News Linux Founder Linus Torvalds Draws Ire for Criticizing Oracle ZFS

When you see the mess Oracle did of Java, OpenOffice and MySQL, Torvalds has a point - a big one, too. Would you compromise your lifetime's project to include something not even 0.1% of your user base wants when it has a high probability of screwing up the whole project?
Linux finally beat Windows (see how many Android-based devices are sold compared with Windows based ones), it would be foolish to get in trouble with Oracle too.
 

bigdragon

Distinguished
Oct 19, 2011
1,111
553
20,160
I find it strange that people would be upset with what Linus said. Oracle has earned its reputation. Linux developers should not go out of their way to accommodate Oracle's restricted contributions and products. Oracle absolutely would screw over the Linux community the moment it thinks money can be made.
 
  • Like
Reactions: bit_user

unis_torvalds

Reputable
Jan 16, 2020
14
6
4,515
There seems to be some confusion in these comments here.

OpenZFS (the only ZFS the Linux community is really talking about) is 100% free and open-source.
It was forked from Sun Microsystems' CDDL-licensed project before Sun was acquired by Oracle, and it has been actively developed ever since as its own separate thing. CDDL (Common Development and Distribution License) is a free and open-source software license.
Canonical (maintainers of Ubuntu) see fit to include CDDL OpenZFS kernel modules with their distro.

Basically, Linus didn't know what he was talking about.

OpenZFS is the best filesystem available today for those who care about data redundancy/stability and smart caching. It's worth a look!
 

CerianK

Distinguished
Nov 7, 2008
260
50
18,870
Oracle has earned its reputation.
I don't have a problem with Oracle one way or the other in that sense (and am certainly not an expert on the subject).
My issue is with how their technology is wrapped for use in our business environment. We have to scrap large portions of what we have to accommodate changes in the Windows/Java ecosystem it is used in to remove more of the Java forms (as I understand it), which require more additional training for hundreds of people. Granted it had done ok for 10 years (with everyone expected to remember the problem areas so they could avoid them), except for the problem with inadequate handling of large external data which required additional OnBase integration with time-restricted seat licenses in order to bring costs down.

That being said, I like Oracle Outside In Technology and use it routinely for viewing all kinds of odd file types I receive.

Prior to Oracle, we used a composite of custom mainframe and Microsoft Access Forms (we developed in-house) to handle corporate operations... with some of the features we were used to not implementable in Oracle (either for cost and/or technology reasons).
 
  • Like
Reactions: bit_user
There seems to be some confusion in these comments here.

OpenZFS (the only ZFS the Linux community is really talking about) is 100% free and open-source.
It was forked from Sun Microsystems' CDDL-licensed project before Sun was acquired by Oracle, and it has been actively developed ever since as its own separate thing. CDDL (Common Development and Distribution License) is a free and open-source software license.
Canonical (maintainers of Ubuntu) see fit to include CDDL OpenZFS kernel modules with their distro.

Basically, Linus didn't know what he was talking about.

OpenZFS is the best filesystem available today for those who care about data redundancy/stability and smart caching. It's worth a look!
Yeah, right - considering Dalvik is a reimplementation of Java (not even a fork, but a clean room re-engineering) and Oracle STILL sued Google for copyright infringement on API CALLS, how long would they wait before they'd sue any and all major company distributing a kernel that includes some of their copyright, nevermind the license?
Ubuntu, being South Africa based and generating not much revenue in the US would not bring much in. Imagine Google including ZFS by default in Android, count the SECONDS before Oracle sues them again.
 
  • Like
Reactions: bigdragon

unis_torvalds

Reputable
Jan 16, 2020
14
6
4,515
Yeah, right - considering Dalvik is a reimplementation of Java (not even a fork, but a clean room re-engineering) and Oracle STILL sued Google for copyright infringement on API CALLS, how long would they wait before they'd sue any and all major company distributing a kernel that includes some of their copyright, nevermind the license?
Ubuntu, being South Africa based and generating not much revenue in the US would not bring much in. Imagine Google including ZFS by default in Android, count the SECONDS before Oracle sues them again.
You're probably right, only this would never happen as Google compiles their own kernel for Android and there's no benefit in ZFS for smartphones. Further, by your same logic I can't imagine Linux Foundation being a juicy target for lawsuits as they are a non-profit and thus even less lucrative than Canonical.
Anyways, I'm not defending Oracle here. Larry Ellison could definitely stand to chill out; this whole Java debacle is a disaster no doubt.
And Linus is allowed to do whatever he wants regarding kernel-space signals etc. It's just that he was out of his depth when he called ZFS a "buzzword" with poor performance and no maintenance that nobody should use. These comments were demonstrably and surprisingly uninformed.
 

bit_user

Polypheme
Ambassador
OpenZFS is the best filesystem available today for those who care about data redundancy/stability and smart caching. It's worth a look!
Years ago, when setting up a file server, I looked at ZFS and BTRFS. I made the decision that any advantages ZFS had were not worth the trouble, and I've been happily using BTRFS ever since.

I love subvolumes, snapshots, and being able to scrub my filesystem on non-RAID devices.
 

bit_user

Polypheme
Ambassador
Ubuntu, being South Africa based and generating not much revenue in the US would not bring much in.
Conical is based in the UK:

Headquarters: London, United Kingdom

However, the name is derived from a southern African word:

Ubuntu (Zulu pronunciation: [ùɓúntʼù])[1] is a Nguni Bantu term meaning "humanity." It is often translated as "I am because we are," or "humanity towards others," but is often used in a more philosophical sense to mean "the belief in a universal bond of sharing that connects all humanity."
 
Last edited:
  • Like
Reactions: unis_torvalds

bit_user

Polypheme
Ambassador
this would never happen as Google compiles their own kernel for Android and there's no benefit in ZFS for smartphones.
Google doesn't only use Linux on smartphones. They also use it on their servers, and probably provide it for customers renting VMs on their public cloud. In some of those scenarios, ZFS could make sense.
 
Last edited:
Google doesn't only use Linux on smartphones. They also use it on their servers, and probably provide it for customers renting VMs on their public cloud. In some of those scenarios, ZFS could make sense.
Except ext4 is enough for their use, and btrfs is there for the rest. ZFS is interesting, yes, but it's nothing that isn't available in other filter systems like XFS or F2FS.
 

bit_user

Polypheme
Ambassador
Except ext4 is enough for their use, and btrfs is there for the rest. ZFS is interesting, yes, but it's nothing that isn't available in other filter systems like XFS or F2FS.
Yeah, I honestly have no clue what they use. Could be their own FS, for all I know (I heard youtube uses/used its own custom FS).

Anyway, there's one feature of ZFS that I still wouldn't try on BTRFS - RAID. It was one of the last big features to be implemented, and was rather unstable, for a long time. It might be fine, but on a RAID I just built, I still used BTRFS on top of a md RAID device - separating out the RAID part from the filesystem layer.

If I were using a hardware RAID controller, there'd be no choice - you'd have to do them separately. However, except for large arrays or NVMe SSDs, I think hardware RAID no longer makes much sense.
 
Yeah, I honestly have no clue what they use. Could be their own FS, for all I know (I heard youtube uses/used its own custom FS).

Anyway, there's one feature of ZFS that I still wouldn't try on BTRFS - RAID. It was one of the last big features to be implemented, and was rather unstable, for a long time. It might be fine, but on a RAID I just built, I still used BTRFS on top of a md RAID device - separating out the RAID part from the filesystem layer.

If I were using a hardware RAID controller, there'd be no choice - you'd have to do them separately. However, except for large arrays or NVMe SSDs, I think hardware RAID no longer makes much sense.
Considering that they were the ones reporting that automatic conversion from ext2 to ext4 worked perfectly on all their servers, I would say they are still using ext4 at least on some machines now.
 

bit_user

Polypheme
Ambassador
Considering that they were the ones reporting that automatic conversion from ext2 to ext4 worked perfectly on all their servers, I would say they are still using ext4 at least on some machines now.
Uh, when was that... like 10 years ago? They were probably also still using COTS hardware, back then.

Anyway, we don't need to get into a whole debate about what FS Google is using, especially in the absence of any good or current info. I guess you could easily rule out any form of ZFS, if their people are not active contributors. Actually, looking at commit logs would probably be the best evidence one could get of what the are using, assuming it's not entirely company-internal.
 
Last edited:
D

Deleted member 14196

Guest
Not open source ...its gotta go!
He’s merely warning developers not to use it until the licensing model changes so that it’s legal to use. What’s wrong with that?

Oracle can stuff it. I hate working with any of their stuff.
 
  • Like
Reactions: bit_user
Uh, when was that... like 10 years ago? They were probably also still using COTS hardware, back then.

Anyway, we don't need to get into a whole debate about what FS Google is using, especially in the absence of any good or current info. I guess you could easily rule out any form of ZFS, if their people are not active contributors. Actually, looking at commit logs would probably be the best evidence one could get of what the are using, assuming it's not entirely company-internal.
At the very least, it's still the filesystem of choice used on Android devices and, AFAIK, on Chromebooks.
Now, it's strue they could be using their own FS - but I have a hunch that the kind of data storage and manipulation Google does is so far above the usual single machine disk storage system that it's been abstracted to heck. And as such, fast and easy like ext4 could very well still be their no1 choice.
 
  • Like
Reactions: bit_user

bit_user

Polypheme
Ambassador
"tomayto, tomahto" ....
You might not care about the difference, but here's how Linus characterized it:
there is no way I can merge any of the ZFS efforts until I get an official letter from Oracle that is signed by their main legal counsel or preferably by Larry Ellison himself that says that yes, it's ok to do so and treat the end result as GPL'd.

As for why he cares about it being GPL'd, he gives two reasons:
it has all the questions about what happens if Oracle ever decides - again - that "copyright" means something different than anybody else thinks it means.

The subtext behind that is that Oracle sued Google for violating the copyright on its Java APIs, even though Google created a separate implementation. This was a novel interpretation of copyright, at the time.

The second reason characterizes a particular view of GPL:
the whole point of the GPL is that you're being "paid" in terms of tit-for-tat: we give source code to you for free, but we want source code improvements back. If you don't do that but instead say "I think this is legal, but I'm not going to help you" you certainly don't get any help from us.

And, with all this being said, he's not even trying to ban people from using ZFS - he's just not offering any assistance or accepting it for inclusion into the kernel source tree.

So, you can argue that you don't care about the distinction between the licenses, but you can't argue that one doesn't exist or is irrelevant.