F2FS is still considered Unstable as of Kernel 3.8, and so far no stable release has yet surfaced from Samsung.

I've been re-reading up on UFS2 under Linux, and I'm surprised little effort was made to get it able to write to the file system rather than be read-only. It's description and features read fairly well and it's stable and solid. If it ever got to Linux it would be a fairly useful little bugger of a file-system.

Thats too bad, especially since more people are getting SSD drives. Don't get me wrong ext*, XFS, JFS, etc are great choices and each have their own strengths, and I personally like JFS/XFS but these FS were never designed with flash devices in mind, only for a conventional drive which for me XFS or JFS will always be a choice since I will always have a conventional drive to supplement my system, as well as an external 1TB. There have been other FS for flash devices like JFFS but don't know how much of that is being worked on at all. E2FS seems a logical step and hopefully it will become stable.

I don't know about U2FS will ever come to Linux other than FUSE, because of licensing if I am not mistaken? Just like ZFS, FreeBSD has it, but it's license allows it, GNU GPL does not - and ZFS seems also a very strong option , but it's only counterpart available is BTRFS which is also not marked as stable (yet in Slackware you do have the option of using that a / partition - but E2FS is not there, oh well ).

I use Linux (Slackware mainly), OS X and Windows 7, and work in a mixed OS environment (Ubuntu is the main Linux there though), so the filesystems I use are:

Ext4: Root filesystem on Linux installations, no real reason, used by many and I went with the flow I guess?
Ext2: Good for portable drives whilst preserving file permissions across Linuxes (I think it's supported by many BSDs as well).
HFS+: OS X needs to boot somehow.
NTFS: Windows use and good for portable drives (read/write support available on Windows / Linux / OS X, allows files >4GB)
FAT32: Good for portable drives (read/write support on multiple OS, for files < 4GB) and for 1 bootable USB stick (contains SystemRescueCd, Slackware 14 installer (32 and 64 bit), Gentoo installer (also 32 and 64 bit))

I have a small collection of external hard drives and USB sticks with different filesystems, it gets confusing at times.

Although it's not too well documented now, UFS2 was possibly under the BSDL last I noticed, and BSDL is GPL compatible.

ZFS is licensed from Oracle under CDDL which isn't compatible with GPL because CDDL doesn't have strict guidelines that a change to the software code, has to be readily published within the core of the project, even though ZFS receives constant updates that are public from all it's contributors and ports.

Oddly however the goals of BtrFS and ZFS are exactly the same, but Oracle absolutely refuses to re-license ZFS even though they solely own the code now. I highly doubt BtrFS will ever be stable enough for usage any time soon.

I did attempt to rebuild Slackware for ZFS support and even reconstitute my LFS build into a ZFS partition and zpool, but the build kept bugging me about a kernel version header not being found, and thusly I'm at an impasse.

Edit:

Apparently, I didn't know this but there is an Open Source version of ZFS being developed independent from Oracle by the Illumos project along with some contributions by FreeBSD, so apparently 2 branches of ZFS exist, one used internally at Oracle, and the other used by FreeBSD and Illumos.

Interesting...

I thought I might add something about ZFS: as nice as it might be, it still doesn't have an utility to check it for errors/corruption. I know it's a CoW filesystem, but still...

Note that JFS has surprisingly good performance on SSDs, much better than even btrfs. I am hoping they release a more comprehensive benchmark of all filesystems available.

http://www.phoronix.com/scan.php?pag...38_large&num=1

I never quite understood this; wouldn't Oracle *want* ZFS to be usable on Linux so as to use it on their own Oracle Linux without risking some kind of legal hot water pertaining to distribution and derived works?

Oracle doesn't own all the copyrights to ZFS.

They owned a majority of it, including the name rights, up till about 2011 and still maintain an in-house version of ZFS with proprietary code, but now "Open Source ZFS" is controlled by the Illumos project who are responsible for OpenSolaris's spinoff project OpenIndiana with contributions being made between them and FreeBSD.

Because, however Oracle still has a stake in ZFS, and because Illumos now controls the majority of the projects development and Illumos is under a CDDL license, they can't change the license of ZFS to anything else at the moment.

The problem would come down to the GNU/GPL instituting a yet to be conceived GPLv4 license that would be compatible with CDDL like The BSD License and The MIT License. And so far the GNU absolutely refuses to do so....

...or Oracle's developers would have to be in complete agreement with the Illumos team to publicly release all the ZFS source code under a BSD or MIT license, and that isn't going to happen any time soon either.

You'd have a better chance of Lennart Poeterring and Kay Sievers come out publicly saying they are dropping systemd development and scrapping the project.

Even if they would do that it would change absolutely nothing, since the Linux kernel will remain GPLv2 licensed.

Any way you look at it, getting ZFS into Linux isn't easy.

Which brings to question, why don't we even have a SlackBuild of ZFSOnLinux, instead of just the ZFS-FUSE package?

Are you volunteering to write a SlackBuild for it? After all, SlackBuilds are created and maintained by volunteers.

There seems to be a zfs SlackBuild in the pending submissions queue though I don't know if it is specifically ZFSOnLinux.

The only project currently on SlackBuilds that is ZFS related is zfs-fuse. If ZFSOnLinux does have a pending submission under ZFS, I was not the one who submitted it, but best of luck to whomever did.

I store my data on XFS partition and (as I test many distros) each of the systems on the partition they suggest (mainly ext4), however with Slackware I usually try JFS.

I'm looking to move away from ext4 because of this issue which is marked as [SOLVED] only because I found a way around it rather than fixed the bug. I switched to BTRFS and have had NO issues whatsoever. But I have not used it heavily that way. I'm also still worried about that issue with ext4 where asynchronous meta changes can be permanently lost under conditions like renaming files, possibly losing files in cases where these changes result in zero links. I don't know that BTRFS is not vulnerable to this issue, but it has happened to me with ext4 so I know it's a real issue.

Everyone says BTRFS is unstable? Why do they say that? Could it be because everyone says BTRFS is unstable? You know if you say it enough, even YOU will believe it. But has anyone shown BTRFS to actually be unstable in the latest kernel? Is it reproducible?

After reading a bit of this thread and doing some research - my mind now opened to non-ext? filesystems - I've decided to experiment with JFS some more. First test case will be an old Iomega StorCenter ix2-200 that I happened to get second-hand; the Kirkwood-based ARM SoC is a perfect fit for SlackwareARM, and it sounds like JFS' low CPU requirements would be great for the embedded Marvell ARM CPU.

The setup is pretty straightforward; 128MB ext2 /boot, 100GB JFS /, 850-ishGB JFS /home, and 1G swap, each on a RAID 1 /dev/md* device (in other words, somewhat similar to the stock firmware setup, but with swap and without funky loopback devices for /). Surprisingly, U-Boot handles the RAID 1 /boot just fine and succeeds with loading the uImage and uinitrd. However, the kernel seems to be doing funny things with the RAID detection, so it'll take a bit of troubleshooting before it's fully up and running.