“Today the ZFS on Linux project reached an important milestone with the official 0.6.1 release! Over two years of use by real users has convinced us ZoL is ready for wide scale deployment on everything from desktops to super computers.”
“Today the ZFS on Linux project reached an important milestone with the official 0.6.1 release! Over two years of use by real users has convinced us ZoL is ready for wide scale deployment on everything from desktops to super computers.”
What license?
CDDL+GPL2. Still a mess, still work
There is no doubt that this is an achievement from a technical standpoint, but because of the license mess, I doubt that very many will use it.
I guess that no distro is permitted to distribute a kernel with zfs. I really doubt Oracle will change the lisense ever, I think the cddl was specifically written to be Gpl incompatible.
Sadly I think their effort would have been better spent on making btrfs better.
This is more of a political/religious issue, rather than a real legal dispute. I highly doubt any open-source distro would ever get sued by any GPL license holder due to the CDDL being incompatible in technicalities. But that’s just my opinion.
While Oracle still holds copyright to large portions of the codebase that went into ZoL, they don’t hold all of it any more. Many other companies and individuals have contributed, and they’d all have to agree to relicense.
Depends on who you ask. Some people say it was, some people say it wasn’t.
The amount of work needed to get a complex storage stack up to production quality is mind blowing, so I can’t blame them for not wanting to reinvent the wheel and instead just port something that works. Look at the mess SystemTap is in (http://dtrace.org/blogs/brendan/2011/10/15/using-systemtap/) and those guys have been at it since 2006.
I disagree, especially since Oracle is one very interesting party. There is an unquestionable legal incompatibility between the GPL and the CDDL, not a political one, so Linux with ZFS can’t be distributed, and that is a fact. On the other hand, if I’m not mistaken, nothing prevents you from building Linux with ZFS, or having ZFS as a module, as long as you do not distribute it.
In a nutshell, you cannot legally have a complete binary Linux distro with ZFS out of the box, but you can work around it by building it on the fly.
The incompatibility arises from GPL’s insistence that all derived work must be GPL. The CDDL prohibits removal of the license, hence the deadlock. Please note, that here the “GPL” party is the one that’s wronged, not the CDDL one, so Oracle would have no standing to sue. It’s the rest of the Linux copyright owners that would have to do that.
Now take a guess, what do you think the odds are that some Linux code author is going to sue, say, Debian for distributing ZFS inside of Linux? We’re not talking distributing some closed-source evil baby-eating binary blob (ahem-nVidia-ahem). We’re still talking copyleft open-source, but simply a different one.
That’s what I mean that it’s a political, or rather, a perception issue. If people aren’t suing Ubuntu for shipping the closed-source nVidia driver, they’re not going to sue simply because they dared to ship a piece of code that’s under a different open-source copyleft license.
Look, from where I’m standing, I couldn’t care less. The longer Linux holds off integrating technologies such as ZFS and DTrace, the more users flock to OpenSolaris-derived and FreeBSD systems. Then we get crazy half-functional duplicates (Btrfs, SystemTap) popping up simply because some religious nut couldn’t get it through their thick skull that as long as the license grants them the same freedoms, it really doesn’t matter whether the letters on the file header say “CDDL” or “GPL”.
Oracle inherited the CDDL licencing on ZFS/Dtrace from Solaris when they bought it. Now that they (Oracle) own ZFS they could re-licence it should they want to, instead they keep their ZFS enhancements proprietary and are continuing their efforts on their own GPL licenced BTRFS filesystem.
The reason Solaris licenced ZFS and DTrace under the GPL incompatible CDDL deliberately was because they were losing market share to Linux and obviously didn’t want to ‘give away’ any advantages to their main competitor.
No, again the reason they are incompatible with GPL is because that was a deliberate choice by Solaris when they licenced them as such to prevent Linux from using Solaris’s ‘crown jewels’.
Now, Linux is GPL and thus obviously can’t use CDDL licenced code so it’s not a question of ‘religious nuts’ or NIH.
Oracle created BTRFS, companies like Red Hat, IBM, Intel have worked on SystemTap. They did this because they can’t legally ship Linux with ZFS or DTrace, and therefore won’t, even if some guy named ‘saso’ on OSAlert says that ‘there’s no legal risk’.
ZoL != OpenSolaris ZFS. ZoL uses ZFS from Illumos, which has many more copyright holders than just Oracle. You are encouraged to download the source and do a search for “Copyright (c)” lines, just to see how many parties would need to agree.
Thanks for sharing your interesting, though unsupported hypothesis.
Sun != Solaris. Minor mistake aside, you are merely asserting your claims without evidence. Claims made without evidence can be dismissed without evidence.
Interestingly enough, some guy named Valhalla on OSAlert makes baseless unsupported claims about people’s and companies’ intent and then turns around and criticizes me for expressing my personal opinion (which I even labeled as such). Kettle meet pot.
Yes, but Oracle’s version of ZFS, the one they got when they bought Sun and are now developing in a proprietary fashion is theirs to re-licence if they so wish.
Same goes for your claims.
Of course mine are not taken straight out of the air, Danese Cooper who was apparently the one who wrote CDDL while at Sun stated that it was written to be GPL-incompatible.
http://en.wikipedia.org/wiki/Common_Development_and_Distribution_Li…
Draw your own conclusions, but really given the statement from the person presented as having created the licence aswell as how it makes perfect sense for Sun to not want Linux to snap up their coveted tech there is no doubt in my mind that this is so.
Hardly baseless, read the Wikipedia link and watch the video if it’s still linked.
Your statements that Oracle, Red Hat, Intel, IBM, created/worked on BTRFS and SystemTap out of ‘religious nuttery’ is indeed baseless, as is your statement that no one will be sued for shipping licence infringing code.
Also, when did Ubuntu start shipping proprietary drivers with Ubuntu? Last I checked (which was admittedly a long time ago) they shipped no proprietary binaries, once you had installed Ubuntu and was online you where informed that there were proprietary drivers available for your system and that you could download and install them from a repository should you want to, but they were not shipped on any installation media nor installed by default.
Again, that’s not the version in ZoL. Therefore, Oracle re-licensing their bits isn’t enough.
Nor are mine. Danese Cooper says one thing, some Sun engineers say otherwise. I didn’t take a position in this, I even specifically said: ” Depends on who you ask. Some people say it was, some people say it wasn’t.” Stop tearing down straw men.
It’s funny you should quote that, because apparently you didn’t read the whole section. Read a bit further down on how Simon Phipps disagreed with what Danese said. All this goes to support my statements, that there were mixed feelings on this even within Sun, and to negate your position that it was a clear cut deal.
No doubt in your mind? I can see that – that’s the religious bit I was referring to. You come to a fixed conclusion and find evidence to support it. I, for a change, don’t know – it might have been, it might not.
I don’t mean to say that everybody who works on these are religious nuts, only some are (such as yourself) – I should have qualified that, mea culpa.
Opinion != statement of fact. Read what I wrote again.
Look dude, you seem hell-bent on making this a discussion about he-said-she-said. I’ve got better things to do than argue about with people on the net. Talk is cheap, what matters is code.
Are you saying that Oracle’s proprietary ZFS contains code from the ‘ZFS on Linux’ contributors? Because if not I don’t see what you are arguing.
I watched the entire video ways back, so yes I know what was said.
Simon Phipps presented Danese as the one who ‘created the CDDL’, she then said that when she created the CDDL there were alot of old time engineers which where reluctant to licence it in a GPL compatible manner.
Meanwhile the upper management wanted a copyleft style licence but couldn’t wait for GPLv3 to be finalized, note that GPLv2 was never on the table as it would have allowed Linux to use the code, GPLv3 if it had been finalized would likely have been used as that would not be compatible with Linux given that Linux is licenced under GPLv2 only rather than the usual GPLvX or later.
Later during this video Simon Phipps does indeed say that he ‘disagrees with Danese to some degree’ regarding the licencing, but I believe more in the person who actually drafted the licence (Danese) and spoke directly to the engineers and management while creating it.
Even more so considering that Simon Phipps job was to mend fences with the Free Software community after Schwartz’s attacks on GPL (of course later Sun embraced GPL, go figure), and of course the idea that CDDL was explicitly drafted to be GPL incompatible was not going to help him in his job. He had every reason to deny those (Danese’s) claims, Danese on the other hand had no reason whatsoever to lie that I can think of.
What does making conclusions have to do with religion??
And no, I did not come with a ‘fixed conclusion’, back when I read about all this I had no conclusion, my conclusion is based on the evidence at hand:
1) Sun’s Solaris suffers greatly in the marketplace, with Linux being it’s main competitor.
2) Sun’s management wants a ‘copyleft’ style licence for their code, they are looking at GPLv3 but it is taking too much time to be finalized, GPLv2 is of course available but that’s not even being considered.
Why? Well let’s see, GPLv2 means that Sun code licenced under it can be used directly in Linux, that same competitor which is eating their lunch, GPLv3 licenced Sun code on the other hand would not be possible to include in Linux, but it’s not ready. Tada! Sun creates a new licence, CDDL, which according to the very creator of that licence (Danese) was crafted to be GPL incompatible.
I came to this conclusion based upon the facts before me, where does this ‘religion’ you keep yapping about come in? For me it makes perfect sense for Sun not to want their main competitor Linux to be able to use their ‘technology’. That’s not ‘religion’, it’s common sense.
Describe what you mean by ‘religious nuts’ and how that applies to the choices these companies have made.
Also I’d like to know what in my reasoning earns me the label ‘religious nut’?
I agree, it’s an incredible amount of work, that’s why I hate to see all this effort expended on a project that will never likely be accepted by most of the Linux community.
It would really be fantastic if it could be, but I doubt Oracle will ever change. In fact it would not surprise me if they close sourced zfs like they did with Solaris.
They did, but the cat’s out of the bag now. Open-source ZFS is now being developed here: http://illumos.org/
If it’s open source, we’ll fork it and reduce number of devs in any given project!
http://omnios.omniti.com/
OmniOS is a distro of Illumos. Illumos is the foundation + kernel, OmniOS is a whole OS distribution (which includes Illumos and a ton of other user-space software). The two cooperate, they don’t compete.
No credible distro will ask itself whether it can get away with something without getting sued. The question is whether it’s legally entitled to distribute it, which it isn’t.
I know OS4, http://www.os4online.com, distributes ZFS. Dont know if its in 13.3, I read something on the Facebook page about it being included.
As I read and understand it* a distribution could package it as a kernel module and include ZFS just fine.
There seems to be an effort to package it for Debian when you look at the package development server of Debian.
Debian who are very strict when it comes to licenses already includes a ZFS implementation in the form of Debian/kFreeBSD.
I agree about btrfs, if only because zfs can never be part of the standard kernel.
* the FAQ entry about the license on the ZoL site:
http://zfsonlinux.org/faq.html#WhatAboutTheLicensingIssue
Edited 2013-03-30 16:38 UTC
That link explains the situation pretty much exhaustively: cannot be distributed as kernel built-in, but can be distributed as a separate module or source-code. Thanks for the link.
No problem, but it wasn’t that hard to find it.
Also this does not mean ZoL could not be used as a root filesystem, a normal Linux distribution always uses an initial ramdisk. As far as I know the ramdisk can also be used to load the filesystem module for the rootfilesystem.
Edited 2013-03-30 17:36 UTC
That’s really good question, any lawyer types hazard a guess?
Why would it involve a lawyer ?
Because there is no question that someone could technically, sooner or later figure out how to make this work. The question is whether or not its legally feasible because of the license issues.
If the FAQ entry on their site is correct, I see no reason to involve lawyers:
“The combination of them causes problems because it prevents using pieces of code exclusively available under one license with pieces of code exclusively available under the other in the same binary. In the case of the kernel, this prevents us from distributing ZFS as part of the kernel binary. However, there is nothing in either license that prevents distributing it in the form of a binary module or in the form of source code.”
http://zfsonlinux.org/faq.html#WhatAboutTheLicensingIssue
I think it’s important to be realistic here though. Folks really into Linux and GPL software are not going to ever accept ZFS because of the license. Those folks also will not run debian with a FreeBSD kernel.
It just makes more sense to run FreeBSD or Solaris if you want ZFS. The only use case I can think of is if you do gaming and want to use ZFS for storage. That would require a Linux kernel due to the video driver situation in BSD.
I think you are generalizing the goals of the users and developers of Debian.
Why would you need ZFS for Frozen Bubble or Tux Racer ?
Have used both ZFS and Btrfs, I think I’d rather see ZFS mature on Linux (or use FreeBSD for file serving) than run Btrfs.
But that’s just my opinion.
The real question is when btrfs is ready.
When it is ready it would rather use btrfs, it is better integrated with the rest of Linux.
It’s not a matter of whether it’s ready, it’s a matter of whether it’s any good:
ZFS CLI tools are a dream to use where as Btrfs CLI tools are a pig. ZFS snapshotting is child’s play, rolling back snapshots on Btrfs is just awkward in comparison (particularly if the volume is /). There’s a few other design decisions behind Btrfs that annoy me, but I really don’t want to get into a rant about it tonight
I just really don’t like Btrfs one bit so if I had to run a CoW file system like Btrfs or ZFS then I’d sooner run a non-Linux OS and have ZFS than run Btrfs just for the sake of running Linux.
Thankfully I’m just as at home on FreeBSD and Solaris as Linux, so I don’t feel as chained to the Linux ecosystem as some *nix administrators do.
Chained, is a bit to strong, it just makes life easier if the system you know already does everything else you need it to.
For example where I work we have both Debian and OpenBSD for some firewall tasks.
Except my whole point was about how I’d rather run FreeBSD on a file server because Linux CoW file systems are pretty naff in comparison to ZFS; which is the complete opposite of your point about Linux doing everything.
But even that aside, your comment implies that other *nix’s don’t do everything a sys admin would need them to, which clearly isn’t a fair statement. Most of the time it’s lack of experience on a platform that leads users into thinking that said platform cannot do a specific task.
Quite frankly, I get a little fed up with just how single minded some within the Linux community are. I mean, I do like Linux, it’s a great platform. But it’s not the only great *nix platform. Though I guess this is the same grievances the engineers at Sun had and why they pushed for CDDL to be incompatible with GPL.
Edited 2013-03-30 22:22 UTC
The license wasn’t an engineers choice.
To give you an idea of what the engineers were thinking, here is a video:
https://www.youtube.com/watch?v=-zRN7XLCRhc
The video is kinda funny, but the reality of it is just sad.
What the Sun developers created was awesome, but to be honest I think most people think OpenSolaris, euh Illumos is dead or gonna die so people are afraid to invest their time in it or bet the business on.
But this is a bit of strange discussion to have, I’ve only had these discussions about Windows in the past
Edited 2013-03-30 22:52 UTC
I didn’t say it was the engineers choice. What I said was that the engineers pushed for a GPL incompatible license. Part of the reason CDDL was written in the way it was, was because the idea of Suns work being ripped out of OpenSolaris and pasted directly into Linux really grated with quite a number of Sun engineers. So they pushed for CDDL to be GPL incompatible. Obviously the decision still lay with upper management.
Thanks, but I’m well versed in that already. In fact I’ve ran various different OpenSolaris OSs on production systems (as well as vanilla Solaris).
1stly, I hate people who claim that “open source project x is dead” just because they personally are disinterested in it. So long as a project has a community, then it’s not dead. OpenSolaris still has a community and there’s still spin off projects forking from it (SmartOS being one of the more interesting ones to watch).
In fact I’ve lost count of the number of times people have stated a *nix is dead when it still has a cult following: Plan 9, FreeBSD, OpenBSD, GNU/Hurd. I’ve heard people make dumb claims about every single one of them.
2ndly, plenty of business do invest in OpenSolaris-based solutions: Nexenta being just one example.
I’m not sure I follow. But in any case, my point was that as great as Linux is, it’s not the be-all and end-all of unix-like / POSIX systems. There are times when it pays to use other platforms; for me when I need a file server, ZFS becomes one of those reasons that a non-Linux solution is preferable as I just don’t like Btrfs one bit.
Edited 2013-03-30 23:41 UTC
“What I said was that the engineers pushed for a GPL incompatible license.”
“Thanks, but I’m well versed in that already.”
I don’t think the _engineers_ _pushed_ for that:
https://www.youtube.com/watch?v=-zRN7XLCRhc#t=22m00s
I guess that would depend on who you speak to. Danese Cooper, who worked as an open source evangelist for Sun, has stated that many engineers did push for a GPL incompatible license.
Obviously there’d be other reasons behind such a decision (unlike you, I’m not claiming an either/or argument) and obviously those leading public talks would dismiss the whole “GPL incompatible” argument because it’s simply bad PR. So you’re clearly going to find plenty of sources that cite other reasons for how CDDL turned out the way it had. But those points aren’t mutually exclusive from what I’ve been saying either; the idea of Sun’s code being copied verbatim into Linux wasn’t a favorable amongst Sun engineers. It really wasn’t and thus many engineers did push for CDDL to be GPL incompatible.
Ahh, that is a lot more nuances ‘statement’. Now you are saying certain engineers pushed for GPL incompatible licensing.
I think that would be realistic and very probably correct.
And we’ll never know how large a part of all engineers that was.
Edited 2013-03-31 00:07 UTC
That’s exactly what I said right from the bloody start:
Though I guess this is the same grievances the engineers at Sun had and why they pushed for CDDL to be incompatible with GPL.
Ok, I’ll concede that the statement does read a bit like I’m saying every engineer pushed for that. I could have been a little more clear on that. But I’d have thought the generalisation was implied.
Oh totally. I was never trying to argue the reason for CDDLs licensing, just some of the feelings from some of the engineers. In actual fact, the end result of CDDL is irrelevant to the point I was making (though you could be forgiven for not realising that with the way how that statement was phrased).
Even with generalization, I’m not so sure it’s close or a lot more than half off all the engineers.
And we can’t be sure if that is true, that is why I think people should stop talking about it like it’s the only truth.
Generalisations don’t have to be based on majority case scenarios (and usually aren’t). I certainly wasn’t implying that all or even the majority of engineers felt that way. In fact I deliberately avoided any such estimations as it would be grossly presumptuous of me to guess such a figure.
To be frank, I’m starting to get a little frustrated because my whole argument that spawning this cyclic debate was that some Sun engineers didn’t want to see their code “stolen” by the Linux community. Nothing more. Now I appreciate that I’ve not explained that point as well as I could have, but you’re equally leaping to quite a number of assumptions (I don’t know if it’s because you want to find counter arguments to my arguments or if it is just an explanation and interpretation issue. but either way you keep raising objections to statements I’ve never made – such as you objection to the claim I never made about the quantity of engineers or about whether they had authority to make the final decision).
Except that we do know it’s true because several Sun employees have stated so. What’s often argued is whether those sentiments were what led to the incompatibilities in the final version of CDDL. However my point was never about CDDL but instead about the engineers sentiments (CDDL was only referenced to give context to the anecdote). So the reasoning behind CDDL is completely irrelevant in the context of this discussion.
Edited 2013-03-31 01:06 UTC
Anyway, let’s get back to the original point.
Let me know when the port nova-compute from OpenStack to OpenSolaris based systems is done. Or when someone starts with a Ceph port. Running cluster pairs is starting to look a bit to much like a hassle.
People would love to see these things, still they haven’t happend. 1 or 2 people did work on an OpenStack port to one of the OpenSolaris based system.
A port to FreeBSD which also has ZFS and DTrace would be very welcome too.
Even better (for us) would be if it is a Debian based OS like Debian/kFreeBSD or Nextena OS.
Whatever mate. I’m not getting into an OS flamewar with a Linux fanboy.
This tangent was about ZFS vs Btrfs not all the other prejudices you have against non-Linux platforms.
Edited 2013-03-31 14:08 UTC
My point is, Linux is what Windows was, it is where the applications are being developed.
I know Linux is just trying to implement everything the OpenSolaris-based systems already had for years:
ZFS – btrfs
Zones – LXC, SELinux, SecComp2, usernamespaces, cgroups, etc.
DTrace – System Tap, etc.
That comparison doesn’t really work because we’re not talking about desktop application support and your point is moot regarding servers because UNIX has always had a strong presence in the enterprise market even with Microsoft’s best attempts to capture a monopoly there as well.
That’s slightly contradictory to your earlier statement because now you’re admitting that UNIX already has a great deal of the applications that are still under development for Linux. So by my logic, it makes more sense to run a platform with mature software than one which is still under beta testing.
I don’t think we’re ever going to agree on this though. I not saying we should never run Linux nor that Solaris / FreeBSD is better than Linux (to argue that one is better than the other is just dumb in my opinion as they’re all solid systems with their own respective strengths and weaknesses). My point was that it’s dumb to pick a substandard file system for a file server just because it’s the only one of that type available for that OS. It makes more sense to use a different OS to enable you to administrate a better file system. So while I have no personal issue with Linux (I run it on a great number of servers and desktops both professionally and personally), I’m not going to limit my options because I’m too lazy to learn another POSIX OS.
The license was not written to be GPL incompatible, this is FUD. The GPL by its nature is incompatible with anything bar public domain and MIT or BSD style licenses.
The CDDL had some other patent provisions (which were good things) which the GPL did not cover, therefore making the two incompatible.
I’ve been using ZFS on Linux for about a year now and it’s been a really good experience. Very stable, low resource usage and the pooling of devices is great. It’s so nice to have flexible pools, snapshots and cloning on a Linux box. I know Btrfs will probably get there someday, but it’s still in heavy development and I’m not ready to trust Btrfs with important data until it’s stable. ZFS fills my needs nicely and permits compatibility between Linux and FreeBSD boxes.
Here is some Linux guys comparing ZFS to BTRFS:
http://rudd-o.com/linux-and-free-software/ways-in-which-zfs-is-bett…
IMO, he has omitted the ability to snapshot and rollback the whole root directory if an upgrade caused problems.
When I boot Solaris, I can choose which snapshot to boot from in GRUB. Each time I do an upgrade, I snapshot. Then I can easily rollback. Or boot into another snapshot to experiment.
This by far, is my most favourite filesystem and I’ve been using it on our servers for 2 years now. We have higher throughput than what we had with Ext3 (haven’t benchmarked against Ext4). Maintenance wise, it’s a blessing – so easy to work with. When used in combination with GlusterFS, I feel I have a solid SAN.
As for licensing, I’d really like to see all distros bundle an easy-to-use build tool for mitigating the licensing issue.
I don’t care about the licensing mess. I have no problems installing Linux and then installing the ZFS module. So far ZoL works very well. I can now finally share my FreeBSD 9.1 ZFS Raid-Z1 data pool with Linux, when I do a dual boot. That’s all that matters to me!
Fantastic job to the ZoL developers. Please keep up the good work!