OpenZFS 2.1.3 bugfix brings compatibility with Linux 5.16

The OpenZFS Project has released version 2.1.3 of what the project calls its "open-source storage platform" for Linux and FreeBSD. The terminology reflects that ZFS is not just a filesystem; it also subsumes the functionality of partitioning and logical volume management. This makes creating and managing what ZFS refers to as …

[Anonymous Coward]

We're absolutely firm on this

SFC (2016): "We do not give up on friendly resolution of a GPL violation easily and are glad Canonical seeks to transparently discuss both sides of this issue in public."

The events of these recent years have tasked me with a difficult question. When does extremism succeed absolutism? I do understand the desire for 'purity', but that easily ends up in flames and death. If you aren't comfortable with being somewhere on the continuum that is real life, you're at a dead end.

[Flocke Kroes]

Re: We're absolutely firm on this

It is not about extremism, purity or absolutism. Firstly it is about "will I get sued" and if the answer to that is a definite no then it is about the copyright holder's intent. CDDL contains a patent license that Oracle can void with a small amount of effort. That alone is enough for me to avoid it. Back when Sun was choosing a license they deliberately chose to base CDDL on a license with GPL incompatibility. I respect their choice.

Over the years it has become clear that different people in Sun had different intentions. As far as I can make out, some people who are not the deciding authority say that mixing CDDL and GPL does not violate CDDL but people who really should know what they are talking about say that mixing CDDL and GPL violates GPL. That by itself is sufficient for me to avoid a mixed work.

There is wiggle room for argument. I do not want to have that argument in court with Oracle's lawyers. Oracle could make things clear by either releasing their code under GPL or stating clearly that their code is not to be mixed with GPL code. My personal interpretation of their decision to keep things vague is they want people to create derivative works without giving permission so that if there is ever enough money involved they can sue.

I do not know if Larry needs a bigger yacht than Jeff but I see no reason to risk contributing to one.

[Anonymous Coward]

Re: We're absolutely firm on this

"contains a patent license that Oracle can void with a small amount of effort."

There's nothing in the CDDL that says Oracle is allowed to retroactively change the license their existing code was granted under.

"Back when Sun was choosing a license they deliberately chose to base CDDL on a license with GPL incompatibility."

They picked a license that would be compatible with existing Solaris kernel licenses, plenty of which were proprietary 3rd party licenses.

Oracle owns some of OpenZFS code, but not all of it. It does not control OpenZFS code. It cannot change the license of OpenZFS. The issue here is with the Linux kernel license.

To make it clear: if OpenZFS were licensed under the GNU GPLv3, the /exact same issue/ would exist.

[Flocke Kroes]

Re: We're absolutely firm on this

The patent licence can be voided without changing the CDDL. For the patent license to be valid you must not change any of the code that implements any of the patents. If a need to change such code does not turn up on its own Oracle can create the need by adding a feature that requires such a change. You then have the choice of voiding the patent license or becoming incompatible. You already have the burden of checking any changes against the patents. Microsoft use pretty much the same trap for anyone brave enough to implement their standards.

[Rich 2]

Re: We're absolutely firm on this

“ If a need to change such code does not turn up on its own Oracle can create the need by adding a feature that requires such a change”

Yes they could. But even if Oracle change the spec of Oracle ZFS, OpenZFS doesn’t have to follow it! And yes, this would make OpenZFS incompatible with oracle’s version but so what? Nobody uses both implementations together on the same drive array, so the incompatibility is irrelevant.

[VoiceOfTruth]

Re: We're absolutely firm on this

-> The patent licence can be voided without changing the CDDL

Change the GPL instead. It will solve a lot of headaches.

[Anonymous Coward]

Re: We're absolutely firm on this

Seriously, who would notice? Do people still use the Oracle ZFS implementation? It's all OpenZFS 2.0 as far as I can see because that is where all the ZFS development has been for the last couple of years, not at Oracle.

Besides, the spec has switched a long time ago (more than a decade?) from version numbers to feature flags to allow various ZFS implementations to graciously deal with bits of the spec they don't (yet) understand.

[Anonymous Coward]

Re: We're absolutely firm on this

> They picked a license that would be compatible with existing Solaris kernel licenses, plenty of which were proprietary 3rd party licenses.

No, Sun deliberately picked a license that was incompatible with GPL because the rabid frothing-at-the-mouth anti-Linux people in the Solaris Group - and their junior groupies - threw a hissy shitfit and threatened to resign if Solaris was ever open sourced under GPLv2 or a GPLv2 compatible license.

I know all this because I was there.

Stop spewing retroactive lies. The only part that is partly true is that there were indeed some drivers that could not be open sourced.

However, releasing OpenSolaris under GPLv2 would have allowed the port and use of the corresponding Linux drivers in OpenSolaris. That did not happen, and it was one of the major reasons OpenSolaris was a complete failure.

Fast-forward 15 years: Solaris is dead and Linux is not.

[Anonymous Coward]

Re: We're absolutely firm on this

You were there too? Ah, you were that guy in the back playing snake on his Nokia, and not paying much attention to anything, right?

If you had listened, you'd have heard that the OpenSolaris code was to be kept as the basis to Solaris as a commercial product, and as you so graciously agreed to admit, there was code in Solaris that had to be kept for customers and that did not belong to Sun, and could not be relicensed (though I do wonder how it can be only "partly true" some drivers couldn't be open-sourced - it can be true or not true, but partly so?)

I've met and worked with enough people working on Solaris to be reasonably sure that frothing wasn't much fashionable among them. Also, you seem to be forgetting that Sun was also contributing a lot of code that went straight into both Solaris and Linux, for example in GNOME and X.org.

So the rabidness must not have been that bad, right? Or that Linux hating thing, they were really doing it wrong.

I peg the death of Solaris not on licensing issues, but on the crap UltraSPARC hardware that Sun was pushing at exorbitant prices ever since the dotcom bubble. It came to a point that even a cheap PC was faster than their entry-level servers. The x86 hardware was decent, but their sales people really hated pushing it I suppose that at lower prices, it brought them lower bonuses.

[VoiceOfTruth]

Re: We're absolutely firm on this

-> Also, you seem to be forgetting that Sun was also contributing a lot of code that went straight into both Solaris and Linux, for example in GNOME and X.org.

Let the Linux people never forget about MySQL and OpenOffice - given away by Sun. But oh no, it's nasty Sun when their fantastic file system doesn't have a licence which works with the stupid GPL.

[Gene Cash]

Re: We're absolutely firm on this

nasty Sun

Yes, the same nasty Sun that kept NeWS proprietary and charged a fee for the source.

If you've ever used NeWS you'd know why I wish it would have won over X Windows.

Sun never did anything that didn't directly benefit Sun and was really short-sighted at times, which is why they no longer exist.

[VoiceOfTruth]

Re: We're absolutely firm on this

Boo, nasty Sun, for charging you a fee for something they created. They should give it all away for free and go begging for a few crumbs from the non-existent 'community'.

[Anonymous Coward]

Re: We're absolutely firm on this

> Ah, you were that guy in the back playing snake on his Nokia, and not paying much attention to anything, right?

I've never owned a Nokia phone.

But, now that you've chosen to continue spewing lies, instead of keeping quiet:

https://web.archive.org/web/20110722120048/http://caesar.acc.umu.se/pub/debian-meetings/2006/debconf6/theora-small/2006-05-14/tower/OpenSolaris_Java_and_Debian-Simon_Phipps__Alvaro_Lopez_Ortega.ogg

That is a Wayback Machine link to a video recording about OpenSolaris. The recording is from May 2006. It's some Debian conference, attended by Simon Phipps and Danese Cooper, both from Sun.

First, we have to suffer through about 22+ minutes of Simon Phipps spewing bullshit. He's lying about Sun and their intentions with OpenSolaris. You can't really fault him for doing that, it was his job. He got paid to lie about OpenSolaris and Open Source.

If you're not into listening to that drivel, feel free to fast forward to 27:27 in the video. Simon likes listening to the sound of his own voice.

And then, we have Danese Cooper. Who is Danese Cooper? Follow the Wikipedia link.

At the time, at Sun, Danese was running the entire show about the open sourcing of Solaris and the beginning of the so-called Open Solaris project, and CDDL.

At minute 27:27 in the video, Danese speaks specifically about the CDDL: why CDDL was/is based on the Mozilla License, and why CDDL was chosen for Solaris and OpenSolaris:

Here's the transcript of Danese's own words:

[ ... ] By the way, Mozilla [License] was selected because it is GPL incompatible. That was part of the design when they released Open Solaris.

There were people in that effort that were really hard pushing for just as wide open spaces licenses. Put it under BSD. Even put it under GPL. There were people who wanted to see that happen.

But we couldn't convince the actual engineers who wrote the Solaris Kernel [to release it under GPL].

These guys worked for 15 years on the Solaris Kernel code base and they had some biases about how they wanted their code opened. [ ... ]

biases in that context being the understatement of the year 2006.

Notwithstanding the fact that, whoever wrote code for the Solaris Kernel did not own that code. Sun Microsystems was the IP owner.

I happen to know who those Solaris engineers were. They were the rabid frothing-at-the-mouth GPL-hating crowd I mentioned in my initial post.

[VoiceOfTruth]

Re: We're absolutely firm on this

If I valued my life less I would waste it listening to some of the rubbish that Torvalds spews

[VoiceOfTruth]

Re: We're absolutely firm on this

-> Sun deliberately picked a license that was incompatible with GPL

That is/was their privilege. GPL is a crap licence. We avoid it like the plague it is.

[oiseau]

Re: We're absolutely firm on this

I do not know if Larry needs a bigger yacht ...

Well ...

Knowing what we all know about how Oracle went about dealing with Sun Microsystem's clients (previously free) access to documentation/software updates/BIOS-firmware files/knowledge bases/fora, etc. right after it took over the company (now $$$$ or no access), I'd say that you are quite right in being wary.

O.

[Rich 2]

Re: We're absolutely firm on this

“… they deliberately chose to base CDDL on a license with GPL incompatibility”

The ONLY license that is compatible with GPL is GPL. I really dislike GPL - it’s an arrogant licence (which should come as no surprise considering who thought it up) that causes much more trouble than anything else. The GPL people will say that the intention is to prevent exploitation without giving back. In the meantime, it cripples lots of use cases.

Consider that FreeBSD have no issues at all using ZFS and the BSD licence is free as a bird and the computing world (including Linux) would be much poorer without it. It’s entirely the fault of GPL if Linux is unable to use something like ZFS - it’s nobody else’s fault. Not even Oracle’s.

[Anonymous Coward]

Re: We're absolutely firm on this

> The ONLY license that is compatible with GPL is GPL.

Bullshit.

> I really dislike GPL [ ... ]

Great. Then don't license the software you write under GPL. There are plenty of other licenses.

But I'm guessing you don't write any software. You just talk about licenses.

[Anonymous Coward]

Re: We're absolutely firm on this

And great, oracle didn't license their software under GPL, but yet you and others like you still rant like the self-entitled cultish fanbois you are.

You don't even realise that your responses have just proved Rich2's point with your completely childish post.

[Rich 2]

Re: We're absolutely firm on this

@ST

No - the only licence compatible with GPL is GPL. You can’t even link GPL code with something else without that something else being Borg’d into being GPL. It’s a horrible license that actively prevents many use-cases. Not bad for a licence that pretends to be “free”.

And over the last 40 years, I have written more software than I can shake a stick at, thanks very much.

[Jamie Jones]

Re: We're absolutely firm on this

Well said. 100 times this.

It should be obvious, but needs to be repeated because the GPL cultism is strong. Just look at all the irrational downvotes!

It's like someone throwing a big informal party for anyone, but GPL says they will only go if there is a formal dress code. Whilst everyone has fun inside, GPL sits outside whining that it's all their fault he can't party with them.

The arrogance of having a restrictive license and then complaining at others that the software they write has the wrong license and they must change it... Mind boggling

[Nate Amsden]

Re: We're absolutely firm on this

FreeBSD may have no issues with the license, but for whatever reason(I won't speculate further) it wasn't successful enough in the marketplace to keep OpenZFS as it's reference platform

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

"As of 2019, OpenZFS (on some platforms such as FreeBSD) is gradually being pivoted to be based upon ZFS on Linux, which has developed faster than other variants of OpenZFS and contains new features not yet ported to those other versions"

I run ZFS myself on linux on several systems. I'm not a die hard fan of the system by any stretch, it has it's use cases. Obviously ZFS-based storage systems never really made a dent in the enterprise storage market.

[Anonymous Coward]

Re: We're absolutely firm on this

That is not correct. FreeBSD pivoted AWAY from the version that was downstream from the Illumos ZFS implementation and TO OpenZFS 2.0. And yes, OpenZFS 2.0 includes a part of the codebase of ZFS on Linux and a bunch of other ZFS implementations.

As of FreeBSD 13, OpenZFS 2.0 is the default ZFS implementation on FreeBSD. That is also why all storage appliances that are based on FreeBSD either have already or are in the process of moving to OpenZFS 2.0. It's great for interoperability that basically everyone (bar Oracle, and rumour has it that they have abandoned this space and closed the relevant dev department) now uses a compatible implementation of ZFS. You can debug a Proxmox ZFS drive in FreeBSD, read an Ubuntu ZFS drive in OpnSense, read a TrueNAS ZFS drive in macOS etc. etc.

[Sudosu]

Re: We're absolutely firm on this

Check out Netapp, they are a fairly popular vendor that uses ZFS as their back end on their appliances....which I was completely unaware of until I was called in to help troubleshoot an issue at one of my larger clients.

The visit turned from "yea, now I have to learn yet another storage vendor technology." to "Hey, this is what I run at home for my media server NAS!!!" They all looked at my like I had two heads, but I got them fixed up and running in no time because I knew how it worked in the back end.

I've been running OmniOS Community ZFS for years (very happily I might add), after I switched from NexentaStor to the OmniTI version.

[VoiceOfTruth]

Re: We're absolutely firm on this

-> CDDL contains a patent license that Oracle can void with a small amount of effort.

Why should it?

Since when is the GPL the gold standard of licensing? How about we look at it from a different perspective - GPL is designed NOT to operate with other licences or does so with difficulty.

BSD is a far better licence than GPL.

[Dan 55]

"Red Hat is also working on a new storage manager called Stratis"

Looking forward to the dependencies with systemd, and probably even pulseaudio and gnome, which means yet another terrible piece of Red Hat software is foisted upon the Linux world.

[Anonymous Coward]

ZFS is just too slow to work on modern nvme drives. Until it can figure out how to provide 7GB/sec per drive throughput via zero-copy and RDMA calls, it's useless to me.

[Kevin McMurtrie]

Why would you demand that one filesystem is perfect for all uses? That's not the way anything works. You use the right tools in the right place.

A ridiculous counterpoint would be asking what good 7GB/s NVMe sticks are when they only hold a itty bitty 2TB.

[Anonymous Coward]

> Why would you demand that one filesystem is perfect for all uses?

No-one is asking for perfect. Just usable.

> That's not the way anything works.

Funny how ext4 doesn't have ZFS' performance problems with NVMe. Not even with NVMe RAID. And it doesn't grow a bloated cache pig either.

Another funny thing: I can buy a $300 NVMe RAID controller card that works just great with ext4. On Linux.

ZFS: solving yesterday's problems, tomorrow.

> [ ... ] what good 7GB/s NVMe sticks are when they only hold a itty bitty 2TB

You need to visit Amazon. Search for 'm.2 nvme 8tb'.

[JoeCool]

I have to give OpenZFS Credit

It is one of the most user-friendly, well documeted pieces of foss software I have ever used.

They did a lot of good work to make it useable by people outside of the developer group.