Ubuntu 16.04 LTS arrives today complete with forbidden ZFS

Canonical will today (April 21st) launch version 16.04 of its Ubuntu Linux distribution, Xenial Xerus, the new long-term-support version of the project. As the name suggests, long-term support versions of Ubuntu get, er, long-term support, a guaranteed five years from today to be precise. The Xenial Xerus will therefore be …

[Anonymous Coward]

@AC - Re: Priorities?

Those who actually use Kerberos and need to correlate events already know about the importance of NTP and how to set it up correctly. No need for systemd folks to nanny us. As for those who don't want to manage systems like it was still 1970, there's Windows for them. What's wrong with it ?

[heasus]

Re: Priorities?

ZFS! Outstanding. This is an incredibly welcome change in behavior and brings a much needed useful fsys to linux; hats off to ubuntu. I might even consider returning to ubuntu - even with systemd.

And, shut the hell up and go back to your cave, rms.

[Ian 55]

Who cares about start up times?

I know OS suppliers like to boast that they've shaved a second off the time, but in the real world uses for Linux, kit is either on 24/7 and gets rebooted once every six months or so with kernel updates or is put into suspended mode.

Either way, the difference between 20 or 30 seconds to boot is irrelevant - even a minute or two would not be 'too long' - and the price of systemd to shave those seconds off is too high.

You want a faster boot? Boot off an SSD. If you need it to be quicker, you're doing something wrong.

[Preston Munchensonton]

Re: Who cares about start up times?

Either way, the difference between 20 or 30 seconds to boot is irrelevant - even a minute or two would not be 'too long' - and the price of systemd to shave those seconds off is too high.

This doesn't just apply to desktops, where that timeframe could be important to users. It also applies to physical servers, which really have never booted fast in my experience with the SAN or SCSI wait times to scan volumes, wait for FLOGI/PLOGI, etc. Just last weekend, a system maintenance had a 15 minute window just for the shutdown and reboot of an ESXi host, and that's without starting any of the guests.

Start up times have accelerated to the point where developers should definitely spend their time solving problems that actually matter, like figuring out how to clean this systemd virus from new Linux distros.

[John Sanders]

Re: Priorities?

Regarding systemd

Please stop saying that the init part comes with ntp or a word processor.

Systemd is made of many parts, they are inter-related but the init process does not have an ntp client built-in, the ntp functionality is provided by a module that you can or can not use depending on your needs.

Please make the effort to learn about systemd and stop spreading FUD.

[AMBxx]

LTS?

5 years support or until the lawyers wade in over licensing.

Not sure if open source fights like this are a sign of approaching maturity or senility.

[AndyS]

Re: LTS?

> 5 years support or until the lawyers wade in over licensing.

What? Is this something I've missed, or can you actually point at anything that resembles what you're blabbering about?

The licence doesn't end after 5 years, Canonical will stop supporting it. Unlike with XP, you're free to set up a company to continue to support it yourself after that time, and sell support in any way you see fit.

[Craigness]

Re: LTS?

RTFA

"...no less an entity than GNU daddy Richard Stallman thinks Canonical is violating the GNU GPL because you can't blend GPL code and non-GPL code. The Software Freedom Conservancy agrees has form funding court actions that test the GPL..."

[AndyS]

Re: LTS?

OK, I misread the OP as saying the legal problems would start after 5 years. He's still spouting nonsense though.

Even if the ZFS issue is found problematic, that won't end support. Worst case scenario I can think of is that they'll stop distributing ZFS with new downloads, or similar.

And as I pointed out, even if Canonical goes under, the product won't die.

Where does anyone get the idea that this could kill the product as it stands? That is pure fantasy.

[Anonymous Coward]

Re: LTS?

> What? Is this something I've missed, or can you actually point at anything that resembles what you're blabbering about?

http://www.theregister.co.uk/2016/02/26/canonical_in_zfsonlinux_gpl_violation_spat/

Or just type "ubuntu zfs licencing" into Google.

Basically:

* it's possible that someone with a copyright interest in the Linux kernel will sue Canonical for violating the terms of the GPL, because they have added a module to the kernel where the module itself is not covered by GPL.

* it's possible that Oracle will sue Canonical over their use of ZFS. The CDDL is more permissive than the GPL, but there are also patent issues. The CDDL says "no patent license is granted for ... the combination of the Original Software with other software or devices"

[Dan 55]

Re: LTS?

Ubuntu doesn't use Oracle's version of ZFS. What's more, Oracle lost interest in maintaining that five years ago.

[Destroy All Monsters]

Re: LTS?

"Oracle lost interest in maintaining that five years ago"

Oracle never lost interest in making moneyextracting money or controlling things though.

[Dan 55]

Re: LTS?

They'd have to argue that Canonical is making them lose money on it in court... A bit difficult I would think.

[Anonymous Coward]

Dan 55 - Re: LTS?

Lost revenue is used as a basis to calculate how much Ubuntu should pay but it is not used to establish that infringement has been committed. Oracle would be just as happy to have Canonical shut down. To make it easier for you to understand, if I steal your property and give it away for free I'm still guilty according to the law.

[Anonymous Coward]

Re: LTS?

Oracle still has the patents though (the ones which it settled with Netapp for an undisclosed sum)

[Roland6]

Re: or until the lawyers wade in over licensing.

This is perhaps going to be the most interesting development, who is going to send in the lawyers and against whom? and will this actually be in both the best interests of either the open source community and its customers/users.

Perhaps someone should be telling Stallman to add his voice to those calling for Oracle, as the majority licence holder in ZFS to simply make a decree changing the licence status of ZFS - just as the University of California did over BSD, and help the open source community resolve the incompatibilities between it's legacy licences.

[TVU]

Re: LTS?

"5 years support or until the lawyers wade in over licensing.

Not sure if open source fights like this are a sign of approaching maturity or senility."

I think that concerns over the implications of the Common Development and Distribution License that applies to OpenZFS are valid. That said, three of the four informed legal opinions I have seen so far support Canonical's position with the only dissenting opinion coming from the Software Freedom Conservancy (no surprise there then). I also strongly suspect that behind the scenes Canonical will have explored the legal position in depth so that they are in a strong and defensible position.

[Anonymous Coward]

Btrfs is a good alternative to zfs and works well on Ubuntu 14.04.

Not sure if btrfs is illegal though, I assume not.

[Anonymice]

Btrfs is *not* a stable filesystem & shouldn't be used for critical data. ZFS has a big history, albeit on other arquitectures. I've not seen any good data on the Linux port.

[Anonymous Coward]

You are just reading outdated docs. It is stable according to the main project and latest versions.

https://btrfs.wiki.kernel.org/index.php/Main_Page#Stability_status

This is where the Internet really sucks for factual information. Most blogs repeat the comments that they read on a previous blog about stability. These are just old 2007 to 2014.

Check you have new enough kernels and packages.

[Androgynous Cupboard]

Does btrfs still decide the disk if full when it's not?. And then continue to maintain it's full after you delete some files?

Honestly, having tried both btrfs (and run into this problem) and zfs, there is just no comparison. There's a reason people rave about it.

[Bronek Kozicki]

There are some large deployments of ZFS on Linux, why else do you think Lawrence Livermore National Laboratory is (indirectly) funding it? Other examples include NASA and CERN

[Adam Inistrator]

"Btrfs is *not* a stable filesystem & shouldn't be used for critical data."

Spouting nonsense I guess. You are way out of date. It is the DEFAULT fs on SUSE and well supported on Debian/Ubuntu just for the lulz?

[Steve 53]

Re: New???

The best advise a friend gave me when looking at BRTFS stability was to look at the mailing list, and see how many puppies it was killing.

While it may be the default for boot disks, boot disks are rarely multidisk raid.

There are a number of shortcomings with RAID5 style configuration (Write holes, poor re-balancing, etc) which made me feel decidedly uncomfortable trusting it with my data.

Which is a shame, as from a convenience point of view, being able to simply add disks of varying sizes would have been much more convenient.

I went with ZFS in the end. (Only about 4 months ago, so the info is reasonably current)

[Anonymous Coward]

Re: New???

BTRFS is the FS on all Netgear ReadyNAS, the OS of which is based on Debian.

Of course a NAS is a special case of server, and they can have greater control of which features are used, but BTRFS is pretty stable on it.

[John Sanders]

Re: New???

Nothing stops you from running btrfs on top of mdadm and get solid RAID5 functionality.

That is how I have been building btrfs pools for a long time and had very little issues.

Having said that, zfs is great.

[Steve Davies 3]

RedHat seem to think Btrfs is ok to use

https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/Storage_Administration_Guide/ch-btrfs.html

[Dan 55]

Re: RedHat seem to think Btrfs is ok to use

Yes, but Red Hat also think systemd is okay to use.

[Paul Crawford]

Re: RedHat seem to think Btrfs is ok to use

Also many RHEL customers are sticking to version 6 and older proven stuff. They value stability and dependability over "new and probably broken on arrival or the next update".

[picturethis]

Re: RedHat seem to think Btrfs is ok to use

This is pretty-much what I've been doing for the last 4 years or so. RHEL/CentOS 6.7 (the latest/last in the 6.x series) for servers and Linux Mint Cinnamon 17.1/17.2 for desktops. Both for professional use and personal use.

For me, this seems to be a good combination of "stuff that just works" and getting latest stuff (that also "just works") and still supported. I'll continue along this path until there's a path with even less overhead.

"Overhead" being defined as my time spent screwing around with shit that should just "work". My time is worth more to me now as I get older (since I have less of it). Computers/Software are mostly just tools for me now, tools used to get more useful work done.

[Anonymous Coward]

Re: RedHat seem to think Btrfs is ok to use

RHEL/CENTOS 6 the new Windows XP/Server 2k3 except the community could decide to support those for a very long time indeed if it was deemed sensible.

[BitDr]

Re: RedHat seem to think Btrfs is ok to use

Red Hat also thinks systemd and crawling into bed with Microsoft is a good idea. One of those has a history littered with the husks of its partners... the other one is just starting out on it's career path.

[Edit] Sorry, I hit [submit] before seeing others referencing the first point.

[Anonymous Coward]

@Steve Davies 3 - Re: RedHat seem to think Btrfs is ok to use

Maybe their lawyers know something we don't know, or maybe RedHat fears Oracle lawyers more than Ubuntu does.

[Steve Davies 3]

Once upon a time...

In a time oh so very long ago, the arrival of a new Ubuntu release was actually big news.

Now?

May I suggest it is now more of a 'Meh' moment.

Sure, the fans will rave on about it but many here we have jumped ship to other distros as they have either outgrown it or got fed up with it.

Mint seems as popular destination for former Ubuntu fans.

I went to the Linux 'dark side' myself and now use CentOS. Still on CentOS 6 so that I can avoid 'systemd' for as long as possible.

As for ZFS, I'm far too old to get excited about a filesystem. I was once very excited about one once upon a time. Well I would be wouldn't I as I wrote my own one as a student back in the 1970's. It was a new Filesystem for RT-11. It worked most of the time still, it was fun for a while. All that bit twiddling deep inside the kernel. Oh, the memories.

[Anonymous Coward]

Re: Once upon a time...

But Mint releases are now based on Ubuntu LTS releases, so this is good news for anyone waiting for the next Mint release, since it means they have a stable base to work from.

Also in the corporate world, LTS Ubuntu releases are the only ones worth considering since the normal cadence releases are supported for such a short amount of time. At work we have a corporate 14.04 image, so I'd hope to see a corporate 16.04 image sometime soon (not holding my breath though).

I'd agree that the non-LTS Ubuntu images barely raise a 'meh', but LTS releases are newsworthy. And it's all the more welcome if they don't try to introduce new UI 'features'.

[Anonymous Coward]

Re: Once upon a time...

I like my OS system releases to be meh to be honest. I like "We made everything a bit better, and we added some new things that you can completely ignore if you like but we think you'll like them"

[Mage]

Centos & Mint

Well, Centos is basically RedHat, a more sensible choice for a server than eye candy Ubuntu.

Mint with Mate desktop is far closer to what a user workstation / netbook / laptop /desktop should be, but it and Ubuntu are basically Debian, Unbuntu with more stupid bits. I used to use Ubuntu, now I use Debian, Mint + Mate, sometimes Centos.

[phuzz]

Re: Centos & Mint

Ubuntu Server ships with no GUI, and thus no eye candy. Mint is based on Ubuntu 14.04, although they'll be moving to 16.04 now it's been released.

[Teiwaz]

Re: Centos & Mint

"Mint with Mate desktop is far closer to what a user workstation / netbook / laptop /desktop should be, but it and Ubuntu are basically Debian, Unbuntu with more stupid bits."

- By 'stupid bits', you mean Unity?

- Well pardon them for trying to ensure 'linux as a user OS remains relevant into the mobile age in their own way. The Gnome ui has some potential as a tablet interface ans KDE also have plans, but that doesn't mean Canonical can't also have their own vision.

It's not as if you can't install a more traditional desktop ui also (from kde to mate or even xmonad).

[frank ly]

Re: Centos & Mint

Mint 17.3 (released January 2016, based on Ubuntu 14.04) is LTS until April 2019. Mint 18 (to be released May/June 2016, based on Ubuntu 16.04) will be LTS until 2021. It's a bit close and a big overlap.

I got Mint 17.3 installed and hammered into shape at the end of January. After that exercise I'm going to wait at least a year before trying Mint 18.

(I was quite happy for three years running Mint 13 which is LTS until April 2017.)

[bombastic bob]

Re: Centos & Mint

"By 'stupid bits', you mean Unity?"

I know *I* would. Unity is like "phone on a desktop", _EXACTLY_ what M$ is doing WRONG these days.

"the mobile age" is HIGHLY overrated. *NOBODY* does "desktop things" on a phone, "Continuum" is a freaking JOKE, and "tablet interface" *ONLY* works on a fondleslab, which has a dwindling popularity if you read the market correctly (basically they're just oversized PHONES and are used in similar ways).

So SERIOUS computing is STILL done with keyboard+mouse on a DESKTOP computer, which needs a DESKTOP interface, like Mint has. Of course, Ubu can run MATE as well. That's what I do with Ubu - MATE desktop.

[Teiwaz]

Re: Centos & Mint

The mobile is overrated (and over hyped) in some respects, but how many people do you see a day using mobile interfaces vs. desktop (provided you actually go out in public).

If 'linux remains a desktop only interface it faces fading into even more into the obscurity of the server room (and maybe only seen running inside Azure).

Serious use will always be desktop oriented, as it's done at a Desk. The IT 'desktop' paradigm could still use some work. The mouse is spectacularly poorly designed for extended use, the keyboard less so (but most layouts designed to prevent long gone mechanical parts from getting tangled).

Unity is one of the few desktops that work really well without recourse to the thrice damned mouse (aside from tling wms)

[oomwat]

ZFS

If ZFS support is implemented as an 'application' which downloads the source and then compiles it in, then that shouldn't break the GPL ... I've not looked, so I could be wrong, but that'd be the way I'd do it.

[Anonymous Coward]

Re: ZFS

It isn't. It's supplied as a pre-built, binary kernel module.

[WonkoTheSane]

Re: ZFS

I believe oomwat is suggesting it would be more legally acceptable if supplied as a DKMS module.

See https://en.wikipedia.org/wiki/Dynamic_Kernel_Module_Support

[bombastic bob]

Re: ZFS

there are MANY reasons why a module needs to have a 'closed source' component. One of them might be FCC rules, for things like wifi drivers, where the requirements for certifying your driver would be to take steps to prevent modifying your software if the software controls things like power tables and frequencies [the things that FCC regs, well, REGULATE].

So Broadcom WiFi will *ALWAYS* have a 'BLOB' because the FCC regs demand it. It also has a "wrapper" so you can re-compile the kernel. I'm not entirely sure how the source for the binary kernel module for ZFS actually works, but I expect that the source *IS* available. The license contention has to do with the Sun/Oracle requirements, and GPLv2 can't have "additional requirements" if it's GPL-covered. So it's being shipped as a NOT GPL component that's (as I understand it to be) dynamically loaded, "bundled" with the OS and not "a separate component" as far as Stallman and others are concerned. THEY *feel* (not think) it's a "violation", but you can bundle non-open-source software with Linux if you want. PURISTS won't, but sometimes it's needed, and the INTENT of the GPL is to allow modification of GPL-covered code and "the system still works".

So if you can re-compile the kernel and the ZFS module for the new kernel, there should be NO issue with the GPL. If re-compiling the kernel PREVENTS you from using the ZFS module, there MAY be an issue, but if the module is "not part of the GPL code repository" then I'd say NO, it's NOT a problem, and I hope the courts agree with me. Otherwise, the "fascist enforcement" of 'open source forever' from the use of a trivial GPL component may COME BACK TO BITE ITSELF IN THE ASS. In other words, the unintended repercussion might be something the FSF and Stallman don't want to deal with even MORE than having ZFS survive as part of Ubuntu.

[Destroy All Monsters]

Re: ZFS

there are MANY reasons why a module needs to have a 'closed source' component

Not a single one. Unless you like to play The Gimp in Zed's Basement.

Also, "regulation" doesn't work by "closing source". It just doesn't. If the FCC "demands it" (haven't checked), the FCC is full of cluless idiots doing Broadcom's bidding (oh surprise!)

If you want "tamperproof", well, you better have the appropriate chip on you board. This is known as "trustworthy computing", i.e. trustworthy from the manufacturer side while the user gets an additional bell on his collar.

So if you can re-compile the kernel and the ZFS module for the new kernel, there should be NO issue with the GPL.

Your personal interpretation of the GPL is well noted.

[thames]

Another routine desktop upgrade.

Most of the big changes - ZFS, LXD, Ceph, seem to be mainly relevant to the server version. Even if you wanted to use ZFS for example on the desktop, you won't get it unless you reformat your drive. Few desktop users will bother, because quite frankly few will really care about it.

For desktop users, it looks like its mainly a series of incremental upgrades. I don't expect to really notice much difference other than that some apps will be at newer versions.

I expect that I'll upgrade my desktop some time tomorrow or the next day when I get a bit of spare time. It's generally pretty painless. There will be more desktop configuration options, but I doubt that I'll look at them. I can't be bothered to change much other than changing which icons are in the launcher (which should carry over from the current version) and the desktop wallpaper. I suspect that most other users are the same.

I know this doesn't make for exciting news, but I think that most desktop users aren't looking for excitement. They're looking for boring predictability with marginal improvements, and I expect 16.04 to deliver on that when it comes to the desktop.

People looking for excitement will need to look to the server, including the new IBM mainframe support.

[Kubla Cant]

Re: Another routine desktop upgrade.

People looking for excitement will need to look to the server, including the new IBM mainframe support.

I'm always on the lookout for excitement, but I don't think my house is big enough or my electricity supply powerful enough to run an IBM mainframe.

[PlinkerTind]

Re: Another routine desktop upgrade.

"...I'm always on the lookout for excitement, but I don't think my house is big enough or my electricity supply powerful enough to run an IBM mainframe...."

Well you can emulate an IBM Mainframe on a laptop, using the open source TurboHercules. An old 8-socket Nehalem-EX would give you 3.200 MIPS, which is a decent midsized Mainframe. If you got the latest 8-socket x86 you would get something around 10.000-15.000 MIPS. And software emulation is 5-10x slower than running native code, so if you ported the Mainframe software to x86, then the latest 8-socket x86 server would actually give 50-75.000 MIPS - which is what IBM's largest Mainframe gives today. In other words, Mainframe cpus are not really that fast. You need 20 ish Mainframe cpus to match one 8-socket x86 server. But Mainframes are extremely expensive for the performance you pay.