It is interesting to note that FreeBSD developers are giving consideration to phasing out sysinstall in favor of a new install program for FreeBSD. A Journaling FS like ReiserFS also seems to be on the cards for future FreeBSD releases, Kerneltrap reports.
FreeBSD Project Discussions on SysInstall, FS
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
I like ESPECIALLY the idea of working on bsdinstaller.com . Looks nicer, sweeter and simply better than ncurses.
Following that, I can hardly argue about virtual keyboard manager, which is obviously intelligent. Heck all of his mentioned ideas seems very wise. I really hope these are changes which get considered quite soon.
Good thinking Scott Long =)))
I really dislike the idea of implementing some GPLed filesystem such as ReiserFS. What would be really interesting though is if FreeBSD could branch HaikuFS and perhaps work on that to implement multiuser features and they have journaling in there already plus a suitable license.
Maybe this is an option?
I don’t know how well the installer works, but it seems like the CGI scripts are a new and useful idea. While I have no problems with ncurses, and think that ncurses “looks nice”, a lot of people have this mental block when it comes down to text only interfaces. Hopefully serving up HTML will get around those mental block and be a more reliable solution than running an X server during the install process.
It may also encourage people to install over a networks, and reduce the number of old *BSD/Linux distro CDs going for landfill.
What ist “haikuFS”?
HaikuFS is the filesystem for Haiku. It used to be called OpenBeOS and that would make the filesystem OpenBFS/OBFS. It’s a reimplimentation of the filesystem from the old BeOS. It’s supposed to be very well done.
PLEASE port reiserfs! then i’ll try bsd… can’t be fucked formatting my hdd.. got over 200gb of data on it… you want more users.. port the filesystems that they use.. if reiser works on bsd, I’ll switch tomorrow… my hardware has full support on bsd, nvidia card, creative soundcard, realtek 8139 network… harddrive however is reiser formatted, and no way in hell am I losing my films for a new os.
A journaled filesystem for FreeBSD? That’s funny, every time it was proposed for the 5.x series it was shot down because USF2 w/ softupdates was supposed to be the better solution. Now suddenly there is a change in heart.
Hopefully, they’ve finally seen the light that UFS based filesystems can’t scale well into current and future storage systems with the need to fsck — in the background or otherwise. Nice to see some common sense for a change. This will definitely help FreeBSD in the long run IF those porting efforts pay off and near religious zealotry for UFS doesn’t get in the way again.
I remembered reading something about BSD LFS, so I searched for it. I found this discussion from September 2004: http://docs.freebsd.org/mail/archive/2000/freebsd-fs/20000924.freeb… , http://docs.freebsd.org/mail/archive/2000/freebsd-fs/20001001.freeb… Reiser FS was being discussed on FreeBSD mailing lists even then.
The discussion also mentioned “two implementations of journaling within UFS.” ( http://docs.freebsd.org/cgi/getmsg.cgi?fetch=115459+0+archive/2000/… ) Perhaps FreeBSD could implement a journaling FS with some degree of compatibility with UFS. (Like Wasabi-JFS)? http://www.wasabisystems.com/products/journaling_filesystem.htm
A journaled filesystem for FreeBSD? That’s funny, every time it was proposed for the 5.x series it was shot down because USF2 w/ softupdates was supposed to be the better solution. Now suddenly there is a change in heart.
Suggested by whom and shot down by whom? I think you’re confusing poseurs dishing out wish-it-was-so ‘facts’ on OSnews with the actual FreeBSD developer community. I’ve seen several FreeBSD developers express the opinion that they need a journaling FS. I’m sure anyone who is considered competent and wants to work on getting such an FS into FBSD would be more than welcome.
What a silly post.
Soft Updates aren’t *always* the better solution – but *sometimes* they are.
A Harvard research shows that the Soft Updates technology is on par with Journaling on the whole, and in certain cases it provides superior performance.
[abstract]
http://www.eecs.harvard.edu/~vino/fs-perf/papers/su-vs-j-abstract.h…
[article]
http://www.usenix.org/publications/library/proceedings/usenix2000/g…
BSD people usually think that having more tools available for solving problems is a Good Thing. The ones who have religious preclusions are usually the GNU folks/zealots.
heck the openbsd fantatics even rewrote cvs in openbsd under the bsd license
The switch to a BSD license is good, but the OpenBSD folks undertook the task of writing OpenCVS for a different (technical) reason.
http://www.openbsd.org/opencvs/
While HaikuFS (OBFS) has the right license and is well documented, it’s written in C++ which makes things difficult for the kernel.
The Open versions that BSD writes are arguably more secure than their GPL counterparts…
”
The switch to a BSD license is good, but the OpenBSD folks undertook the task of writing OpenCVS for a different (technical) reason. ”
no. people can try to pass it off as such but if they really wanted it to be secure they could have just provided the necessary patch instead of wasting time rewriting it
“The Open versions that BSD writes are arguably more secure than their GPL counterparts…”
arguably is the key word here. It would have much better to secure the existing version. they do this for gcc. why not for cvs?
Sure, but only parts is in C++ as far as I know, and there might be parts which can be changed to plain C I guess? If the filesystem is brilliant which it might be considered as, this must be an obstacle that is possible to overcome.
If they could work together, FreeBSD would become a brilliant serversystem for Haiku as desktop. Something I would love to see for the future, what a winning combination.
I hope that devs on both ends would discuss the issue, maybe Axel D + Scott Long that is.
You could alwyas back up your hard drive to removable media or a network location. It’s not so bad.
Because once you go GPL, you never go back…
Soft Updates aren’t *always* the better solution – but *sometimes* they are. A Harvard research shows that the Soft Updates technology is on par with Journaling on the whole, and in certain cases it provides superior performance.
From what I understand, the comparison made in that paper isn’t really relevant to modern journaling fs, but regardless, the point of contention with soft updates is neither performance not data integrity, but rather the need to fsck after an unclean shutdown. Disks are getting mighty big.
there has been a lot of debate on the installer. Usually th side that says “it serves it’s purpose so why chang it’ usually wins.
I would welcome a new installer eveven though I’m developing my own..
Argh. Don’t get rid of sysinstall. The last thing I want is to have to run an X server to install freebsd on a slow machine! With sysinstall I can configure my old slow laptop with little overhead. I find it very useful and just as intuitive as any graphical administrative tool I have used on other systems.
Argh. Don’t get rid of sysinstall. The last thing I want is to have to run an X server to install freebsd on a slow machine! With sysinstall I can configure my old slow laptop with little overhead. I find it very useful and just as intuitive as any graphical administrative tool I have used on other systems.
Who said it would be replaced with some X11 GUI? Take a look at http://www.bsdinstaller.com/ , it has a text-mode front-end and a CGI (http) front-end. No overhead for your old slow laptop.
ReiserFS is nice
this new installer they are talking about would it be omne like anaconda or something nice and graphical?
this new installer they are talking about would it be one like anaconda or something nice and graphical?
Leaving aside for the moment the fact that Anaconda is graphical, you can see screenshots of the installer they are discussing at a link posted in a previous response:
http://www.bsdinstaller.com/
It’s funny how they want now want to focus on reworking an installer when what they should really be working on is getting more hardware support, and getting SMP to work properly.
Does anyone here miss the days when FreeBSD team was headed up by JKH (Jordan Hubbard – who now heads up the BSD component of OS X)
Tsk tsk tsk – I’m disappointed in the FreeBSD group – since when do frilly things like a reworked “sysinstaller” matter more than proper kernel development?
Milione – FreeBSD junkie since 2.x
Ummm, what good is an advanced OS with an advanced kernel if you can’t install it?
Sysinstall is not the least bit userfriendly. The FreeBSD group wants to encourage _more_ people to use FreeBSD — which equates to more testing and more potential developers to work on advanced things like SMP. Having a good installer is kind of a pre-requisit to encouraging growth of the userbase.
Ummm, what good is an advanced OS with an advanced kernel if you can’t install it?
It is not the case. You can install FreeBSD perfectly. FreeBSD is well known by *nix admins because the OS quality. They really don’t care about the install. I prefer a very-good OS rather than an easy to install/use one like windows for example.
Ummm, what good is an advanced OS with an advanced kernel if you can’t install it?
I disagree that sysinstall prevents a proper FreeBSD installation. I don’t think it is a priority more important than hardware support. Is sysinstall more important than making ACPI work for example?
It’s funny how they want now want to focus on reworking an installer when what they should really be working on is getting more hardware support, and getting SMP to work properly.
They are not at all focusing on a new installer. It’s just on their wish list.
Tsk tsk tsk – I’m disappointed in the FreeBSD group – since when do frilly things like a reworked “sysinstaller” matter more than proper kernel development?
Again; it does not matter more than kernel development. Far from it. But it would be nice to have an improved installer.
I suppose we’re talking about reiserfs4 here.
For me these would br the key motivation points of using reiserfs4:
1. no more fsck – no more waits after the machine crashed
2. tails – efficient storage of huge amount of small files (check your man,etc,Maildir,locale,perllib dirs to see why this does matter)
3. metadata – storing metadata will be the next big thing as soon as desktop searching has become standard, ACLs could be stored here too
4. speed – reiserfs4 is darn fast, even compared to the already not so bad reiserfs3
5. filesystem plugins – someone to implement a CVS on th fs-layer?
I think reiserfs4 would be a very nice addition to the bsds especially since there is no other journalling fs available for them (Net/Free/Open/DragonFly) (afaik).
“Hopefully, they’ve finally seen the light that UFS based filesystems can’t scale well into current and future storage systems with the need to fsck”
So fsck in background. IIRC thats possible. Deraadt wrote about this in an interview a while back, referring to above Usenix paper.
Hopefully, they’ve finally seen the light that UFS based filesystems can’t scale well into current and future storage systems with the need to fsck — in the background or otherwise. Nice to see some common sense for a change. This will definitely help FreeBSD in the long run IF those porting efforts pay off and near religious zealotry for UFS doesn’t get in the way again.
Journaling still has trouble dealing with cases where a file is being written to as power is lost. ext3, for example, will simply delete the file. XFS fills it with zeroes. I have not tested reiserfs, and would be interested to know how it behaves.
With UFS + SoftUpdates, the file simply exists in the form of whatever changes have been written out before power was lost. It isn’t damaged by filesystem recovery in the way that journaling filesystems are.
You should read the thread: Scott’s intent was just to generate some interest and discussion about new projects.
with all this talk of ReiserFS 4, has it been included inthe main kernel tree yet?
1. The installer is most likely being worked on by people who are not doing advanced SMP stuff. So to complain that nobody should work on the installer is foolish.
2. I think it was clearly stated that they were thinking of collaborating with the dfbsd guys and using an installer that has already been (for the most part) written.
3. These are volunteers here, not paid programmers. They work on what interests them and what they feel is most useful.
I hope this clears up any misconceptions.
Sysinstall is not a unfriendly installer. Trying installing Gentoo or Slackware if you want to know what unfriendly is.
I don’t see what the problem with the existing FreeBSD installer is. Sure it won’t satisfy the needs of Joeno Nuthing, but that isn’t FreeBSD’s target audience. If Joeno wants to use Unix, he always has Linux or Mac OS X. The main advantage I see to the new installer is its use of CGI, because ethernet and http are quickly replacing a terminal emulator on a serial port.
Oh, and I see multiple file systems as a strength of Linux. It has allowed Linux to change file systems while maintaining backwards compatibility. It allows you to easily exchange data between different operating systems on the same machine. It even has some advantages if you like playing around with emulators (it is dead simple to exchange data between KEGS and Mac-on-Linux in Linux and Mac OS X). I have always found the BSDs lacking in this respect.
with all this talk of ReiserFS 4, has it been included in the main kernel tree yet?
reiserfs and reiser4 are GNU-GPL, in order to incorporate them into any BSD kernel, a liscense change would be required, meaning reiser would have to be rewritten from scratch, or BSD would have to adopt the GNU-GPL.
Namesys, Hans company, is also sponsored by Linux distributions: Lindows and SUSE.
Couldnt the freeBSD guys just port HFS+ from openDarwin, i’ve not read the apl, but im sure that with darwins similarity to FreeBSD it cant be too difficult
With UFS + SoftUpdates, the file simply exists in the form of whatever changes have been written out before power was lost. It isn’t damaged by filesystem recovery in the way that journaling filesystems are.
Ext3FS has several journaling modes. Quote from the manpage:
journal
All data is committed into the journal prior to being
written into the main file system.
ordered
This is the default mode. All data is forced directly
out to the main file system prior to its metadata being
committed to the journal.
writeback
Data ordering is not preserved – data may be written into
the main file system after its metadata has been commit-
ted to the journal. This is rumoured to be the highest-
throughput option. It guarantees internal file system
integrity, however it can allow old data to appear in
files after a crash and journal recovery.
Which one are you referring to?
There are already some tools for working with HFS volumes in the ports tree. Only last time i checked no support for journaling, but that shows at least some work is being done in that area.
uhm… isn’t Matt a little sore about Scot and team?