Native port of ZFS for Linux | Hacker News Reader

Native port of ZFS for Linux

github.com

59 pointscrankyadmin31 comments

It makes sense that this surfaces 1 week after reformatting my Ubuntu box to use OpenSolaris and native ZFS.

Also of note:

This ZFS on Linux port was produced at the Lawrence Livermore National Laboratory (LLNL) under Contract No. DE-AC52-07NA27344 (Contract 44) between the U.S. Department of Energy (DOE) and Lawrence Livermore National Security, LLC (LLNS) for the operation of LLNL. It has been approved for release under LLNL-CODE-403049.

I find it interesting that both LLNL and Sun employees are involved in the port. Does Sun being directly involved help the case for it being used in Linux?

Did the ZFS patents ever get sorted out? The US government needn't worry about this, they have an exemption, but the rest of the world should care. Sun has a truckload of patents on ZFS that, last I'd heard, they were not granting free to ZFS users, and NetApp thought they had a patent on it as well, at least enough to sue Sun.

ZFS is released under the CDDL 1.0. The terms of the CDDL (§2.1) grant a "a world-wide, royalty-free, non-exclusive license" for any patent claims "infringed by the making, using or selling of" ZFS now owned or later aqcuired by Sun/Oracle.

Thus, typical ZFS users cannot be sued for infringing on the patents that Oracle holds on ZFS technology. However, this does not preclude third-parties (e.g., NetApp) from pursuing patent claims that they believe are infringed by ZFS -- but this is just as true with any other technology (think MS suing Garmin over FAT in Linux).

Liability for infringement on Oracle's ZFS patents is possible if a user does not comply with the terms of the CDDL. For example, modifying the ZFS source code and distributing it using a license other than the CDDL would be a breach of the license terms (§3.1), and thus forgo any patent license. The Apache 2.0 license has similar terms regarding patent rights.

  (think MS suing Garmin over FAT in Linux)

OT: That was TomTom, not Garmin

If the patents were a big issue I don't think we would have ZFS in FreeBSD.

I don't think the rest of the world cares that much. US software patents don't mean much to anyone outside the country...

Yeah I doubt RIM would agree with you. That is just one example.

I don't get this reference. Links?

http://www.wirelessweek.com/News/2010/05/Policy-and-Industry...

I'd like to know how this could be legally redistributable, given the license conundrum CDDL and GPLv2 make. Any idea?

ZFS is redistributable and Linux is redistributable; you just can't redistribute the combination. So you'll have to download and install ZFS separately.

Thank you. I was looking for some discussion here, but everyone was just spouting off about patents, and it being "illegal", which isn't the case. OpenAFS has a similar license incompatibility, yet it's still very widely used.

Mmmh, interesting. Apparently they did this with production in mind, so it must be usable. I'll give it a try...

Note that this is just a port of the OpenSolaris code base -- not a reimplementation. CDDL copyrights, patents, etc., are still big considerations before this can be used.

Seems bit old, originating from 2008? Any notable progression lately?

It went nowhere. Linux distros and the mainline kernel are using btrfs, which has a similar layer-spanning design as zfs but is btree based like Reiser (which should make it faster).

Most features are done and you can install with btrfs out of the box today in the RHEL 6 or Centos 6 betas.

http://en.wikipedia.org/wiki/Btrfs

ZFS has the following that BTRFS doesn't * Deduplication - Huge in any virtualization environment * Cache layer - Required for high IOPS * fsck - BTRFS cannot fix itself * raid - BTRFS raid currently cannot repair itself and current git of the latest linux kernel still has these isses as well as space issues. I wouldn't trust BTRFS to store /dev/random

Good, because it's an experimental filesystem. Don't write it off completely though, because all of those (except for "cache layer" -- wtf does that even mean?) are planned.

They claim that they're legally free and clear, but a lot of the kernel devs are of the view that kernel modules are derived works of the kernel and therefore have to be GPL.

> therefore have to be GPL.

Nope, they just have to be compatible with the GPL. You could, for example, release a kernel module under the BSD or ISC license and be in the clear. The problem is that the CDDL is intentionally incompatible with the GPL.

Given all the binary drivers available for Linux, those people are clearly toothless.

They may or may not be "derived works" (whatever those are), but they aren't "derivative works", which is what matters as far as copyright law is concerned.

GPL people have a tendency to forget that GPL is a copyright license, and so if someone is doing something that doesn't require copyright permission GPL is irrelevant.

Compliance with a published API doens't mean it's a derived work.

While it obviously can't be distributed with the mainline kernel due to licencing issues, it can be distributed standalone to work with Linux, or any other hypothetical system that provided the same API.

Would love to see ZFS (or at least something better than HFS) on OS X. Anyone has info on that? Anything in the pipeline?

Apple had one point had read/write support for ZFS, but it was pulled because Sun at the time was not able to provide certain guarantees. There was an email by one of the main guys in ZFS that said something to that effect.

As such ZFS on Mac OS X is a dead project.

Some work is still being done outside of Apple: http://code.google.com/p/maczfs/ Maybe it's an undead project.

I think PureDarwin folks are working on this. http://www.puredarwin.org/downloads/xmas

Just to note: The GitHub account user name is Brian Behlendorf.

Not the Apache one... the other one! :D

I tend to avoid complexities of any kind or form. As long as there's Slackware and JFS, Linux will be sane.

Reply on news.ycombinator.com