step in and offer updates (as they have with older RHL distros).
C) … “I get the feeling that apps from 3 years back don’t work that well on Linux systems released today.”…
OK, you get that feeling, but go out and try it. I’m running the XFM file manager here, packaged for Red Hat 5.2 (ie 1998), on a 2003 Slackware 9.1 installation! I’ve had success with many other older packages too. Anyway, that’s not really the issue; on production servers and workstations you don’t want to be on the bleeding edge. Red Hat backports security fixes, so there’s no worry about compatibility.
In all, FreeBSD is a great server OS (I’ve relied on it many a time), but there are a couple of genuine benefits something like RHEL offers.
“Besides, if business stability is on a 5 year perspective, Sun is the way to go…”
Might be true, but you were saying that FreeBSD can compete with any Linux distro. And I’m just pointing out that FreeBSD can’t always compete with RHEL in serious business scenarios, for the above reasons.
Sam:
“I think that Linux is currently suffering from MS-syndrome”
How can Linux suffer from anything? Linux is a kernel; you mean that certain distros are trying to emulate Microsoft. And that’s fine. But that’s not Linux. There are plenty of distros out there heading for completely different goals. Don’t let Fedora and Mandrake skew your whole view of Linux.
“FreeBSD’s strength is that it does not pander to these needs and concentrates on keeping things sane, simple and stable.”
But you can say exactly the same things about Slackware. It’s cleanly designed, simple and stable too — in fact, the closest to FreeBSD. So FreeBSD’s elegant design is not a unique thing; it just happens that some of the mainstream distros are trying to do more on the desktop. Again, don’t let them skew your view of Linux as a whole.
Fair enough you might have some points there…
so RHEL is the way to go basically if you consider Linux that is… 1 billion in the bank sound convincing I agree, at least on a 5 year perspective… but RH is also very expensive in comparison.
Interesting issue to discuss anyway…
I try to look at this argument from a different POV… from a technology standpoint, BSD was far ahead just a few years back, now that has changed… the Linux kernel has caught up quite quickly.
Assuming that the GPL is the licence for all GNU/Linux parts of the OS (which is a correct statement), GNU/Linux will continue to catch up and pass FreeBSD (it’s debatable if it’s already passed BSD).
The basic reason for this is that GNU can take BSD licence code, and relicense under GNU, but BSD can’t take GNU code and relicense under BSD, so GNU has much more code at their disposal. This also goes for code outside the OSes themselves. All code and libraries used on a GNU system benefit from companies using them and improving them, BSD doesn’t get this benefit, just look at Microsoft, they took much of BSD’s networking and socket code.
BSD comes from a much more ‘closed’ development group. It was originaly a university project, in fact, the original 4.1-4.3 releases of BSD (not freebsd) had much Unix code left in them. In fact even 4.4 Release was found to have 6 files of unix code that were removed and later repackaged as 4.4-Lite (which hurt BSD’s development in the early days). GNU/Linux is a much more open development enviroment… anyone can make changes, and there’s a pretty good chance that if they’re good, they’ll find their way into the official packages, but not likely with BSD.
BSD also takes much longer to release software, which can be considered a positive or a negative depending on the view point.
Back to the original point of this post… companies and people seem to destroy GNU/Linux IMO… there’s too many differences between distros. I don’t see a reason to have 4 different ways of splitting up the rc.d files, or other configuration scripts. Also there’s far too many packaging systems. The next problem is that business controls most of these things now. GNU/Linux went from an underground effort to a corporate effort, which is casuing much confusion and politics. Because of BSD’s roots and development style, this probably won’t happen.
This leaves GNU/Linux as a much more commercial endeavour, and BSD as much more underground toy, choose which one you prefer, but I suggest sampling both.
>The real question about 5.2.1-RELEASE is whether it will make the ULE scheduler the default
No, it won’t be. I asked.
“This leaves GNU/Linux as a much more commercial endeavour, and BSD as much more underground toy, choose which one you prefer, but I suggest sampling both.”
Toy is it? Hmm. I love when people try to further their point of view by appearing to be objective. Good try.
ULE has just been made the default in -CURRENT (just after the 5.2-RELEASE, as planned).
As planned where? Certainly not here:
http://www.freebsd.org/doc/en_US.ISO8859-1/articles/5-roadmap/sched…
5.2-RELEASE was supposed be the first production quality release with all the new features on by default, and it was supposed to be released in September 2003. Instead 5.2-RELEASE seems like nothing more than a slightly slower and less stable version of 5.1-RELEASE, with no new technologies enabled. Hopefully 5.2.1-RELEASE will address these concerns.
This allows all the people who run -CURRENT to start testing ULE, so that it’s stable in time for the 5.3-RELEASE
Wow, now *THAT’S* wishful thinking. How many people in the world do you think run FreeBSD 5-CURRENT? Perhaps a few thousand? Do you think these people alone are enough to take the largely untested ULE scheduler and KSE threads implementations to production quality in a period of 3 months?
The current schedule calls for 5.3-RELEASE near the beginning of April 2004. It is only after this time that ULE and KSE will see widespread use. Do you expect to see the 5-STABLE branch with the 5.3-RELEASE? I would certainly hope not.
Looking at http://www.freebsd.org/projects/busdma/ there is still a considerable amount of work to be done. If you read http://www.freebsd.org/doc/en_US.ISO8859-1/articles/5-roadmap/goals… you will see that one of the requirements for 5-STABLE is transitioning these drivers to use the new interface. The same goes for http://www.freebsd.org/smp/#status
The bottom line is the current information is putting the release of 5-STABLE at at least a year off, when the previous release schedule would’ve put it around now, with RELENG_5 having been forked back in September 2003.
For a scalable, production quality server, Linux 2.6 + NPTL is ready to use today. If you want this sort of thing on the FreeBSD side of things you’re going to have to wait at least a year, or stick with 4.x, its deplorable userspace threading, its antequated 44BSD scheduler, and its horrible SMP performance.
For the last time….
You cannot just take BSD licensed code, strip the license off it and put a GPL on it. Go read up on:
1. copyright law
2. the BSD license itself.
Here is the relevant part for you since you are probably too lazy:
1. Redistributions of source code must retain the above copyright notice…
My, you seem to have quite a lot of time to hang around and evangelize Linux. How much do you contribute to either Linux or BSD in terms of patches or official bug reports?
How many people in the world do you think run FreeBSD 5-CURRENT? Perhaps a few thousand? Do you think these people alone are enough to take the largely untested ULE scheduler and KSE threads implementations to production quality in a period of 3 months?
A few thousand people/systems is a big amount of testers and certainly more than enough to get it ready.
The current schedule calls for 5.3-RELEASE near the beginning of April 2004. It is only after this time that ULE and KSE will see widespread use. Do you expect to see the 5-STABLE branch with the 5.3-RELEASE?
I’m using ULE in 5.2-RELEASE right now on 3 systems (1 workstation, 1 server/firewall and 1 laptop). Soon I’ll start using KSE too. Just because it isn’t made default doesn’t prevent people running 5.1.x-RELEASE from using it. Just check previouse news here on 5.2 and ULE and you’ll see many others trying the same thing so ULE and KSE will see a lot of USE even in 5.2
For a scalable, production quality server, Linux 2.6 + NPTL is ready to use today. If you want this sort of thing on the FreeBSD side of things you’re going to have to wait at least a year, or stick with 4.x, its deplorable userspace threading, its antequated 44BSD scheduler, and its horrible SMP performance.
production quality server?!! With a .0 major new release? Are you nuts? No serious (as in “doing this for buisness”) person would even consider that.
NPTL is (on the paper at least) an inferior threading-model for SMP-systems compared to KSE and M:N. M:N threading is much more complex to implement than 1:1 threading which is what has delayed this project. Once KSE gets some use expect 5.4 and onwards to really start shining in the performance section. I’m guessing that Linux eventually will move to M:N threading also but then FreeBSD will be far ahead in implementation.
Given this, I have no reason at all for using Linux over FreeBSD especially given the interactivity level ULE is giving me on my FreeBSD/Gnome-desktop on my laptop. It’s just that snappy
Just because it isn’t made default doesn’t prevent people running 5.1.x-RELEASE from using it.
should have read:
Just because it isn’t made default doesn’t prevent people running 5.2.x-RELEASE from using it.
For a scalable, production quality server, Linux 2.6 + NPTL is ready to use today. If you want this sort of thing on the FreeBSD side of things you’re going to have to wait at least a year, or stick with 4.x, its deplorable userspace threading, its antequated 44BSD scheduler, and its horrible SMP performance.
I probably will. I’m yet to upgrade from 4 Stable, and I must say that performance isn’t that different from 2.4 linux kernels on my “older” hardware.
Until Linux 2.6 really is stable and really is part of the mainstream distros I probably won’t upgrade. Same goes for FreeBSD 5.x. I’m fairly happy with what I’m running now, so I don’t mind waiting.
As for scalable production quality servers. I know of few setups in a production environment that need the kind of scalability & stability you are claiming. I doubt that many that do are running bleeding edge kernels like FreeBSD 5.x or Linux 2.6 at this stage of the game. They would have to be a lot braver than I am.
The networking stack panics all the time. This is the
worst quality fbsd release ever.
Yes, Linux 2.6 is definitely ready today. They took care of a few remote root exploits, and I’m sure your servers will help find the rest.