Hi,

I want to use my external discs (and internal partitions) with both Linux and MacOSX and I am wondering about the best file system.

The file system I want should at least support:
- common Unix stuff (symbolic links, etc.)
- extended attributes
- ACL
- journaling
- very stable (also for crashes / sudden power loss)
- fast

On Linux, I am using ReiserFS3 right now. On MacOSX, I am using HFS+. The Linux HFS+ driver doesn't support journaling. MacOSX doesn't seem to support ReiserFS at all. So, no good situation right now.

It would also be nice if both system could also be installed on the disc (and boot from), whereby this is not necessarily needed.

I was not using ext3 because I had some bad experiences some years ago with sudden power loss (I destroyed the file system by that). But as this was a long time ago, probably the situation is better now. Anyway, I only found information about ext2 support on MacOSX (with a quick search). Does MacOSX support all my requirements with ext3?

Which other alternatives are there?

Thanks,
Albert

I use Ntfs because it support files >4gb and it is r/w.

I'd like to use ext3 but I didn't have found a driver that can write on it.

MacFUSE & FUSE ftw

Which of those is supported by NTFS?
- common Unix stuff (symbolic links, etc.)
- extended attributes
- ACL

Also, is the journaling fully implemented in the FUSE-driver? And what about the stability and performance?

As far as I know, the NTFS-driver is usable to access data now once in a while (read&write), but it is far from stable/complete as any driver in the Linux kernel (compare it to ext3 or reiserfs).

If I am doing my work, I never want to have any problems because of the filesystem and I cannot be perfectly sure with NTFS.

I don't think there IS a filesystem that satisfies your requirements. The ONLY filesystem which both OSes can read and write natively is FAT32.

NTFS-3G is pretty good in OS X, with MacFUSE, much faster & more stable than it used to be. Still not as good as native obviously. I use it mostly due to the 2gb filesize limitation of FAT32. UNIX Symlinks are not supported on NTFS-3G. I'm really not sure about the extended attributes and ACL.

ntfs-3g for mac is quite stable but with many concurrent writings you could corrupt it.

There are no alternatives to NTFS today.

Actually speed has lower priority for me than the other requirements. But the other requirements (at least symlinks, extended attributes and stability) are absolutly mandatory.

It seems I have to continue to use my external drive only on a Linux machine and access the files via Samba or SSHFS (or some other method; often I am just working with SSH X-forwarding directly on the Linux machine if I have to run it anyway) right now to get this.

I am really wondering why there isn't a modern Unix filesystem yet (modern means, it should support all of my requirements, but all FS in the last 5 years should do that anyway) which is supported by all major *nixes (or at least MacOSX and Linux).

Personally I am of the opinion that filesystems are one of Linux weakest points. I don't know much about ext4 yet, and I find it really a shame that so many years work on Reiser4 were wasted (regardless of Hans Reiser's conviction).
Adopting ZFS could be a solution, but they won't because of the licence.

ZFS is installable, and working under Linux.. Ubuntu at least has a ppa for adding it through Synaptic ( https://wiki.ubuntu.com/ZFS ).. I'm running it in a mirrored zpool on my (Ubuntu) server's data drive.

The problem is, it's not as people conceived of it.

It's got to be imported/exported every time you reboot into a different OS on both OS's no less, manually, through the command line, which while isn't difficult, isn't necessarily efficient either.

my favorite is ZFS, as pointed out before, it has clients on both. it also has very very good portability, and if you were removing it from a osx machine to a unix machine, you would find that you need to "park" most file systems so that the journals are updated. in unix/linux that is done automatically on shutdown.

ZFS, does not require being parked, however it will be happier if you park it.

I have done tests with live systems, pulling the plug while read/writing to a ZFS mount, and then mounting it in another system.

in all my tests it came back without Any file loss.

JMHO

Hbp

IMO, filesystems are one of the strongest point of Linux. No other OS has support for so many different file systems. And it's not only the quantity, also the quality (technic/implementation/features) of most of the more modern filesystems is very interesting and amazing.

ext4 is considered as being stable right now (at least in the Kernel) and some distributions are starting to use it as the default.

Reiser4 development was absolutly independent from ext4. So even if you would say that work was wasted here, that didn't had any effect on other filesystem development. I for myself would also not say that the work for Reiser4 was waisted. It introduced a lot of nice ideas which are going to be implemented in other filesystems. And Reiser4 itself could also make it to the stable Kernel tree once when the code is cleaned up.

Any development on any filesystem on Linux will only help the general development of filesystems on Linux.

There are also further very stable filesystems available on Linux, for example JFS or XFS.

Also other further development, for example on Btrfs, seems very comprising. (Again, that doesn't effect the development on ext4 or Reiser4; that are mostly different people.)

You cannot see such wide-spread and heavy development on filesystems and such wide-spread support in any other OS than in Linux. And any argument like "they" should concentrate more to one filesystem is just invalid, because there is just no single group like "they". There are thousands of groups which different interests and if they need, they will develop their own filesystem, for the good of everybody.

Btw., ZFS sounds interesting, I will try it out. (Just because the licence will force ZFS outside of the kernel does never mean that it will/cannot be used by Linux users.)

Although development has stagnated a bit until a few months ago, FS'es are still very strong on Linux. There's ext2/3, XFS, JFS and there used to be ReiserFS. All of them are mature, stable, fast and feature-rich. And recently there've been developments on ext4, btrfs and nilfs, so the future is bright.

And hey, anything beats HFS+. ZFS is awesome, but with Apple you never know what kind of restrictions they are going to impose on this otherwise beautiful piece of software by Sun.

Btw, NTFS-3g doesn't seem to support posix permissions, let alone ACL's. Everything is root:admin 777 here.

Besides, performance on OS X really sucks. While Linux can read as fast as 105MB/s, OS X tops at 40MB/s with a disappointing 30MB/s average.

Your best bet would be HFS unjournaled, but I'm not sure you want that.