Ubuntu will be adding ZFS as standard

[protocelt]

According to this mailing list post, Ubuntu will be integrating ZFS in the near future. I have wonder if Canonical is losing business to other operating systems with ZFS like SmartOS, Solaris, and of course FreeBSD.

I was also under the impression that ZFS couldn't legally be integrated into the Linux kernel due to the incompatibility between the CDDL and GPL licenses. This possibly isn't the case.

 

[Beastie7]

near future

It's going to be a loooooong near future. They can't even finish Btrfs, let alone integrate something extremely complex as ZFS into Linux.

 

[kpa]

The apparent license incompatibility is easily solved by not importing the CDDL licensed code to the kernel source tree but building the kernel module(s) implementing ZFS separately from a separately maintained codebase. GPLv2 as used with the Linux kernel poses no restrictions on third party loadable kernel modules, they can contain what anyone wants to put there and be from any source imaginable.

 

[NewGuy]

This was sort of a given since Debian (Ubuntu's upstream) announced they plan to include ZFS support in the near future. (This was mentioned back in April 2015.) Ubuntu isn't pioneering this, they might not even have to do any work on it, when ZFS support lands in Debian, Ubuntu will just inherit it.

This really just makes the support official. ZFS has worked fine in both Debian and Ubuntu for years now.

 

[Oko]

ZFS works fine on RHEL as well and you would be surprised how many Linux people swear by it.

 

[protocelt]

It's going to be a loooooong near future. They can't even finish Btrfs, let alone integrate something extremely complex as ZFS into Linux.

Not so sure about that. Although if there are so many Linux users/admins pushing for ZFS support, and it seems like there are, then why are they continuing to push Btrfs instead in a lot of the major Linux distributions? Last I read, and it may be FUD as I didn't corroborate anything, Btrfs had some major shortcomings in design that most likely won't be fixed. I don't remember what those shortcomings were.

Personally I would be happy to see better ZFS integration with Linux. Many users/admins use multiple operating systems and it just makes interoperability that much better between them. It's too bad Apple dropped ZFS. That could have been nice for many as well.

 

[Beastie7]

Not so sure about that. Although if there are so many Linux users/admins pushing for ZFS support, and it seems like there are, then why are they continuing to push Btrfs instead in a lot of the major Linux distributions? Last I read, and it may be FUD as I didn't corroborate anything, Btrfs had some major shortcomings in design that most likely won't be fixed. I don't remember what those shortcomings were.

Personally I would be happy to see better ZFS integration with Linux. Many users/admins use multiple operating systems and it just makes interoperability that much better between them. It's too bad Apple dropped ZFS. That could have been nice for many as well.

I'd wager it's more practical for them to create their own implementation of ZFS that's well integrated into the kernel, and of course, licensing issues. I know Linus would never accept anything other than the GPL in Linux, which raises future inclusion issues that could otherwise be smoother. My confusion is the whole nuances you have to go through just to get everything (including all supported feature flags, all SPA/DMU work) working. With Deb is it in kernel or the FUSE implementation? I did quick skim of their (Ubu/Deb) documentation and my god is it cumbersome. And recompiling? Ew. Getting all of this working smoothly without upstream, mainline support is going to be a tough task; albiet as a kernel module. But who knows. If they can manage, that'll be surprising.

 

[gofer_touch]

ZFS gains more support, that is a good thing! I wonder though why do some prominent names, notably Henning Brauer of OpenBSD in some of his latest talks, continues to say that the license is going to come and bite? His position is that it is a risk to use ZFS. Are there any real merits to this claim?

 

[Oko]

ZFS gains more support, that is a good thing! I wonder though why do some prominent names, notably Henning Brauer of OpenBSD in some of his latest talks, continues to say that the license is going to come and bite? His position is that it is a risk to use ZFS. Are there any real merits to this claim?

For uninitiated gofer_touch is referring to Henning Brauer's "OpenBSD sucks" talk. The talk is of course about OpenBSD not about ZFS nor about FreeBSD for that matter. ZFS is mentioned in the context of the fact that OpenBSD doesn't have a decent modern file system. Historically OpenBSD was always a first rate network OS and a second sometimes third rate storage OS.

The licensing discussion between Free and Open BSD folks is largely the conversation of deaf parties. OpenBSD continues to strive in-spite "open hardware, open source, free license" only policy while FreeBSD have adopted more "pragmatical" approach where even proprietary binary blobs are OK in the kernel (infamous NVidia video drivers come to mind).

One should see ZFS discussion in that light. FreeBSD is at this point so heavily vested in ZFS that FreeBSD at this point is just (or I would say mostly) storage OS where ZFS is a first class citizen. CDDL is a big problem and any legal move by Oracle, which owns ZFS regardless of what folks from OpenZFS think about it, will cause almost instantaneous collapse of the entire FreeBSD project. I have no idea what are historical reasons for such situation but the lack of innovation and lack of ownership of new interesting technologies are probably part of the story (sorry guys but ZFS was just ported to FreeBSD which have nothing like HAMMER, vkernel, OpenBGPD, OpenIKED, LibreSSL or OpenSSH). Licensing issue aside ZFS meant reimplementation of large parts of Solaris kernel in FreeBSD and many people fell very uneasy about such complex file system even if it was free.

DragonFly guys are of course not making situation easier and even a alpha release of BSD licensed, light, open source modern file system HAMMER2 might render FreeBSD irrelevant just like mighty NetBSD become a hobby project 18th of October 1995.

 

[kpa]

CDDL is a big problem and any legal move by Oracle, which owns ZFS regardless of what folks from OpenZFS think about it, will cause almost instantaneous collapse of the entire FreeBSD project.

Can you elaborate on this? As far as I can see Oracle can do nothing to change the licensing terms of the code that originally released under the CDDL by Sun. As long as you comply with the terms of the license you can not be taken to court for any kind of licensing violation.

 

[gofer_touch]

Can you elaborate on this? As far as I can see Oracle can do nothing to change the licensing terms of the code that originally released under the CDDL by Sun. As long as you comply with the terms of the license you can not be taken to court for any kind of licensing violation.

I too would be very interested in an elaboration on this. The only thing I've found concerned the Oracle NetApp patent lawsuit that was dismissed

http://www.eweek.com/c/a/Data-Storage/NetApp-Oracle-Settle-Old-Patent-Litigation-Over-ZFS-121130

 

[ANOKNUSA]

This was sort of a given since Debian (Ubuntu's upstream) announced they plan to include ZFS support in the near future. (This was mentioned back in April 2015.) Ubuntu isn't pioneering this, they might not even have to do any work on it, when ZFS support lands in Debian, Ubuntu will just inherit it.

This really just makes the support official. ZFS has worked fine in both Debian and Ubuntu for years now.

This could just be a quirk of perception--it didn't become apparent to me until after I'd started using FreeBSD full-time--but the official or semi-official adoption of ZFS by various Linux distros and adoption by Linux users seems to be moving more rapidly these days than the development/adoption of BTRFS. There appears to be a fair amount of anxiety and misunderstanding remaining over BTRFS: I still occasionally see someone state that they refuse to use BTRFS until it gets fsck support, for example.

Again, that could just be an illusion, but it wouldn't surprise me to see that happen once ZFS gained more exposure in the Linux world. I used BTRFS for a year or so and found some things about it impressive, but I didn't have to use ZFS for long to see just how far BTRFS has yet to go, and unlike the latter ZFS has a pretty well-established reputation at this point.

ZFS works fine on RHEL as well and you would be surprised how many Linux people swear by it.

The folks over at the Funtoo project seem pretty taken with ZFS. A couple years ago when I was playing with it, their wiki had only a few articles; the only article other than the installation guide that dealt with disks/filesystems was about installing and maintaining a system on ZFS. They didn't even have anything on BTRFS or comparing/contrasting other filesystems.

 

[Oko]

Can you elaborate on this? As far as I can see Oracle can do nothing to change the licensing terms of the code that originally released under the CDDL by Sun. As long as you comply with the terms of the license you can not be taken to court for any kind of licensing violation.

I am trained as a research mathematician not a lawyer. That question should have been carefully vetoed by lawyers working for FreeBSD foundation before porting ZFS to FreeBSD. The fact that Apple gave up on porting ZFS as soon as Oracle acquired Sun tells me (and I do know many very competent engineers and few good corporate lawyers who work for Apple) that FreeBSD foundation didn't do its homework.

Personally I am more concern about the part of Henning Brauer's talk where he claims that even small users of ZFS like myself will be eventually bitten for their rear end. I don't use ZFS at home and I have no plan ever to use it. However I do have over 300TB of data at work sitting on ZFS pools and that makes me very jittery.

My personal opinion of BTRFS is that it is just another HURD. After 20 years of development the best RHEL has on 7.1 is XFS which I used in mid nineties on IRIX/SGI. Linux has a proven track record of vapourware technologies and that is where I see BTRFS.

 

[gofer_touch]

The last time I checked there were no fewer than 26 companies that have built OpenZFS into their core products, including the wonderful folks over at iXsystems http://open-zfs.org/wiki/Companies

I am not knocking the above post, because it is a valid point, but have all of these companies not done their homework either?

Surely by now someone would have raised some kind of alarm, apart from Henning Brauer. Everyone else seems to be completely silent. One positive sign is that the original coders of ZFS under Sun are contributing to the OpenZFS effort. Are they oblivious to possible licensing dilemmas?

Perhaps the folks at OpenBSD had more practical reasons for not wanting to port ZFS into their tree... like it being a gigaton of extra stuff to audit for security holes and it being such a complex filesystem/volume manager in one? NetBSD is working on a port (I guess it is beneficial to have ZFS on the Sega Dreamcast), and the Dragonfly project initially thought of porting it but then Matt Dillon decided against it in favor of coding their own home-grown file system.

Also, while were on the subject of the CDDL license, isn't Dtrace in the same boat?

 

[protocelt]

Making the claim that the FreeBSD project could collapse because of legal issues with ZFS is pure FUD. All the BSD projects have different goals and philosophies, and as such, differences of opinion and bias exist between them. Stating an opinion is fine as long as it's made clear that's it's an opinion, otherwise, be prepared to offer references to facts that back up that claim.

Historically and even now Oracle has been extremely aggressive at protecting any and all of it's intellectual property. Many companies, such as Joyent for example, base their business partially or entirely around OpenZFS. These companies directly compete with Oracle in some markets. If there was any chance Oracle could wipe them out by suing over ZFS, they most likely would have by now and further, it's unlikely ever to happen in the future as long as the CDDL license for the original ZFS code or any CDDL licensed code is followed.

 

[Oko]

Perhaps the folks at OpenBSD had more practical reasons for not wanting to port ZFS into their tree...like it being a gigaton of extra stuff to audit for security holes and it being such a complex filesystem/volume manager in one? NetBSD is working on a port (I guess it is beneficial to have ZFS on the Sega Dreamcast)

You are 100% on-money when with respect to OpenBSD folks motivation. ZFS on NetBSD is vaporware IMHO. The core group never really wanted it.

 

[Oko]

Making the claim that the FreeBSD project could collapse because of legal issues with ZFS is pure FUD. All the BSD projects have different goals and philosophies, and as such, differences of opinion and bias exist between them. Stating an opinion is fine as long as it's made clear that's it's an opinion, otherwise, be prepared to offer references to facts that back up that claim.

I was hopping that it was clear from my posts that they are my personal opinion loosely based on some circumstantial facts. As such they should not be taken too seriously.

 

[protocelt]

Not a problem Oko. I just wanted to make sure there was no confusion in licensing or issues for anyone stumbling across this thread in the future when considering FreeBSD.

The point of this thread was simply to highlight the fact that ZFS, a first class citizen on FreeBSD, is so well regarded, Ubuntu has decided to include it by default. The FreeBSD project had obviously made the right choice in porting it.

 

[phoenix]

I am trained as a research mathematician not a lawyer. That question should have been carefully vetoed by lawyers working for FreeBSD foundation before porting ZFS to FreeBSD. The fact that Apple gave up on porting ZFS as soon as Oracle acquired Sun tells me (and I do know many very competent engineers and few good corporate layers who work for Apple) that FreeBSD foundation didn't do its homework.

How do you know this wasn't done? Have you discussed your concerns with anyone at the Foundation? There's a lot of work done behind the scenes by the Foundation for things like this. This smells greatly of uninformed FUD.

 

[Beastie7]

Not to mention differences in ZFS implementation that would probably nullify any potential lawsuit.

 

[NewGuy]

You are 100% on-money when with respect to OpenBSD folks motivation. ZFS on NetBSD is vaporware IMHO. The core group never really wanted it.

Vaporware implies that the feature has not been implemented and probably won't be. But ZFS is included and works well on NetBSD 7.0. I played around with it and the implementation of ZFS appears to be about on par with FreeBSD and Linux.

 

[gpatrick]

NetBSD 7.0 has ZFS albeit v23, but it works.

I also take issue with NetBSD being described as a hobby project since 1995. This is pure FUD.

And if ZFS had legal problems then Joyent wouldn't base SmartOS on it, and OmniTI wouldn't have OmniOS.

 

[beastDemian]

which owns ZFS regardless of what folks from OpenZFS think about it, will cause almost instantaneous collapse of the entire FreeBSD project

Could you clarify how you got to that conclusion? Oracle can't change the ZFS license anymore than the FSF can't change the license on old GCC releases (which is why they had to release new versions of GCC/GDB related utilities when they upgraded to GPL3). What is the legal peril you see FreeBSD developers facing should Oracle decide to damage them? In any case, we still have UFS, should Oracle decide to suddenly start acting like Jerks towards the Project for no reason.

 

[Oko]

Could you clarify how you got to that conclusion? Oracle can't change the ZFS license anymore than the FSF can't change the license on old GCC releases (which is why they had to release new versions of GCC/GDB related utilities when they upgraded to GPL3). What is the legal peril you see FreeBSD developers facing should Oracle decide to damage them? In any case, we still have UFS, should Oracle decide to suddenly start acting like Jerks towards the Project for no reason.

I am a research mathematician so I have no formal education in that area. I have lived half of my life in the Roman law legal system and half of my life in the country which uses common law legal system (U.S.). I am not formally educated in the matter by I have seen what was going around me for the past twenty plus years. Now I also know how to read and I am sure you do. So lets read Sun Public License which is the only License Oracle is accepting when it comes to ZFS, Solaris and similar

http://opensource.org/licenses/SPL-1.0

feel free to compare to CDDL-1.0 used by FreeBSD foundation

https://opensource.org/licenses/CDDL-1.0

Little Wikipedia background is helpful

https://en.wikipedia.org/wiki/Common_Development_and_Distribution_License

Please let me know if you as a private citizen without law degree find any of the things in SPL-1.0 scary. We can then compare our notes.

 

[ronaldlees]

Personally I am more concern about the part of Henning Brauer's talk where he claims that even small users of ZFS like myself will be eventually bitten for their rear end...

LOL: But in Europe it would be a schnauzer rather than the pit bull? Excuse my ignorance.

I've always been frustrated by Debian's strong adherence to GPL. It's sometimes very maddening, thus I'm surprised that they'd do anything like has been suggested.

I agree with the others that Oracle sure likes to keep its IP lawyers busy. How did they come out on Google/JavaAPI lawsuit?