[IDEA] Reiser4

[Quikee]

Re-evaluate the already existing reiser4 specs (https://wiki.ubuntu.com/Reiser4) for Edgy. Reiser4 was deffered in Edgy but ignored for Feisty.

Reiser4 is in some situations preferred to other file systems (even if it is experimental) because it supports transparent compression, which can be very useful at installation to a USB stick for example.

If the specs will be deferred again, at least add either the reiser4 module to restricted modules or make the installation easier for the people that want reiser4.

[Erunno]

As long as Reiser4 isn't included in the upstream vanilla kernel I doubt that Ubuntu will add it on their own. With Hans Reiser current private problems and the announced sale of Namesys to fund his lawyer the future of Reiser4 seems to be on the edge of a knife.

[Quikee]

As long as Reiser4 isn't included in the upstream vanilla kernel I doubt that Ubuntu will add it on their own. With Hans Reiser current private problems and the announced sale of Namesys to fund his lawyer the future of Reiser4 seems to be on the edge of a knife.

If you read the specs you'll see that this was the idea even if it is not included in the vanilla kernel.. I don't see a reason why it should not be included as a "universe" package for those that want it. The sooner it is included into a distro with a wider test base, the sooner it will "stabilize" and will be added to the vanilla kernel. Reiser4 is currently the most advanced filesystem in Linux (until somebody ports ZFS of course) - it would be a shame to throw it away just because Hans Reiser has some "personal" problems.

[qamelian]

It's also the filesystem that gave me the most grief when I used it for a while. I agree it would be good to have it as as option for folks who want to use it, but I'll never touch it again.

[gnomeuser]

Andrew Morton publicly stated that if one or two vendors picked up Reiser4 for testing it would be much more likely to go into the kernel. Adding support as an optional feature would help get testing done and anyone enabling it would likely understand that it's experimental software.

We already include things which isn't in the upstream kernel (drivers, devicescape wifi stack, etc.) so that shouldn't be a definitive blocker. What could block it would be that for optimal implementation it would be nice to have someone like Nate Diller working fulltime on filesystem/vfs bugs and that means hiring a kernel hacker (convincing Canonical to do that might be a toughie even if it would mean raising the overall quality of our kernel).

[bicchi]

I think ext4 (http://en.wikipedia.org/wiki/Ext4) will make it into Ubuntu before Reiser4. In the end ext4 has more chances of being tested since its already in the kernel. The benchmarks don't look that bad considering is still in development. But who pays attention to benchmarks anyway. Those things shift depending on the end user testing scenario.

[metnik]

I'd like to see Reiser4 at least in Universe.
Ubuntu supports proprietary drivers, so I don't think it's a bad idea to have a kernel patched for Reiser4.
I'm an happy Reiserfs user, and I used Reiser4 in gentoo in the past, all the benchmark I see are very positive about this fs

http://kerneltrap.org/node/7992

[Enigmatic]

I don't know what everyone's experiences with Reiserfs are like, but I am regretting that I reformatted with an ext3 partition for / and /home/ - reiserfs had much less overhead and seemed speedier for any disk related activites.

Would it be possible to just convert the partition type using a partitioner, and assuming that fstab is changed to reflect the new fs, not lose any data?

[TheMono]

Nope. Doesn't work like that. You need to move the data off, then reformat, then alter the fstab, then move the data back.

[hugmenot]

[QUOTE=Enigmatic;2510113]I don't know what everyone's experiences with Reiserfs are like, but I am regretting that I reformatted with an ext3 partition for / and /home/ - reiserfs had much less overhead and seemed speedier for any disk related activites.

I reformatted as XFS after an HDD failure and regret this as well. Who again claimed that XFS was faster?

[bicchi]

Just recently this came on kerneltrap Linux: Reiser4's Future (http://kerneltrap.org/node/8102)

[foerdi]

+1

we need reiser4

[aamukahvi]

Reiser4 is currently the most advanced filesystem in Linux (until somebody ports ZFS of course)
ZFS is already ported, sadly only a FUSE implementation (licensing issues) . http://www.wizy.org/wiki/ZFS_on_FUSE

[waster]

we need ZFS. shame about the licence.

[gnomeuser]

So we don't carry Reiser4 even as an experimental feature because it's adding a patch to the kernel.. but we are now carrying AppArmor, a wrongly designed security feature which as it stands has no chance of getting in the kernel.. seems like bad logic to me.

The only reason not to carry Reiser4 given that policy is that it's not currently being developed a lot and given that it's a filesystem we risk data corruption - but being labelled experimental (or even supported in the way Fedora supports non-ext filesystems but specifically listing them during the install as enabled).

By getting Reiser4 in a real distro we could kick start development and get it the widespread testing Andrew Morton called for before it could go into the mainline kernel.

[Trevice]

I have probed ext3, reiserfs, xfs, reiser4 and ext4

and after make a lots of measures... the winner is ext4
yes, reiser4 its very fast, much faser than ext3 or xfs

but last weeks i probed ext4 and I surprise myself....Its increible faster!
if you dont believe me, just take a look the numbers

Ext3 Reiser4 Xfs Ext4 Ntfs
14 10 5 3 23 (1)
13 12 12 12 - (2)
27 25 102 23 289 (3)
1:05 56 187 22 +13min (4)
34 36 167 4 14 (5)

(1) Copy of a single file of 180mb inside the partition
(2) Copy from externlar usb disk to the partition (again 1 single file of 180 mb)
(3)Decompression of the kernel source (over 42.697 files, a total of 435mb)
(4)Copy inside the partition of all de kernel source (over 42.697 files, a total of 435mb)
(5)Delete of the kernel source (over 42.697 files, a total of 435mb)

all the measures are in seconts

Conclusions: xfs is very slow with small filest, ext3 is not optimiced, ntfs is a ****.....and ext4 is god!!! :P
sorry for the horrible table


and yes, I have a lot of free time.....

[ubu-for]

ZFS is already ported, sadly only a FUSE implementation (licensing issues) . http://www.wizy.org/wiki/ZFS_on_FUSE

Could ZFS be the next default filesystem?

but last weeks i probed ext4 and I surprise myself....Its increible faster!

Awesome!

So we need ZFS or ext4! :mrgreen:

[aamukahvi]

Could ZFS be the next default filesystem?
Only if Sun relicense ZFS under GPL. The CDDL license it is currently under is incompatible with the Linux kernel.

[gnomeuser]

Only if Sun relicense ZFS under GPL. The CDDL license it is currently under is incompatible with the Linux kernel.

Actually it's more tricky than so.. if they GPLv3 ZFS we still can't use it as you cannot mix GPLv3 and GPLv2 code in the kernel (since the kernel doesn't enable the version 2 or later clause).

It's very unlikely that we'll see ZFS on Linux outside of the FUSE port.

What I like the most about ZFS is the easy of use they designed into it, 2 well defined commands to control all of ZFS, contrast that to if we have to go for the same functionality on Linux, lvm + filesystem + raid + compression + encryption is a staggering amount of commands and they are some what poorly documented and thought out. We could wrap all that functionality in a similar way I guess but nobody has been interested in doing the work.

[gnomeuser]

This sounds ZFS-like (http://www.ussg.iu.edu/hypermail/linux/kernel/0706.1/1732.html). Maybe we'll get some of the functionality at least.

[einalex]

We will get ext4 withour doing much about it anyway since everybody will take and test it.

I can't really see points not to include reiser4.
- There are devs, Patches are coming in.
- Including one doesn't mean not including the other
- More testing leads to more patches

Hans Reiser or other filesystems like ext4 and ZFS have absolutely nothing to do with the question to include it or not. And I really don't see any reason why it should not be included. It may damage the users files? Light up big sign "This will eat your files." Upon installation.

Sorry for the rant. I just hate to see a good piece of software, we all could profit from, getting flushed down the toilet.

[jcpunk]

I also vote for reiser4. If you ignore awesome, but weird features like the file is a file and directory thing, there is one big bonus. On the fly inode creation. I did some sizing and formating tests on a 1Tb drive and inodes really are starting to weigh in here....

http://uhacc.org/forums/index.php?board=2;action=display;threadid=3148;sta rt=0;boardseen=1

with ext4 doing static inode allocation and zfs being an incompatible license, reiser4 makes the user experience more "Windows like"

The average Windows user has no idea what an inode is or what it is for, but they will declare Linux broken if they cannot save more files with 30% of the disk free.