Native ZFS Is Coming To Linux Next Month

An anonymous reader writes "Phoronix is reporting that an Indian technology company has been porting the ZFS filesystem to Linux and will be releasing it next month as a native kernel module without a dependence on FUSE. 'In terms of how native ZFS for Linux is being handled by this Indian company, they are releasing their ported ZFS code under the Common Development & Distribution License and will not be attempting to go for mainline integration. Instead, this company will just be releasing their CDDL source-code as a build-able kernel module for users and ensuring it does not use any GPL-only symbols where there would be license conflicts. KQ Infotech also seems confident that Oracle will not attempt to take any legal action against them for this work.'"

Comments

Freedom ain't free

[binarylarry]

How open are the two open source licenses if they prevent you from using the software?

Re:Freedom ain't free

[Anonymous Coward]

They don't prevent you from using it. However, the GPL prevents redistribution of less free combinations. You can use anything GPL internally (in your home or inside your organization) combined with anything non-free which you like, you just are not licensed to redistribute it to external organizations or people.

Re:Freedom ain't free

[h4rr4r]

Sun used the CDDL just to make sure Linux never got ZFS. Even that move is not going to save solaris, only open sourcing it earlier would have done that. I say this as a linux user who likes solaris and thinks it will be a shame to see it die. Well I like it once the GNUtools are installed, the solaris versions sucked.

They are both quite open, how free they are some might argue about.

Re:Freedom ain't free

[stinerman]

They don't prevent use. They prevent redistribution as part of the whole.

I can download, build, and install fglrx (which is completely non-free) or this ZFS module. I just can't distribute either module linked into the kernel.

Re:Freedom ain't free

[outZider]

No, Sun used the CDDL because they hate the restrictions on GPL. The sharing issues go both ways, Sun wanted to keep some ownership. It's not like the BSD license exists just to spite GPL.

Re:Freedom ain't free

[h4rr4r]

No, they are a company that exists to make money. Saving Solaris would make them more money. Very simple. Corporations do not hate like that, they only do what they must to maximize profit.

BSD is a fine license, it was created for a real purpose, not to just protect a doomed product.

Re:Freedom ain't free

[larry bagina]

There's a despair poster, I believe, with a caption along the lines "it could be, your main purpose in life, is to provide a warning to others". (Damn it, the internet made me check ... "It could be that the purpose of your life is only to serve as a warning to others."

ZFS's purpose was not to be a next generation file system, but to encourage next generation file systems to be built. Free Software has a tendency to get stuck at "good enough" sometimes. And someone has to come along and show that there is a better way. Competition is good. Sometimes it's internal (gcc vs egcs), sometimes it commercial (CVS vs perforce and bitkeeper).

What if ZFS was GPL? What if it went into Linux? It might get incremental tweaks, but it would stagnate at "good enough". Instead, btrfs, hammer, etc were developed -- much better, much cleaner file systems.

ZFS has some cute tricks. What could be better than taking a sledgehammer to a disk drive without causing problems? But ultimately, ZFS would hold linux back.

Re:Freedom ain't free

[symbolset]

No, Sun used CDDL to be deliberately incompatible with the GPL because they knew the good bits would get assimilated into Linux if they didn't.

Re:Freedom ain't free

[EvanED]

Free Software has a tendency to get stuck at "good enough" sometimes.

Not just free software. IE6 has the poster child of getting stuck at "good enough" until it was kicked out of place.

Re:Freedom ain't free

[Anonymous Coward]

they knew the good bits would get assimilated into Linux

Linux is Borg now?

Re:Freedom ain't free

[coerciblegerm]

No, Sun used the CDDL because they hate the restrictions on GPL. The sharing issues go both ways, Sun wanted to keep some ownership. It's not like the BSD license exists just to spite GPL.

This is the third time I've seen someone post something to this effect in the past week. I smell a smear campaign. Nonetheless, I'm calling BS here. Daneese Cooper, one of the individuals who helped draft the CDDL, stated that they based the CDDL on the MPL "partially because it is GPL incompatible. That was part of the design when they released OpenSolaris." It was made deliberately GPL-incompatible, but this has nothing to do with 'restrictions' in the GPL.

Re:Freedom ain't free

[jedidiah]

> ZFS's purpose was not to be a next generation file system, but to encourage next generation file systems to be built. Free Software has a tendency to get stuck at "good enough"

This is a tad ironic considering that a 3rd party had to bail out Sun in this regard.

Solaris was the perfect example of "rut-ware".

Re:Freedom ain't free

[hedwards]

Precisely, FreeBSD for a long time (Possibly still) shipped with two math emulators, one which was compiled in by default was kind of bogus but BSD licensed, the other which only included source was GPLed, but happened to work better. It was a bit of a pain, but ultimately worked, people didn't usually redistribute custom kernels anyways, so the requirements of the GPL weren't particularly restrictive on the typical user and everybody could have what they wanted.

Re:Freedom ain't free

[fucket]

Thanks for linking to the thumbnail. I'll take your word for it that that's what it says.

Re:Freedom ain't free

[afidel]

Pretty much, it absorbed the best parts (except SMIT) from all the Unix OS's from the 80's and 90's.

Re:Freedom ain't free

[Anonymous Coward]

If by good enough you count "we'll pay the families of the victim, it'll be cheaper than fixing the Pintos" as good enough...

Re:Freedom ain't free

[EvanED]

Personally, I still think IE was better than the alternatives for some time. Netscape 4 and IE 5 were about on par; each had advantages over the other. IE improved a bit to 5.5 and to 6, while Netscape didn't. And it was a long time until Firefox showed up on the scene.

Actually your analogy is kind of funny, because while I'm by no means an expert, I did read the Wikipedia article on it about a week ago, and there's a section about an article in the Rutgers Law Review called "The Myth of the Ford Pinto Case" that basically said (1) the Pinto really wasn't worse than most cars of the era and (2) that memo you alude to is incredibly misdescribed. (The latter point is supported by the fact that the trial judge actually ruled it inadmissable.) In particular, it was not a "the cost of the lawsuits will be less than the cost of a recall" analysis.

(Of course this means that in some sense you could say IE6 was worse than the Pinto. ;-))

Re:Freedom ain't free

[CAIMLAS]

I'll be the first to say that ZFS has some shortcomings and limitations.

However, it's like Active Directory is to the workstation/server model of enterprise networking: it does everything. There is nothing else which comes close (or shell we say, 9/10ths of the way) to it in terms of it's "completeness" and feature set. Yes, it has some severe limitations ("Windows only") and shortcomings ("OMG it's a pain to troubleshoot"), with a fairly deep learning curve as well as a limited domain of applicability beyond the base subset (network administrators/storage techs).

But at the end of the day, they do things easily which most other products can't even do in such a complete fashion. Before such capability can be surpassed, it has to be met. AD and ZFS have been out now in more-or-less their current incarnation for close to 7 years, and only a bare subset of those features are elsewhere (and in less-than-ideal development status).

Frankly, ZFS is a (the) "next generation filesystem" for these reasons. It made assumptions (which are wrong), but do not make it lack utility. The management toolset is clean (very clean).

How, exactly, was being the conceptual "next generation" not the conceptual goal for the first 64 bit, CoW, RAID-built-in filesystem available? Seems to me that's why it's the fox everyone is chasing...

Re:Freedom ain't free

[Anonymous Coward]

Sun used the CDDL just to make sure Linux never got ZFS.

[citation needed]

Re:Freedom ain't free

[Anonymous Coward]

This is the third time I've seen someone post something to this effect in the past week. I smell a smear campaign.

Nonetheless, I'm calling BS here. Daneese Cooper, one of the individuals who helped draft the CDDL, stated that they based the CDDL on the MPL "partially because it is GPL incompatible. That was part of the design when they released OpenSolaris." It was made deliberately GPL-incompatible, but this has nothing to do with 'restrictions' in the GPL.

And Cooper's assertion was reject by Simon Phipps, Sun's Chief Open Source Officer for quite a while (before leaving Oracle in the last few weeks):

http://www.opensolaris.org/jive/message.jspa?messageID=55013#55008
http://en.wikipedia.org/wiki/Common_Development_and_Distribution_License#GPL_incompatibility_controversy

Re:Freedom ain't free

[Thundersnatch]

Instead, btrfs, hammer, etc were developed -- much better, much cleaner file systems.

How can filesystems that don't exist in stable release form yet be "better" than ZFS?

ZFS is far ahead of btrfs, both in terms of stability, features, and usability. Btrfs doesn't have parity RAID, dedupe, or replication yet. These are critical features for large-scale systems. In short, it isn't even close to ZFS. ZFS is also "cleaner" in my opinion, in both design and UI. Oracle funding most btrfs development also raises a question of btrfs momentum now that they own ZFS and Solaris.

Re:Freedom ain't free

[Jaxoreth]

What could be better than taking a sledgehammer to a disk drive without causing problems?

Shooting it with a .45?

Re:Freedom ain't free

[r7]

Sun used the CDDL just to make sure Linux never got ZFS

That's a paranoid way of looking at it. Why Sun chose the CDDL instead of the GPL was to retain the right to fork a proprietary version while retaining GPL-like lincensing in all other respects. The CDDL is a better choice for businesses than the GPL as it offers some degree of protection from FSF lawsuits while still requiring that modifications be contributed back to the community.

Quoting http://blogs.sun.com/chandan/entry/copyrights_licenses_and_cddl_illustrated
   

A common misconception is about CDDL and GPL incompatibility. (Incompatibility in the sense: to combine two source files, one under GPL and another under CDDL, to create a common executable.) GPL is incompatible with most licenses like Mozilla Public License, Apache, and CDDL. GPL wants you erase those licenses and use GPL in that place, where as these licenses do not permit erasing them. Hence the incompatibility deadlock.

Re:Freedom ain't free

[Anonymous Coward]

The GPL prevents redistribution of more free combinations as well.

Re:Freedom ain't free

[Anonymous Coward]

They could have adopted the MySQL model (since the GPL can't really put restrictions on (re)distribution authorized by all the copyright holders (except maybe the patent clausules)).

Re:Freedom ain't free

[outZider]

It's not hard to be incompatible with the GPL. The GPL prohibits a lot of actions. Good for some, not for others.

Re:Freedom ain't free

[Anonymous Coward]

And I'm sorry, but Danese Cooper is *not* the authoritative source on this. Despite whatever she claims, there are far more people that are at Oracle now, or that are Ex-Sun that will tell you the opposite.

Re:Freedom ain't free

[onefriedrice]

The designers of the CDDL may have had some sort of agenda against the GPL, but the fact remains; when it comes to license compatibility issues in general, it is the GPL which is decidedly incompatible with every other license. The CDDL would be the rare exception of a license that is incompatible with the GPL on purpose, and Sun obviously had business reasons to do it. Therefore, while the GP is very probably wrong in asserting that Sun uses the CDDL because they hate GPL restrictions, he is also probably correct (from what I've seen) that some GPL advocates tend to view those who choose a non-GPL license as trying to thwart GNU and/or Linux so they don't have to admit that maybe other licenses have terms and conditions that have their own merit.

Re:Freedom ain't free

[the_womble]

IE 6 got stuck a long way short of good enough.

Re:Freedom ain't free

[Tubal-Cain]

Larger version

Re:Freedom ain't free

[jonbryce]

Good enough at that time was better than Netscape 4 or the very early unstable builds of Mozilla.

Re:Freedom ain't free

[node 3]

No, they are a company that exists to make money.

Companies don't exist to make money. They exist to serve the desires of those involved in the company. One aspect of that is usually to make money (non-profit corporations, for example (and by definition), do not exist to make money).

In the case of the license for Solaris was chosen to serve a purpose, and that purpose was almost certainly not specifically to make money. Sure, making money was taken into consideration, but acting like it was the sole motivator is not exactly rational. No more so than saying a person takes a specific job with the sole motivation of making money. That's almost impossible, as there are plenty of motivations for the exact jobs people choose, and while making money is the most obvious, it's not as simple as just that.

Now, to be sure, those in charge of corporations in the US tend to run them with the sole goal of making money, that doesn't make it some sort of universal truth. Whenever I see someone who tries to portray it as such generally has as their motivation a way of promoting actions that would otherwise be indefensible for their inherent repugnance.

Re:Freedom ain't free

[kainosnous]

I would also hate to see Solaris die. I remember looking at it a while back, and thinking that it had promise. However, with all the talk about copyright issues and incompatibilities, I thought it best not to get too familiar with it. I think that at the time, one reason that I liked it was ZFS. The other reason is that it was one more OS out there for my list of "Other OS's that are not Windows or Mac". It comes in handy in conversation sometimes.

Re:Freedom ain't free

[fusiongyro]

Only open sourcing it earlier? So they could undermine a larger version of their own community?

Open sourcing Solaris clearly did not save Solaris. Solaris is probably a better operating system than Linux, but apparently the only way Oracle thought it could be “saved” it is by charging a fortune to shove it down the throats of its own customers. Whatever. Worse certainly seems better in this case.

Re:Freedom ain't free

[Anonymous Coward]

It's not like the BSD license exists just to spite GPL.

Yes. It's the other way around. GPL exists because some people fear their free work could be used in commercial products.

Re:Freedom ain't free

[Lennie]

Maybe someone at Oracle started the btrfs project, but there are many contributers to btrfs.

Re:Freedom ain't free

[Anonymous Coward]

Sun used the CDDL just to make sure Linux never got ZFS.

Nah. There was no way that Sun would use the GPLv2 for OpenSolaris because
o GPLv2 would limit Sun's paying customers too much
o GPLv2 does not contain a patent grant - an evil entity could publish or contribute code under the GPLv2 and yet sue for patent violations.

Re:Freedom ain't free

[Anonymous Coward]

Could you explain why you can't make a proprietary fork from GPL software if you require copyright assignment? The linked blog post doesn't explain anything because the diagrams he talks about are not actually visible...

Also notice that when he talks about the misconception about incompatibility, he doesn't actually mention what the misconception is.

Re:Freedom ain't free

[Anonymous Coward]

when it comes to license compatibility issues in general, it is the GPL which is decidedly incompatible with every other license.

That's FUD if I've ever seen FUD. Check out the FSF's list of free software licenses; there's many licenses that ARE GPL-compatible. Excluding the GNU licenses themselves, there's at least Apache 2.0, Artistic 2.0, Berkeley DB, Boost, Modified BSD, CeCILL, Clear BSD, Cryptix, eCos 2.0, Educational Community 2.0, Eiffel Forum 2, EU Datagrid, Expat, FreeBSD (!), FreeType, iMatix, Independent JPEG Group, imlib2, Intel Open Source, ISC, NCSA, Netscape Javascript, OpenLDAP, Perl 5, PD, Python 2, Python up to 1.6, Ruby, SGI B 2.0, SML/NJ, Unicode, VIM 6.1+, w3c, webm, WFTPL 2, X11, XFree86 1.1, zlib and Zope 2.

And keep in mind that these are *licenses*; in reality, most projects won't even bother making up their own licenses. "Decidedly incompatible with every other license". Sheesh!

some GPL advocates tend to view those who choose a non-GPL license as trying to thwart GNU and/or Linux so they don't have to admit that maybe other licenses have terms and conditions that have their own merit.

Who are those mysterious "GPL advocates" you mention, then? Also, what does this have to do with a situation where Sun really WAS trying to "thwart GNU and/or Linux", by its own admission?

Look, the CDDL isn't a bad license per se, and the FSF page linked above lists it as a free software license, too, if a GPL-incompatible one (it does urge you not to use it for that reason, but hey, this *is* the FSF). But the original point was that Sun wanted to make sure that ZFS etc. would not be available on Linux, and they chose/engineered a GPL-incompatible license specifically to ensure that. You're not even contesting that anymore, so why are you still arguing about the whole thing?

It's a fact. Sun didn't want Linux to get ZFS. Get over it.

Re:Freedom ain't free

[TheRaven64]

Comparing hammer to ZFS is also a bit silly. Hammer was developed precisely because ZFS did not solve the problem that DragonflyBSD wanted solved. ZFS is designed for large SANs controlled from a central server. Hammer is designed to allow you to treat every disk on a network as part of the same storage pool. They are diametrically opposed objectives, and a filesystem designed to do both would need to either make painful compromises or have so much variation in code paths that it would effectively be two different filesystems.

You can do something similar with ZFS in FreeBSD, because ZFS slots into the GEOM system and can use any GEOM provider as the backing store, meaning that you can use remote partitions exported over the network, but you'd need a massive amount of configuration and get a lot of fragility for something that hammer does automatically and reliably. Conversely, hammer has incredibly poor performance on a number of workloads where ZFS does very well and doesn't provide the same level of redundancy on a single machine.

Btrfs, at the moment, is largely vapourware. It might become something impressive in the future, but for now it is not.

Either way, porting ZFS to Linux is probably a mistake. The FreeBSD port has some performance issues from the mismatch between the design of the ZFS code and the rest of the kernel, but more importantly it's not as flexible as it could be. ZFS is highly modular. The FreeBSD GEOM stack is also incredibly modular. If you were doing a native ZFS implementation for FreeBSD, you'd rewrite each of the components of ZFS as a separate GEOM module. Instead, the entire ZFS stack is exposed, more or less, as a single GEOM module. A lot of the potential flexibility of ZFS is lost by doing this, but it's done because it's much easier than a complete reimplementation.

Re:Freedom ain't free

[Anonymous Coward]

> ZFS's purpose was not to be a next generation file system, but to encourage next generation file systems to be built. Free Software has a tendency to get stuck at "good enough"

This is a tad ironic considering that a 3rd party had to bail out Sun in this regard.

Solaris was the perfect example of "rut-ware".

Sun is/was an engineering company, not sales. They're marketing has also generally sucked. Awesome product is fine, but if you can't money off it then you're screwed.

Re:Freedom ain't free

[aliquis]

Ohwell, Oracle can always run their database on Linux once they managed to kill Solaris. And by then ZFS on Linux will be good for them ;)

Re:Freedom ain't free

[Anonymous Coward]

"Well I like it once the GNUtools are installed, the solaris versions sucked."

That is because you were weened on a GNU diet (GNU is not UNIX) and don't have a clue about a true System V Release 4.0 UNIX.

Your loss is my gain. That's why I make the big bucks and most likely will never have to worry about finding a highly paid job.

Please keep being ignorant so that I won't have any competition. Thank you kindly in advance.

Re:Freedom ain't free

[GreyWolf3000]

It doesn't sound like there was an agenda against the GPL itself, just the intention to release the code under a license that prevented bundling with Linux.

Re:Freedom ain't free

[WNight]

No, that's cool. We just don't want it used in any way where the users don't get the code when they want it.

Re:Freedom ain't free

[Xeleema]

And it was a long time until Firefox showed up on the scene.

True. However, while MSFT was spoon feeding everyone IE5 5.5 & 6, we had the Mozilla Suite 1.7x (02apr03), which became SeaMonkey (30jan06) after Firebird and Thunderbird were forked out of the code base. THEN Firebird was renamed Firefox, an ad was put in the NY Times, and the world started to pay attention.

Re:Freedom ain't free

[Anonymous Coward]

Btrfs doesn't have parity RAID, dedupe, or replication yet.

Surely parity RAID is a function of the MD layer, and replication is a function of LVM?

Re:Freedom ain't free

[Urkki]

The GPL prevents redistribution of more free combinations as well.

...more free combinations which allow further redistribution of less free combinations, to be more exact.

Without this, GPL would be rather pointless. If somebody wants to keep their code free, and by extension, allow all future users of the code have certain freedoms that come with having the source code (what freedoms exactly, depends on GPL version, due to tivoization), then that's what is needed.

Freedom (of any kind) is not black and white thing, nor is it one-dimensional scale.

Freedom to take freedom away sure is a freedom, but I can see why some would want to restrict that freedom when it's about something they've created and want to remain free.

Re:Freedom ain't free

[EvanED]

I used the Mozilla suite for a while (web & mail), but it definitely wasn't a clear win over IE. It was slower, at least for me, much more of a memory hog, and didn't work with as many websites. It was only once Firefox was stable that there was something that was clearly better than IE. I probably switched from a combination of IE/Mozilla to Firefox a bit late into the party, but there were still 3 1/2 years between the release of IE 5 and the release of Firefox (let alone when it became stable; it didn't hit 1.0 for 2 years after that, though it was good before then); during those 3 1/2 years, IE was king IMO.

Re:Freedom ain't free

[hjf]

If you read the ZFS documentation you will understand WHY RAID was built into ZFS. Until then, STFU.

Re:Freedom ain't free

[Thundersnatch]

Surely parity RAID is a function of the MD layer, and replication is a function of LVM?

Only if you're stuck in the LVM-on-Linux mindset, which frankly sucks. There are many reasons why; any of Jeff Bonwick's writings will explain these ZFS design choices. It's also one of the main reasons well-heeled shops run Linux volumes from SANs instead of just using LVM.

For example, LVM snapshots require space reservations, and copy a block to every snapshot with every write. This is extraordinarily slow, and gets worse with each additional snapshot. This makes LVM snapshots effectively unusable for high-volume server systems, where they are needed most.

Re:Freedom ain't free

[ToasterMonkey]

So has everything else in production today. What's next, it'll absorb the best features of Windows 95?

Re:Freedom ain't free

[Cato]

Porting ZFS to Linux is only a mistake for those who don't like ZFS or Linux... I am contemplating a new Solaris / NexentaStor NAS just to get access to ZFS, but it would be much easier if it was available in Linux within the kernel.

Re:Freedom ain't free

[trasz]

Looks like some folks try very hard to notice a very simple fact: that in every case involving license incompatibility one of the parties is always GPL. There are GPL-compatible licenses, there are GPL-incompatible licenses, but there are no e.g. Mozilla-incompatible licenses other than GPL. In other words - in the Open Source world license incompatibility problem just doesn't occur when not dealing with GPL. Also, GPL-incompatibility is not the only problem with that license. GPL code cannot be incorporated into software licensed under most GPL-compatible licenses, such as BSD, Apache, Xorg, whatever. But I digress.

Re:Freedom ain't free

[trasz]

Actualy, CDDL is years ahead of Btrfs _wishlist_. Looks like Btrfs developers didn't even start thinking about e.g. hybrid pools yet, or proper support for SMB shares (including NFSv4 ACL support), or deduplication.

Re:Freedom ain't free

[VulpesFoxnik]

So has everything else in production today. What's next, it'll absorb the best features of Windows 95?
We have Wine now, don't we?

Open Source != Free Software

[cpicon92]

It's open source in the sense that the source is open. Free to view, and free to use as long as you don't distribute it.

Re:Open Source != Free Software

[guruevi]

I don't know if that's true. I know you probably can't redistribute the kernel with the CDDL bits but you can redistribute them separately (CDDL = Common Development and Distribution License). Then all you have to do is make sure that your software (or customer) installs the right bits and then you can get a pretty decent NAS box.

Besides the legal issues, I would love to see them tackle the technical issues. ZFS itself is very clean in code, very well documented and pretty simple once you get down to the wire. The issue (and selling point) is going to be performance and upkeep and for commercial implementations support. If the upkeep is going to be similar to BSD's implementation (several versions behind) or the performance as bad as FUSE, people are just going to stick to OpenSolaris (or one of it's commercially supported decendants like Nexenta).

Re:Open Source != Free Software

[Anonymous Coward]

CDDL is a free software license, so what's your point?

Reference: http://www.gnu.org/licenses/license-list.html#GPLIncompatibleLicenses

Re:Open Source != Free Software

[mysidia]

I don't know if that's true. I know you probably can't redistribute the kernel with the CDDL bits

Just like you can't distribute a Linux distribution and include nVidia drivers? Tell that to the distros.

But I think it may fall on death ears. Or you might hear back a reminder that mere aggregation on a storage medium is exempted in the GPL... as long as the non-free package is separated from the GPL'd package.

Re:Open Source != Free Software

[quercus.aeternam]

This is both Open and Free, just not quite as free as Stallman would like.

CDDL licensed code can be freely distributed and modified, so long as it is compiled with a compatible license.

This is why BSD has no issues with including ZFS. The BSD license is less restrictive than the GPL.

Re:Open Source != Free Software

[straponego]

Death ears? Be very, very quiet... the death ears will get you.

Re:Open Source != Free Software

[mysidia]

I meant deaf ears, mainly because the distros. ignored whatever complaints they got and continued to include non-GPL-compatible drivers.

Re:Open Source != Free Software

[hedwards]

The main reason why the BSD implementation is several versions behind is that ZFS has been under rapid development lately. Last time I checked, FreeBSD had the important features ported, with a few nice to haves not yet available. The important work was on the performance end of things and squashing OS specific bugs. This won't remain like this forever, at some point ZFS will mostly be fleshed out and in need of only a small number of tweaks, at which point I doubt other OSes will be that far behind. A filesystem isn't terribly useful if only one specific release can properly utilize it.

Re:Open Source != Free Software

[Xtifr]

I know you probably can't redistribute the kernel with the CDDL bits but you can redistribute them separately

Not necessarily. I know programmers like to interpret the law as if it were a computer program--rigidly, and with no room for interpretation--but it doesn't necessarily work that way. The law will consider your intent, and if your intent is to deliver a derivative work, then forcing the customer to obtain the pieces and glue them together won't necessarily let you off the hook.

This came up most famously with Objective C. Originally, Steve Job's NeXT made a proprietary front-end to GCC, and they made their customers get the rest of GCC for themselves. When the FSF objected, Steve consulted his lawyers, and decided to release the Objective-C front end source, which is why GCC includes Objective-C today.

Of course, the laws may be different in India, but I wouldn't touch this module without consulting with my lawyers first, since I'm in the US.

Re:Open Source != Free Software

[Xtifr]

Where has Stallman ever said that it's not as free as he would like? I know the FSF disapproves of licenses that aren't GPL-compatible, but not on the grounds that they're not free enough. They simply don't like code that they can't use for their own projects--and who can blame them? I don't either, and I use a wide variety of licenses, including ones that the FSF isn't necessarily thrilled with.

Re:Open Source != Free Software

[Kjella]

This is both Open and Free, just not quite as free as Stallman would like.

No, the problem is that both licenses have the same kind of copyleft clause:

CDDL: "You may not offer or impose any terms on any Covered Software in Source Code form that alters or restricts the applicable version of this License or the recipients rights hereunder."
GPLv2: "You may not impose any further restrictions on the recipients' exercise of the rights granted herein."

Anything that is a requirement in one would be seen as a restriction in the other, the only way it would work was if one license was a strict superset of the other like the GPL and LGPL. The smallest of differences and it's incompatible because either the CDDL imposes restrictions on the GPL code or the GPL imposes restrictions on the CDDL code.

Two copyleft clauses go together about as well as matter and anti-matter. Sun knew that and there's nothing radically different in the CDDL to suggest they couldn't have used the GPL. They simply chose to pick their own license and try creating their own CDDL ecosystem instead.

It was honestly another example that as far as I can tell, Sun doesn't really "get" open source. Oh, they know how to open up the source of OpenOffice and OpenJDK and OpenSolaris, but they sucked at community building. If you make it a one-way street where only you push releases out, no wonder you don't get value back.

Re:Open Source != Free Software

[Anonymous Coward]

So that's what happened to Sun!

Re:Open Source != Free Software

[TheRaven64]

The Objective-C case never went to trial, so it can't be used as precedent, however, it differed in one important way: NeXT distributed GCC. There are two ways in which the GPL can affect your code:

The first is if your code is a derived work of some GPL'd code. This is obvious - you have no right to distribute any derived work of any copyrighted work (except occasionally under some fair use / fair dealings provisions), but the GPL provides a specific grant to do so if you use a license which is equivalent to the GPL.

The second is if you distribute the GPL'd code. In this case, the GPL is not limited by the legal interpretation of a derived work. The only right that you have to distribute the GPL'd code is because the GPL explicitly grants you this right, and this can be withdrawn if you violate the GPL. Distributing code that links with the GPL'd code but is not under a compatible license is one such violation. The GPL can not prevent you from distributing your code, but it can prevent you from distributing the GPL'd code. This is why Linux distributions do not include the binary nVidia drivers (instead, they include a script to fetch them from nVidia) - you can distribute the kernel, or you can distribute the nVidia code, but not both.

NeXT, in distributing the Objective-C blob, was violating the GPL by also distributing GCC. They could have worked around this by requiring their customers to grab the GCC code from the FSF then compile it, and link it to the Objective-C blob, and they decided it was simpler to just release the code.

It'a actually quite amusing that this is considered a win for the GPL. An Objective-C implementation is two parts: a compiler front end and a runtime library. The runtime is the more complicated part, and this was not released by NeXT. The code they did release was around 5,000 lines of really, really, terrible C code, which was useless without the runtime. The FSF wrote a new runtime, with slightly different interfaces, and made the code even more ugly by littering it with #ifdef statements. NeXT and Apple people never worked on this modified version, they worked on their own branch and left it to the GCC people to port the changes over after each release.

No one has touched the Objective-C code in mainline GCC for almost a decade. Meanwhile, clang (which is BSD licensed), has support for all of the new Objective-C 2 features on non-Darwin platforms, while GCC only supports them on Darwin with the Apple runtime, and only if you use Apple's branch, which lacks any of the new optimisations added since GCC 4.2.1.

Back on topic, this company should be completely safe distributing the kernel module, as long as they don't also distribute the kernel. They are safe from Oracle, because the CDDL is a per-file license and has no problem being linked with GPL'd code. They are safe from Linux developers as long as they don't distribute the kernel and their code is not a derived work of the kernel.

As both the GPL and CDDL are Free Software licenses, neither includes any limitations on use of the programs. This means that, as an end user, you can do whatever you want with them. Linking them together violates the GPL, but that just means that you lose your right to distribute them, it doesn't mean that you have to stop using them.

Re:Open Source != Free Software

[TheRaven64]

No, the problem is that both licenses have the same kind of copyleft clause:

Not quite. The CDDL is a per-file license. You may not alter the terms of the CDDL for code in that file, but you can link it with other files that include different code. The GPL, in contrast, applies the copyleft conditions to the complete work.

Re:Open Source != Free Software

[Xtifr]

You're right that none of these cases (and ObjC was just one example--maybe not the best) have been to court, but I don't want to be the first. Even if I believed I'd prevail, court is expensive and a massive waste of my time.

A better example might be the fellow who was distributing a patch to link RSA with FSF code, back when RSA was proprietary, patented, and only available from one source. Even though it was just a patch, it was pulled after an FSF challenge.

I'm not saying you're wrong, or that it's definitely not legal. I'm just saying, from what I've heard, it may not be as clear-cut as you seem to believe. I would consult a lawyer before attempting anything like it. Especially since rumor has it that Sun chose their license in large part to prevent ZFS and a few other technologies from being added to Linux.

Oracle owns ZFS now, and they still claim to be pro-Linux, despite the Google suit. If they want ZFS in Linux, they can easily relicense it. Until then, I'd rather not rely on legal trickery to get around the licenses, no matter how firmly some random guy on slashdot suggests that it "should be" ok. :)

Re:Open Source != Free Software

[cthulhu11]

I worked for Stepstone at the time. NeXT poached the lead ObjC developer from us, and his wife miscarried during the move to California from Connecticut. ObjC support in GCC rather effectively killed Stepstone.

Good Article

[bill_mcgonigle]

No, really. I had a bunch of questions going in, and they were all answered. This is rare enough to warrant a shout out to Michael Larabel.

I disagree with some of his subjective claims like x86_64 being a substantive limitation or ZFS on Linux remaining niche (I guess that depends on how you define the niche...) but he got the national lab project, the zpool version, the Oracle (nee Sun) patent problem. Kudos.

FreeBSD 9 is probably where ZFS will wind up finding a proper home, I'm guessing.

Re:Good Article

[h4rr4r]

How do you think it is not a substantive limitation?

My phone runs linux and is not x86 of any shape or register size, nor is my workstation, nor are many other machines I have running linux. This is just like people who think flash working only on x86 32bit linux is good enough.

If FreeBSD ever gets a good ZFS implementation expect lawsuits to commence.

Re:Good Article

[Monkius]

Sadly, yes, there is every reason to expect it.

Re:Good Article

[Just Some Guy]

If FreeBSD ever gets a good ZFS implementation expect lawsuits to commence.

None have been filed since it was production-ready last year. Besides, what would they sue over? The FreeBSD team using code that Sun deliberately and explicitly licensed for such things?

Re:Good Article

[bill_mcgonigle]

How do you think it is not a substantive limitation?

ZFS now only runs on servers. Yeah, they wanted to see it on digital cameras, but in the current market that's not real. Nobody who's currently using ZFS would blanch at the x86_64 requirement.

Sure, there are opportunities that will open up when it's ported further, but they're doing the right thing by getting it out where it'll get the most use.

Re:Good Article

[bill_mcgonigle]

None have been filed since it was production-ready last year.

It's not. Yet. There are many reports of lock-ups with uptimes on the order of a week. Soon, I hope, but don't set people up to hate on it.

Besides, what would they sue over? The FreeBSD team using code that Sun deliberately and explicitly licensed for such things?

It's not Sun you need to worry about, it's NetApp.

Re:Good Article

[h4rr4r]

It is not production ready, I know I tested it. The next version should fix those gripes. Patents are what Oracle will sue over.

Re:Good Article

[h4rr4r]

Because their are no servers running linux using non-x86_64 CPUs? You had better tell IBM that.

Re:Good Article

[Kjella]

Because their are no servers running linux using non-x86_64 CPUs? You had better tell IBM that.

If it's important enough for IBM, let them supply the patches for POWER chips. I'd think the x86_64 requirement is mostly for 64 bit addressing, which as far as I know th POWER chips do as well so it shouldn't be that hard given enough interest. Personally I'll wait and see how well this GPL/CDDL module business will work out...

Re:Good Article

[Lunix Nutcase]

Why? NetApp has been having it's ass handed to it over the patent claims.

Re:Good Article

[Anonymous Coward]

If FreeBSD ever gets a good ZFS implementation expect lawsuits to commence.

This is a ZFS system
# uptime
  9:38PM up 750 days, 5:19, 1 user, load averages: 0.16, 0.09, 0.08
# uname -sr
FreeBSD 7.0-STABLE

Re:Good Article

[bill_mcgonigle]

Don't be intentionally dense. The majority of the market that can help these guys refine their code is fine with x86_64.

Re:Good Article

[bill_mcgonigle]

Legal intimidation is a common tactic.

Re:Good Article

[Lunix Nutcase]

What's intimidating? In their case against Sun they've had a patent struck down and summary judgements against them saying that Sun doesn't violate others.

Re:Good Article

[mysidia]

That doesn't stop them from pursuing more claims against targets they perceive to be vulnerable. If they think they can get a quick win or quick settlement with BSD folks (to stop distributing), they will do so, and then use it as leverage to attempt to reach a settlement with Oracle.

Probably requiring Oracle to get out of the storage business and stop making ZFS available to ISVs/partners to build NASes with, or as open source

Re:Good Article

[bill_mcgonigle]

What's intimidating?

Being a hobbyist OSS developer and getting hit with a patent infringement lawsuit from a large corporation.

Re:Good Article

[mysidia]

Because ZFS is not production quality on a 32-bit CPU or with less than at least additional 2GB of RAM available for ARC, even on Solaris where ZFS is most mature. Bare minimum for ZFS: 1Gb RAM, 64bit proc.

If you have a 32-bit CPU or less than 2GB system RAM, use UFS or Ext3, forget about ZFS for such hardware configurations, unless you want to experience pain (system hangs, memory starvation, crashes / Panics due to 32-bit address space squeeze causing fragmentation and ultimately inability to allocate ARC efficiently).

Re:Good Article

[hedwards]

It is, however, it's not BSD code, it's a port of the code from Sun. Well, now belonging to Oracle, I'm not sure how they could sue the FreeBSD people without first filing suit against Oracle for infringing on whatever patents they claim are relevant.

Re:Good Article

[mysidia]

ZFS is just about as "production ready" on the latest stable release of FreeBSD, as BtrFS is "Production Ready" on the latest Linux kernel release.

Re:Good Article

[Anonymous Coward]

I disagree with some of his subjective claims

That's funny, I disagree with some of his objective claims. For example, the claim that native ZFS is coming to linux next month...

Re:Good Article

[mysidia]

It's worth mentioning that the latest version of Windows Server (2008 R2) is 64-bit only as well.

And ZFS has always had 64-bit as minimum system requirements for production systems, even on Solaris.

That is, 32-bit is considered okay for limited testing, unsuitable for production use, particularly for use with zpools larger than a few hundred GB in size or so.

If you have a 1TB or larger storage pool with ZFS, you need 2gb of RAM and a 64-bit CPU to have something acceptable and stable. This is true whether you used Solaris or BSD.

I consider it a good thing that the person porting to Linux is actually enforcing the basic 64-bit requirement. Maybe fewer people who don't read docs and 'system requirements' sheets will get burned that way, by not noticing that "32-bit is not suitable for enterprise use", and say ZFS on Linux 'sucks', because they screwed up basic configuration and deployment requirements ?

Re:Good Article

[arth1]

If FreeBSD ever gets a good ZFS implementation expect lawsuits to commence.

FreeBSD already has a fairly good implementation of ZFS. It's other file systems (ext4, xfs, jfs, jffs2...) that it lacks. If FreeBSD ever ditches archaic UFS, expect heads to explode :-)

Re:Good Article

[Anonymous Coward]

I suggest you brush up on the legal case before you put your foot in your mouth again.

Re:Good Article

[CAIMLAS]

If FreeBSD 8.1 RELEASE, STABLE and CURRENT are any indication, FreeBSD 9 has a long, long way to go to be the 'home' for much of anything outside of the odd server here and there the base for other projects (pfSense/freeNAS). It certainly is not a 'general purpose' OS - if for no other reason than its woefully lacking (and often unstable) hardware support.

And by 'long way to go' I mean it needs to regress - in stability and commitment to making all the subsystems actually work out the door. The current state of FreeBSD as a project is, well... lacking. I wish it were otherwise, but no.

(They could jump out of 1994 while they're at it and upgrade to a packaging system/package distribution architecture/anything more modern than what they're got while they're at it too, but that's probably too much to ask for.)

Re:Good Article

[Anonymous Coward]

Just because you have had problems does not mean other people have these problems. Many organizations are using FreeBSD with ZFS in production environments.

Oracle cannot sue over ZFS patents because the CDDL gives a patent grant.

I think it would be a good idea to actually know what the fuck you are talking about instead of pulling shit out of your ass.

Re:Good Article

[bill_mcgonigle]

Yes, that's why this talk of FreeBSD's ZFS being "production-ready" is disconcerting.

But, hey, your filesystem will be consistent when your OS hangs. ;)

Re:Good Article

[Galactic Dominator]

This person likes trolling FreeBSD. Notice there are no facts, only FUD.

http://slashdot.org/comments.pl?sid=1763546&cid=33346740

Re:Good Article

[Anonymous Coward]

FreeBSD uses UFS2, which is incredibly fast. However, my understanding is that its design does not scale well to large disk arrays due to long FSCK times. This is where ZFS comes in as its a perfect fit.

Re:Good Article

[Deagol]

UFS supports the geom journal (gjournal) journaling system, which reduces fsck times to something reasonable. It's not 100% absent, as with ZFS, but it works very well. Still not as fast a filesystem as straight UFS2, but faster than ZFS. I hear there's a journaling system being bolted onto soft updates, so that should prove to be an interesting development. Not sure what it offers over gjournal.

Re:Good Article

[Deagol]

You forgot the output to "mount" or "zpool status" or "zfs list".

Re:Good Article

[grepya]

How do you think it is not a substantive limitation?

My phone runs linux and is not x86 of any shape or register size, nor is my workstation, nor are many other machines I have running linux.

Because it would not be a substantive limitation for any genuine candidate for a ZFS deploy. Your phone or workstation are not those candidates... nor -- I'd be willing to bet -- are the highly unspecific "many other machines you have running linux"

Re:Good Article

[Deagol]

I've used FreeBSD since 6.0 as my daily workstation, and I have few complaints. I'm currently running 8.1-RELEASE/amd64 on a 4-core machine, 8GB ram, and almost a TB of ZFS space. ZFS isn't the best performer, granted, but I beat it up pretty good and have nightly snapshots running, and stability has never been an issue.

I can count the number of times FreeBSD has locked these past 5 years on 1 hand. 3 were due to the 3rd-party fuse-ntfs3 port (non-stock, try to avoid it when I can), and 2 from reading the entirety of /dev/mem with certain programs (not sure why this happens, and it surely must be a bug).

And I don't know why people dis the ports system, as I find it far more functional than anything in the Linux arena I've ever used. Yes, gentoo fans, this includes portage.

Re:Good Article

[grepya]

What's intimidating?

Being a hobbyist OSS developer and getting hit with a patent infringement lawsuit from a large corporation.

bill_mcgonigle -- what you're doing here is the very definition of FUD.

Re:Good Article

[Bert64]

I have a machine running FreeBSD 8.1 which has been up a couple of months with a zfs filesystem, is there anything in particular which could make it crash? I have successfully completed a number of compiles on it at least...

Re:Good Article

[David Gerard]

"Many organizations are using FreeBSD with ZFS in production environments."

Is there a list handy?

Re:Good Article

[David Gerard]

"They could jump out of 1994 while they're at it and upgrade to a packaging system/package distribution architecture/anything more modern than what they're got while they're at it too, but that's probably too much to ask for."

This is part of why I moved from FreeBSD to Ubuntu. (The other half being my laptop at the time being supported by Linux but not FreeBSD.) I really hate sysadminning a Linux kernel, BSD is so much saner - but apt is so ridiculously good that BSD should adopt it yesterday. Of course, they won't.

Re:Good Article

[David Gerard]

Speaking as someone who's used both - and fixed both when they ate themselves - apt knocks ports into a cocked hat.

Re:Good Article

[TheRaven64]

My phone runs linux and is not x86 of any shape or register size, nor is my workstation, nor are many other machines I have running linux

I can't speak for the Linux version, but ZFS on FreeBSD needs x86-64 for three reasons:

First, and most simply, this is the platform that all of the ZFS developers use, so it is the one that is most tested. This doesn't mean that it won't work elsewhere, it just means that it is not well tested anywhere else.

The second is a performance consideration. ZFS uses a lot of 64-bit arithmetic for computing checksums and so on. On most 32-bit platforms, doing 64-bit arithmetic means that you need to split the operands between two registers, effectively halving the number of GPRs that you have to work with. On x86-32, this basically limits you to 2 registers, which cripples performance - every operation involves some stack spills. This is an x86-specific limitation. On ARM, for example, you have 16 32-bit registers, which can be viewed as 8 64-bit registers for certain instructions. Doing a lot of 64-bit arithmetic on an ARM chip still doesn't generate as much register pressure as even doing 32-bit operations on x86.

The final limitation is memory. ZFS likes to have 600MB or so of kernel memory. On x86, the divide between kernel and userspace memory is typically done using segmentation. The kernel has one segment, marked in the GDT as requiring ring-0 permission to access. When you switch to kernel space, the segment register points to this entry. In userspace, you use other segments (sometimes just one per process, sometimes one for stack, one for heap, and so on, sometimes one for all processes with some churn between them). With other implementations, this is done at the page level, although that's more expensive. The kernel's memory, however, is always mapped into the userspace process's address space - it just isn't always accessible.

The reason for this is that x86 lacks sensible TLB controls. If the kernel's address space were not mapped in this way, then every system call would require a TLB flush, which would impact performance. The more address space that you allocate to the kernel, the less you give to userspace apps. If the kernel has 2GB of address space, userland apps can only have 2GB each. On ARM, each TLB entry is tagged with an ASID. The kernel and userspace programs' address spaces are entirely separate, but transitions between the two don't require a TLB flush because the userspace process can't see entries tagged with the kernel's ASID.

Rather than saying that ZFS requires 64-bit, or requires x86-64, it's more accurate to say that it won't work (well) on x86-32 due to inherent limitations of the platform. That doesn't mean that it won't work well on other 32-bit or 64-bit architectures which are less braindead.

Re:Good Article

[TheRaven64]

Perhaps you could describe what exactly it is about apt that makes it so much better than ports? I've not used apt much, but everything that I did with it had a direct equivalent command with portupgrade or one of the other pkg_* tools.

Re:Good Article

[David Gerard]

Watching the ports tree eat itself when I tried to upgrade Gaim in 2004 (when that was a classic example of Linux weenie software) was quite instructive as to ports' failure modes.

After that I installed Ubuntu on a laptop and found the ease of use of Synaptic in the GUI and apt-get on the command line haunting compared to faffing about with ports.

Not the question you asked, bu related: it's a lot easier to get current software into an apt repository than it appears to be in ports. The amount the FreeBSD ports' maintainers mess around with the software is frequently ridiculous. (Nearly as much as Debian's do, to be fair.)

Don't let me stop you using FreeBSD and ports all you like, of course. The important thing to remember is that all software sucks and this is really a matter of opinion.

Re:Good Article

[Just Some Guy]

Raises a hand, waving.

Re:Good Article

[Ltap]

I'd argue that there's nothing special about apt, although unless you want to write a pile of aliases for each of the options (apt-get upgrade install ...) it can get annoying to type out, and the Ubuntu repos always seemed to have very idiosyncratic names for their packages. In this regard, I prefer Arch Linux's 'pacman' and the Arch repos, just for simplicity and convenience. Although I agree, apt is fine as a basic package manager.

Re:Good Article

[drolli]

i agree about the architecture. C code which can only works on a single architecture points to the programmers having no clue. Usually that is a pretty strong indication not to use the code.

Re:Good Article

[David Gerard]

Actually, the main thing is the concept of a well-maintained repository. Debian's is of superlative quality, Ubuntu inherits that. It's not really apt itself. (Fink uses apt and its repo is really badly maintained; Fedora has a well-maintained repo meaning yum is highly usable in practice. YMMV, all software sucks.)

Re:Good Article

[inode_buddha]

Nice numbers. I haven't seen anything like that in a *long* time. ISTR the last time I saw something like that was on slackware v.3. However I wonder what would happen with more users? It's something I would like to experiment with, however I'm very wary and cautious with Oracle's current style. Oracle (nee Sun) is the only one who could repair this relationship. Thoughts?

Re:Good Article

[inode_buddha]

Still, though: how often do you actually need to FSCK? In my situation, not so much anymore. Granted though that I'm a home linux user with no servers whatsoever. UFS might actually be a good idea here: I find that hardware is the more limiting factor nowdays. Even with spinning rust as opposed to SSD.

Re:Good Article

[noidentity]

This is just like people who think flash working only on x86 32bit linux is good enough.

I kind of agree with them. It'd be even better if it ran on nothing.

Re:Good Article

[cthulhu11]

Um, maybe because ZFS is not restricted to x86/x64? What gave anyone the idea that it is?

Re:Good Article

[JamesTRexx]

I've been running Ubuntu for a couple of years now because like you I got a new laptop with some hardware not supported by FreeBSD.
In that time I've seen apt do just as much damage as ports could do. Nothing that couldn't be fixed manually, but I wouldn't call it ridiculously good (unless the definition for "good" has a wide range).
If anything I prefer ports because I can still get XMMS out of it (there just isn't a better player/interface) while the Ubuntu repository has dumped it a long time ago.

Re:Good Article

[David Gerard]

YMMV. I have a habit of running Ubuntu alphas, which I submit gives one solid Debian sysadmin experience fixing broken systems ;-)

XMMS is so ... ten years ago. You know the thing that finally put me off XMMS? That it wrote absolute pathnames in .m3u files. AAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

Re:Good Article

[bill_mcgonigle]

bill_mcgonigle -- what you're doing here is the very definition of FUD.

It's identifying a risk, from a known aggressive litigator, for the kinds of things said litigator has said they'll litigate. Does identifying any risk result in FUD?

Re:Good Article

[bill_mcgonigle]

I have a machine running FreeBSD 8.1 which has been up a couple of months with a zfs filesystem

hey, great news. Thank you for the feedback.

is there anything in particular which could make it crash

I don't think 'zfstest' has been ported to anything (from Solaris) but zfsstress runs on Linux:

ftp://ftp.tummy.com/pub/tummy/zfsstress/

so probably getting it going under BSD has been done or would be straightforward. I haven't tried it myself yet.

Re:Good Article

[Anonymous Coward]

good but linux is too complex i think . all the codes commands harder than windows. now i use android os its very useful.

http://www.izmirotokirala.com

Re:Good Article

[badkarmadayaccount]

CFLAGS, dependency masks, version locks - what is the apt equiv? (Just asking - I'm a fedora person, myself)

Re:Good Article

[David Gerard]

No idea! Never had to unfuck a broken RH/Fedora this decade ...

Re:Good Article

[trasz]

NetApp did sue Sun. Their patents are on their way to being thrown out in the court, FYI.

Re:Good Article

[trasz]

Oracle can't sue, because CDDL license guarantees the right to use the patented techniques that come with CDDL-ed source code.

Re:Good Article

[CAIMLAS]

Most of that can be set in one way or another:

* CFLAGS - quite easy to do with apt-src. There is rarely a point to do so.
* dependency masks? I assume you're referring to portage-like masking? This isn't even close to an issue: debian defaults to sanity, IE "pick the stable crap, not development crap which isn't vetted". This is done with release repositories classified 'stable', 'unstable', 'testing' (also permanently named 'sid'), and 'experimental'. IMO, lacking this basic (default) functionality is one of the biggest faults of pre-portage ports derivatives. (I just never had a term for it before now; didn't realize that's what it was called.)
* /etc/apt/preferences.d - `man apt_preferences` - for tagging/version locks.

Likewise, debian makes it trivial to maintain your own (local) repository. That's a huge boon vs. the complexity of doing that for other systems.

Re:Good Article

[badkarmadayaccount]

What if you want to mix packages from different repos, and have cross-repository dependencies? Does apt-src have persistent global CFLAGs with per package overrides, again persistent? AFAIK Portage can also handle binary packages, where appropriate.

Re:who cares?!

[h4rr4r]

BTRFS will end ZFS if Oracle does not kill that too.

Using SSD as an HD cache / ZFS L2ARC

[Anonymous Coward]

I hope this idea http://blogs.sun.com/brendan/entry/test also works with the port (and will be working with btrfs one day!?):

Transparently adding an SSD into the slow-HD-to-fast-CPU-register cache hierarchy. Are there ways to make something like this work with current FSes?

Re:Using SSD as an HD cache / ZFS L2ARC

[h4rr4r]

Mount a nice fast SSD as swap would be a good start. Then point your /tmp to a tmpfs.

Re:Using SSD as an HD cache / ZFS L2ARC

[Anonymous Coward]

I hope this idea http://blogs.sun.com/brendan/entry/test also works with the port (and will be working with btrfs one day!?):

Transparently adding an SSD into the slow-HD-to-fast-CPU-register cache hierarchy. Are there ways to make something like this work with current FSes?

Mount a nice fast SSD as swap would be a good start. Then point your /tmp to a tmpfs.

My swap is practically always empty, so that would not help much. And how much stuff gets written into /tmp? Mine is 839kiB right now, that is neglible and probably in RAM cache already. What I want cached are all the binaries and libs linked from there. Then starting any app for the second time would be near-instant and this cache would survive reboots and be one or two orders of magnitude larger than RAM.

Re:Using SSD as an HD cache / ZFS L2ARC

[h4rr4r]

I would honestly then say mount it as /. Seems the easiest way to do something similar to what you want. Not exactly the same, but with 32GB SSDs being very affordable not a huge hurdle either.

Re:Using SSD as an HD cache / ZFS L2ARC

[drsmithy]

Mount a nice fast SSD as swap would be a good start. Then point your /tmp to a tmpfs.

That's not even remotely close to the same thing.

Re:Using SSD as an HD cache / ZFS L2ARC

[foniksonik]

I seriously doubt that mounting /. As your filesyatem will help in any way. You'd end up just reading and commenting on each post and forget TFA altogether.

Hey if Phoronix says it, it has to be true!

[Beelzebud]

I hear that every install of ZFS for Linux comes with a pre-installed Steam client, and a free copy of Team Fortress 2 For Linux!

Re:Hey if Phoronix says it, it has to be true!

[TheGratefulNet]

...or the yahoo toolbar.

(why is oracle pimping the yahoo toolbar, btw, on their java installs? boggle!)

Re:Hey if Phoronix says it, it has to be true!

[binarylarry]

Sun started that BS.

Re:Hey if Phoronix says it, it has to be true!

[h4rr4r]

They get paid to include that. Just like Microsoft is now paying Verizon to add bing search to phones, and NASCAR pays Sprint to include their apps.

Re:Hey if Phoronix says it, it has to be true!

[CoolGopher]

I came across this job post at Valve just recently. To save you from having to follow the link, it includes the item
"Port Windows-based games to the Linux platform".

Just sayin'.

Re:Hey if Phoronix says it, it has to be true!

[Beelzebud]

You do realize that Valve has Linux server clients for all of their games, and the Steam infastructure itself runs on Linux right?

Just sayin'.

Re:Hey if Phoronix says it, it has to be true!

[MrHanky]

Linux server sounds like a fun game.

Re:Hey if Phoronix says it, it has to be true!

[dreamchaser]

You do realize that the job posting is about porting games, not server clients or internal server applications, right?

Re:who cares?!

[EvanED]

ZFS has becoming vapor ware since apple announced snow kitty wasnt gunna support it.

I do not think that word means what you think it means.

New Project

[bragr]

Looks like I'll finally have a reason to get around to rebuilding my home server. Well, as soon as WD finishes warrantying one of the drives in the array that is. Current estimates put ZFS and the drive arriving at the same time!

If it comes out and works well

[Sycraft-fu]

Seems a little early to be putting faith in that. It's feature list looks good, on par with other modern desktop file systems like HFS+ and NTFS. However it is currently unstable. When will that be fixed? Who knows? Maybe it moved full steam ahead and we have a stable, capable file system next month. Maybe the project loses steam and languishes and 4 years from now it is still "unstable" and "coming soon."

You can't really say how well it'll work until there is stable code to test. Remember designing a file system isn't the real hard part. I'm not saying it is trivial work or that it is unimportant but it is by far the easier part of all this. You can write out a specification that sounds great on paper, but then you have to implement it. That is the much harder part. You have to make it fast, stable, not corrupt data, able to do everything it should and so on.

This is part of the reason why NTFS on Linux has been so tricky. It is actually pretty well documented in the Windows Internals book, and other places, but it is a complex file system. FAT, on the other hand, is real simple and thus not hard to implement.

As an example you can look at driver sized. The NTFS driver in Windows is 1.6MB. The FAT driver, on the other hand which supports multiple versions of FAT, is only 200k. The NTFS kernel driver is one of the very largest in the system, only the ATi video driver (much larger) and TCP/IP stack (a bit larger) are bigger than it on my system.

So we'll see what happens with btrfs. As of late, there's not been much activity. The last version update was June 2009. Maybe they are rolling up final testing for production release, or maybe things have slowed down and release is not near. We'll just have to wait and see, but it is foolish to believe this will be the Next Big Thing(tm) at this point.

Re:If it comes out and works well

[UnknownSoldier]

Not that I doubt you, but have any references to verify that claim please?

Re:If it comes out and works well

[Sycraft-fu]

If you think LVM + ext3 is better than NTFS sorry that speaks only to ignorance and possible Linux zealotry, not to knowledge. I don't care if you like MS, their file system is high class. It offers good performance and a very wide feature set, and they update it all the time. They don't rename it, but NTFS has been improved with each version of Windows.

No shame in not knowing about file systems, it is rather esoteric, but then please don't go shooting off at the mouth about how $your_chosen_platform has the bestest FS and everything else sucks.

Linux actually is a good deal behind in the FS game, which is one of the reasons there's so much interest in btrfs or ZFS. Linux could use a more modern file system for many tasks.

Finally Btrfs doesn't "Blow NTFS out of the water," because btrfs cannot be used on production systems. NTFS is used on basically every production Windows system, desktop or server, in the world. Btrfs is still under development. It may be a far superior file system when it comes out, but it isn't out yet. You can't set it up and use it and expect all your data to be ok.

There's a big difference between something under development and something in production. IE9 doesn't "Blow Firefox 3 out of the water," either. In theory it will, I've tried the test builds and its rendering is amazingly fast because of the hardware acceleration. However, it isn't a stable, release browser. It is just a test. How it does in realty when it is released, and equally important how FF is doing at the same time, remain to be seen. Making any claims of it at this point would be extremely premature.

Re:If it comes out and works well

[Christophotron]

BTRFS is not that unstable really.. I have been running for a few months now, since the on-disk file structure was finalized. it's in a raid 1 configuration across 2 300gig drives on one of my home servers and it hasn't had a hiccup yet, even with lots of file i/o. i think it would like more than the CPU and RAM I gave it, but its still less resource intensive than ZFS. AFAIK ZFS would not even run on that machine due to the 32 bit processor and only 512mb of RAM. Some of the features are not implemented yet but it is certainly stable enough to test..

Re:If it comes out and works well

[Anonymous Coward]

FAT32 is better than NTFS.

Ever try to use an NTFS-formatted flash drive in anything that's not a PC?

QED

Re:If it comes out and works well

[h4rr4r]

When did NTFS get snapshotting? A rather simple feature.

Re:If it comes out and works well

[h4rr4r]

When did it start letting people do the equivalent of "rm /bin/rm"?

Re:If it comes out and works well

[EvanED]

NTFS doesn't do COW, but it's had snapshotting for a while under the name "volume shadow copy". This was added in XP or 2003, and even given somewhat of a UI in the form of "previous versions" in Vista.

Re:If it comes out and works well

[h4rr4r]

So you are suggesting I can freeze IO to the machine, then run a snapshot command on NTFS?

I would be glad to hear it.

Re:If it comes out and works well

[EvanED]

NTFS doesn't do COW

I think I misspoke here; volume shadow copy can do COW.

What NTFS doesn't do is the ZFS "always allocate a new block when writing" thing, and that's what I was thinking of when I wrote that.

Re:If it comes out and works well

[Sycraft-fu]

Well when it comes to filesystems, I wouldn't want to run one on any production machine until the developers say they are stable. Bad FS driver can equal file system corruption and data loss.

The problem with testing early release stuff is that it is not necessarily representative of the final product. It can, of course, be too slow because of lack of optimization. However it can also be too fast. What do I mean by that? I mean perhaps during implementation, they discover that more extensive checks and processing of certain kinds is necessary to maintain data integrity. This ends up slowing things down.

I'm not saying that will happen, just saying it can, hence I don't like tests until the final version is there and everything is in it.

Also the other problem I have is that you don't know if something will ever reach completion. I've seen OSS projects that rocketed to a stable release in record time, I've seen them proceed slow and steady to a release, and I've seen them peter out and stagnate. You really don't know how it is going to be until it comes out.

I remember when at work we were trying to decide on an e-mail client to recommend to users (I work in an academic environment so we can't mandate things). None of us were that happy with Outlook for IMAP (though in 2010 it has gotten a lot better). I was using Thunderbird and found it to be acceptable. Since it was free and worked, I said lets do that. My boss wanted to wait for another one, a project he'd been watching for some time (can't remember the name). Was going to be awesome when it released... Ya well that day never came. They kept talking and talking, but no releasing of final code. Maybe some day it will be awesome, but right now Thunderbird is 100% more awesome since it is out, running and useful.

I have high hopes for btrfs because our large, inexpensive, non-backed up storage (our backed up storage is NetApp) at work is Linux and it could use something better than ext3. However I won't say "Yes this is a good FS," until it is a released product and I've seen how the final thing works.

Re:If it comes out and works well

[Cyberax]

"Seems a little early to be putting faith in that. It's feature list looks good, on par with other modern desktop file systems like HFS+ and NTFS."

LOLWUT? NTFS is about as far from 'modern' file systems as ext2/ext3.

NTFS has a lot of problems with performance (like a tendency to fragment seemingly while the HDD is still en-route in plastic package). It doesn't support deduplication, symbolic links to files (yep, it's done above the VFS layer in Windows Vista), no support for RAIDs, no support for dynamic resizing, etc. It's also SLOW. NTFS is an old filesystem, it was conceived in the beginning of 90-s when journaling was the state of the art. But now the state of the art has moved far far away.

NTFS driver is so large because the internal structures of NTFS are fairly complex. NTFS kernel driver is also not the largest one, XFS is by far larger. ext4 is comparable.

Re:If it comes out and works well

[EvanED]

No need to actually freeze I/O; the volume shadow service will ensure that the snapshot is taken at a consistent time.

I've not actually used this feature, so I spent a bit of time just now looking around for how to actually activate it. It sounds like it's easy if you have a server edition of Windows ("vssadmin create shadow /for=c:"), but I'm not sure how you can trigger a copy manually on non-servers. It's possible that creating a system restore point does exactly what you want, but I'm not sure. You can tell it to make snapshots automatically (for use with the "previous versions" feature), but I don't see a way to specify the interval, just stuff like the maximum amount of space to use.

Re:If it comes out and works well

[jedidiah]

> If you think LVM + ext3 is better than NTFS sorry that speaks only to ignorance and possible Linux zealotry, not to knowledge.

Too bad you spent all of that space and didn't actually come up with any real reasons to justify this position.

Re:If it comes out and works well

[shutdown -p now]

The whole point of snapshots is that you don't freeze the IO. The snapshot service provides you a, well, snapshot of how things were at the moment it was requested, and maintains that snapshot even as other applications keep writing data. It's roughly similar to MVCC, only the units are FS blocks, not database records.

Re:If it comes out and works well

[tepples]

Another Slashdot user recommends against using "citation needed" in a Slashdot post. Apparently, Slashdot users are supposed to first list what they tried in Google. Something like "I tried ext3 vs ntfs in Google but couldn't tell what was reliable" might help.

Re:If it comes out and works well

[value_added]

They don't rename it, but NTFS has been improved with each version of Windows ... No shame in not knowing about file systems, it is rather esoteric, but then please don't go shooting off at the mouth about how $your_chosen_platform has the bestest FS and everything else sucks.

1. Sorry, mate, NTFS does indeed have unique version numbers.

2. Like most Micorosft apologists, you're spending considerable time and effort going on about "features", an approach that's more appropriate to a marketing brochure than a technical discussion. So while NTFS may have lots of "features", so do most file systems. What matters is the implementation, and how that translates to the real world. On that count, few people are as impressed as you seem to be.

Re:If it comes out and works well

[tepples]

Parent post supports only the assertion "NTFS is better on PCs; FAT32 is better on all other devices." If you'd believe the Windows XP/Vista/7 formatter, FAT32 is aptly named: Microsoft believes that compatibility with non-PC devices isn't worth the performance drop on volumes bigger than 32 GB.

Re:If it comes out and works well

[benjymouse]

So you are suggesting I can freeze IO to the machine, then run a snapshot command on NTFS?

I would be glad to hear it.

The Volume Shadow Service (VSS) is always running (by default). Backup utilities - including the ones which come with Windows - use VSS to create a snapshot and perform backup from that point in time. It doesn't freeze IO; rather it goes to copy-on-write.

On server versions you can also create snapshots interactively by using the vssadmin tool.

Shares can be set up to create a shadow copies multiple times per day. This is not copy on every write - but it *is* copy on write once a block is part of a snapshot. Any client (plugin needed for XP, IIRC) can display previous versions which are available snapshots.

VSS actually goes beyond NTFS integration (which is probably why it is a service and not just a NTFS feature). Certain applications - e.g. Exchange, SQL Server and Hyper-V - also integrate with VSS. Instead of VSS operating directly on e.g. SQL Server files, it integrates with the server to create a snapshot for the database files. During restore the system knows how some applications took part in the shadow copy. This ensures that I can correctly restore *all* the files needed to bring a SQL server database back to a certain point-in-time. It also allows the SQL server to prune the log automatically.

I have a Server2008R2 which has several Hyper-V images (development and testing). When I perform a backup of the server, VSS interacts with Hyper-V to perform backup of the virtual machines as well. A Server2003 which hasn't been set up to support VSS is actually "hibernated" by Hyper-V/VSS - then backed up - then brought back into running state. That could be considered "freezing IO", I suppose.

Re:If it comes out and works well

[EvanED]

...symbolic links to files (yep, it's done above the VFS layer in Windows Vista), no support for RAIDs...

To be fair, if these are done in another layer (as they are, if we take your word for it), then you can hardly count it as a strike against the file system.

I'd definitely like to see Windows get a ZFS-like file system, and I do think NTFS is comparable to ext3 etc. (Each has a benefit or two over the others, e.g. ext3 doesn't store a file creation time and Windows has Transactional NTFS, but at least NTFS on Windows has some obnoxious file naming limits and locking problems.) That said, NTFS beat the pants off of the file systems available in 2.4.

Re:If it comes out and works well

[Anonymous Coward]

Start
Run
cmd
vssadmin

Re:If it comes out and works well

[Christophotron]

I agree. BTRFS is definitely not ready for production or for storage of anything important that is not backed up elsewhere. It has known bugs, like for example the reported free space on a raid 1 will show the total disk size and not the actual free space, so it may be dangerous to fill the array too close to 100% (shown as 50% in df). It is unclear when (or if) it will be ready, but it is being worked on -- I've seen updates for the userland tools in Debian testing, and the newer kernels have updates for the fs driver. The bug I mentioned is fixed in 2.6.33, I believe. I was only countering the argument that it is too unstable even to test it out. That is untrue. Heck, even Linus Torvalds reportedly uses BTRFS as the root filesystem on one of his laptops.

Re:If it comes out and works well

[Anonymous Coward]

http://technet.microsoft.com/en-us/library/cc754968%28WS.10%29.aspx - vssadmin has existed since Windows XP

Re:If it comes out and works well

[mysidia]

I/O is only frozen for 10 seconds. I'm sure Microsoft has an article about this

Yes, one method shadow copy uses is COW. Shadow copy is not NTFS.

It is a volume-level mechanism that lies under the NTFS file system. There is an area for files, and there is a volume shadow copy region on each volume.

There are two methods shadow copy can be implemented, one involves re-directing writes to the reserved region. Copy on write is the other option.

Re:If it comes out and works well

[mysidia]

You ready to try it on a live Oracle database server with a 10TB filesystem that averages at least say 5,000 DB transactions per second? :)

Re:If it comes out and works well

[Anonymous Coward]

Around Windows XP/2003, backported to 2000. It's called VSS

Re:If it comes out and works well

[Galactic Dominator]

A long time ago.

http://axelilly.wordpress.com/2009/01/15/creating-vss-snapshots-on-windows-2003-server/
http://en.wikipedia.org/wiki/Shadow_Copy

FreeBSD had it around '05, Windows 2003 introduced COW snapshots. The new Win 7 and Server 2008 offer diskshadow, a decent cli interface to manage them.

EvanED is apparently not familar with VSS. I don't know of UI to VSS which is not 3rd party one.

If you've used any of the COW snapshot capable FS's, you know why LVM snapshots are inadequate.

Re:If it comes out and works well

[afidel]

Other than embedded devices that will only talk FAT/FAT32/exFAT, the best cross platform filesystem is actually UDF, on all modern OS's it's supported including write. It supports greater than 2GB files and large disks (2TB with 512B clusters, 8TB with 2K clusters though the current standard doesn't list 4KB clusters which is an issue for the newest generation of HDD's).

Re:If it comes out and works well

[afidel]

Better, you can ask all VSS aware apps to freeze their own I/O to make a crash consistent backup. This is how all non-brain damaged backup software works on Windows and how array based snapshots for all but the cheapest of disk arrays works.

Re:If it comes out and works well

[afidel]

There's also hardware VSS providers which are free to do things as they please wrt how they achieve the snapshot.

Re:If it comes out and works well

[afidel]

I'm very impressed with NTFS, the last time I lost data on an NTFS volume that wasn't caused by a hardware fault was back in the NT4 days. It handles non graceful shutdowns better than just about any other FS I've worked with.

Re:If it comes out and works well

[afidel]

*deduplication, correct
*symbolic links to files, incorrect NTFS has supported reparse points since Windows 2000
*no support for RAIDs, incorrect Server versions support RAID 0, 1, and 5
*no support for dynamic resizing, incorrect Windows 2003 added support for dynamic growth for non-system/non-boot volumes, 2008 added dynamic grow and shrink for all volumes.
It also supports compression, encryption, ACL's, Metadata, and ridiculously large volumes and files.

Re:If it comes out and works well

[Beelzebud]

So you're suggesting that drives don't get fragmentation on Linux? Speaking of raving idiot...

Re:If it comes out and works well

[Anonymous Coward]

there is a difference between the philosophy of how ZFS and BTRFS were architected.

ZFS has notoriously been called a "rampant layering violation", but IMHO that makes it easier to use - since it eliminates one whole layer.

BTRFS is not similarly architected: it is not an apples-for-apples replacement for ZFS.

Re:If it comes out and works well

[EvanED]

*symbolic links to files, incorrect NTFS has supported reparse points since Windows 2000

Reparse points can only occur at directories, which means that NTFS did not support symlinks to files until Vista.

Re:If it comes out and works well

[Anonymous Coward]

> Wow. What a rant from a raving idiot. Last I checked, windows users were still having to 'defrag the drive'. WHAT THE FUCK! That problem was solved in 1967.

I stopped reading after that.

Re:If it comes out and works well

[ion.simon.c]

*blink*

I spent the past five years working in a small (~25 developer) Windows XP Pro based software development house. Our file server was running Win2K3 Enterprise and was using a large (1TB, later upgraded to 10TB) hardware RAID 5 disk array. All of this equipment was sourced from Dell.

Once a month, we needed to call in an admin to bring down the server (and once every other month for someone's desktop machine) to delete files that were "screwed up". "Screwed up" means:
* Cannot delete, rename, read, or modify the file.
* The only tab available on the "File->Properties" dialog for the file in question is the "General" tab. (This means that the Sharing, Security, and Customize tabs shown here are not present.)

Note that *every* developer performed work as an unprivileged user. Noone on staff possessed an Administrator account, with the exception of the admins.

I've never *ever* seen this behaviour with *any* filesystem on Linux. I've abruptly pulled the plug on my home machines hundreds of times and never had *any* filesystem issues. (Not even with reiserfs V3. :D)

Re:If it comes out and works well

[drsmithy]

NTFS has a lot of problems with performance (like a tendency to fragment seemingly while the HDD is still en-route in plastic package). It doesn't support deduplication, symbolic links to files (yep, it's done above the VFS layer in Windows Vista), no support for RAIDs, no support for dynamic resizing, etc. It's also SLOW. NTFS is an old filesystem, it was conceived in the beginning of 90-s when journaling was the state of the art. But now the state of the art has moved far far away.

These are pretty much all wrong and/or irrelevant.

* *Actual* performance problems due to fragmentation - outside of a few corner cases - are basically nonexistant.
* Deduplication outside of SAN and NAS-level appliances is basically nonexistant, so saying that makes it obselete seems ridiculous.
* Can you explain what you mean by "it's done above the VFS layer" ? Surely you're not trying to argue symlinks and shortcuts are the same thing ?
* RAID is handled at the block device level, not the filesystem level (and many, many people believe putting RAID into the "filesystem" is an architecturally bad thing, so that's hardly something it can be plainly criticised for).
* NTFS filesystems can both grow and shrink.
* Do you have a source for up-to-date benchmarks ?

NTFS is an old filesystem, it was conceived in the beginning of 90-s when journaling was the state of the art. But now the state of the art has moved far far away.

NTFS today is not the same NTFS that was in Windows NT 3.1.

Re:If it comes out and works well

[Cyberax]

"*symbolic links to files, incorrect NTFS has supported reparse points since Windows 2000"

Incorrect. Reparse points apply only to directories, not files.

"*no support for RAIDs, incorrect Server versions support RAID 0, 1, and 5"

On block level. No filesystem support, like in BTRFS or ZFS.

"*no support for dynamic resizing, incorrect Windows 2003 added support for dynamic growth for non-system/non-boot volumes, 2008 added dynamic grow and shrink for all volumes."

Only for 'dynamic' disks which are undocumented and shrinking also doesn't always work.

"It also supports compression, encryption, ACL's, Metadata, and ridiculously large volumes and files."

Linux filesystems support compression (btrfs), ACLs (POSIX, SELinux), metadata (extended attributes) and ridiculously large volumes.

Re:If it comes out and works well

[Cyberax]

"* *Actual* performance problems due to fragmentation - outside of a few corner cases - are basically nonexistant. "

Yep. That's why I have to run defragmenter on our build server every week...

Also, Windows is notoriously slow with file operations. It's not directly related to NTFS, but more to extremely inefficient VFS stack.

"* Can you explain what you mean by "it's done above the VFS layer" ? Surely you're not trying to argue symlinks and shortcuts are the same thing ? "

http://neosmart.net/blog/2006/vista-symlinks-revisited/

"* RAID is handled at the block device level, not the filesystem level (and many, many people believe putting RAID into the "filesystem" is an architecturally bad thing, so that's hardly something it can be plainly criticised for)."

However, filesystem-level RAIDs have a lot more functionality than block-level RAIDs. Look at ZFS or BTRFS.

"* Do you have a source for up-to-date benchmarks ?"

I have my own set of benchmarks. Well, NTFS on Windows is almost always slower (and quite often like 100 _times_ slower) than Linux filesystems.

http://rsdn.ru/File/37054/benchmark.zip - this is the source.

http://rsdn.ru/forum/philosophy/1710544.1.aspx - this is a post with benchmark results (in Russian, sorry - I can translate if you have any questions)

http://rsdn.ru/forum/philosophy/1712431.aspx - this post contains this benchmark, slightly adapted.

I regularly re-run these tests. So far, Windows is only getting slower compared to Linux.

I've recently created a multithreaded version of this test. Well, let's say that NTFS sucks so badly, that it's hard to understand how MS has managed to achieve this.

Re:If it comes out and works well

[FauxPasIII]

It's actually closer than it looks like from their website, per some recent activity on the mailing list:

http://www.mail-archive.com/linux-btrfs@vger.kernel.org/msg05749.html

      While Btrfs is stable on a stable machine, it is currently possible
      to corrupt a filesystem irrecoverably if your machine crashes or
      loses power. This will be fixed when the fsck tool is ready.

And
http://www.mail-archive.com/linux-btrfs@vger.kernel.org/msg05882.html

      We're still actively developing (btrfsck). I don't have a release date planned
      yet but we should have betas coming out over the next few months.

Note both threads are from the last two weeks. Things are still happening rapidly; just not, for whatever reason, on the project web site.

Re:If it comes out and works well

[Sir_Lewk]

When did preventing that become the job of the FS? File permissions work just fine in any linux FS I've tried. Besides that what more do you expect to be done?

Re:If it comes out and works well

[ToasterMonkey]

LOL.. you were moderated funny because readers had to scroll past abut twenty informative replies about NTFS & VSS to find your little gem.

Re:If it comes out and works well

[Anonymous Coward]

mod this up informative

Re:If it comes out and works well

[stiller]

Note that Btrfs does not yet have a fsck tool that can fix errors. While Btrfs is stable on a stable machine, it is currently possible to corrupt a filesystem irrecoverably if your machine crashes or loses power on disks that don't handle flush requests correctly. This will be fixed when the fsck tool is ready.

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

Re:If it comes out and works well

[TeknoHog]

BTRFS is not that unstable really.. I have been running for a few months now, since the on-disk file structure was finalized.

Which kernel are you running? 2.6.35.4 is the latest stable version, and menuconfig says that Btrfs has an unstable disk format.

Re:If it comes out and works well

[Anonymous Coward]

Yeah, "production ready" is situation dependent: MeeGo uses btrfs already and it's definitely ready for that sort of use.

Re:If it comes out and works well

[g4r0]

There are still some big issues...

Try this for example:
1) take 2 usb drives and plug them in, they will become /dev/sda and /dev/sdb
2) create a btrfs partitition on each one so you'll have /dev/sda1 and /dev/sdb1
3) plug them out and plug them back in the reverse order so /dev/sda1 will become /dev/sdb1 and /dev/sdb1 is now /dev/sda1
4) try to mount them, it will be impossible

You can only mount btrfs partitions if they keep their original name

Re:If it comes out and works well

[truedfx]

Note that on 32-bit systems, btrfs will only work well until your inode numbers no longer fit in a 32-bit int, which unlike with other file systems can easily happen. stat() will return an error code, and critical things like the ELF loader will no longer work.

Yes, I've seen this happen.

Re:If it comes out and works well

[Anonymous Coward]

You're talking complete bullshit. NTFS performance is terrible. NTFS is a pile of featureson top of a hack extension of a bad operating system, with bad performance as a result.

Re:If it comes out and works well

[Harik]

He asked when you could do the equivilant of rm /bin/rm - which you can't in windows. Although I'm not entirely sure that's a filesystem thing, it's a VFS layer thing where an open is an automatic exclusive read lock.

Which is why you get the ridiculous volume shadow copy bullshit and forced reboots to update anything.

Re:If it comes out and works well

[EvanED]

Which is why you get the ridiculous volume shadow copy bullshit

Volume shadow copy is somewhat independent of the locking BS Windows gives you. The locking makes VSS more necessary (okay, essential), but it brings a lot of benefits even if you don't have locking.

Re:If it comes out and works well

[Christophotron]

From the btrfs wiki:

Btrfs is under heavy development, but every effort is being made to keep the filesystem stable and fast. As of 2.6.31, we only plan to make forward compatible disk format changes, and many users have been experimenting with Btrfs on their systems with good results. Please email the Btrfs mailing list if you have any problems or questions while using Btrfs.

OK, I guess the word "finalized" was a bit premature. It's "forwards-compatible", lol. I run 2.6.32 btw.

Re:If it comes out and works well

[Anonymous Coward]

Realtek's AC97 driver is about 70MB. HP's printer driver is close to 100MB.

Re:If it comes out and works well

[Anonymous Coward]

"*symbolic links to files, incorrect NTFS has supported reparse points since Windows 2000"

Incorrect. Reparse points apply only to directories, not files.

Yes and no; there's some mixed up terminology there. You're thinking of junctions, which are used to implement a sort of symbolic link for directories in Windows 2000 and later. Windows Vista and Server 2008 introduced support for symbolic links to both directories and files. (The directory version has different semantics from junctions in certain situations.)

All of these link types are implemented using reparse points, a generic mechanism of redirecting filesystem object operations that has been present in the NTFS spec since Windows 2000.

Re:If it comes out and works well

[Anonymous Coward]

NTFS... High class? That must explain why when I run cfdisk /f on many of my XP machines, it claims there are errors and then that it fixed them. Then when I reboot and run it again, it claims the errors are still there and that it "fixed them" again.

Re:If it comes out and works well

[afidel]

You could actually do symbolic links in 2000, they were just very hacky and not supported by all OS methods. It's because NTFS had reparse points that this worked.

Re:If it comes out and works well

[LWATCDR]

Why not use XFS or JFS if you need a better file system than EXT3? XFS has good performance on big file systems.

ZFS recap

[oldhack]

We've heard much about ZFS, but being a slashdotter, I can't recklessly go on and RTFA. So, maybe someone here can recap its main benefits. Maybe a power point slide?

Re:ZFS recap

[colinrichardday]

But we're all slashdotters here, so who can read the article to do the recap? I believe this is a belling-the-cat problem.

Re:ZFS recap

[Miseph]

Perhaps you'd like a colonel to read it for you?

Re:ZFS recap

[Anonymous Coward]

We've heard much about ZFS, but being a slashdotter, I can't recklessly go on and RTFA. So, maybe someone here can recap its main benefits. Maybe a power point slide?

Here's a good PDF on it:

        http://hub.opensolaris.org/bin/download/Community+Group+zfs/docs/zfslast.pdf

Here's the PDF being presented by the co-creators, Jeff Bonwick and Bill Moore:

        http://blogs.sun.com/video/entry/zfs_the_last_word_in

Three parts, one hour each. Streamable blip.tv as well as a downloadable M4V file.

Two, ten minute videos:

        http://www.youtube.com/watch?v=gthel59G56c
        http://www.youtube.com/watch?v=OdHUub462pM

Though I recommend you set aside the three hours (even if it's over several days) to really get a good understanding of how things work.

Re:ZFS recap

[mysidia]

OpenSolaris ZFS Screencast: 100 Mirrored Filesystems in 5 minutes

OpenSolaris Screencast: ZFS Self Healing

Oracle Demo: ZFS Discovery Day

Presentation: (PDF) ZFS: The Last Word in File Systems (2008)

Presentation (VIDEO) ZFS: The Last Word in File Systems" (2008)

Nexenta 3.0 webinar (Slide 18, Deduplication)

Sun ZFS drive destruction Demo

Re:ZFS recap

[oldhack]

Dang, this explains a lot of what's going on around here.

Re:ZFS recap

[ArsonSmith]

ZFS is full of features we all wanted 20 years ago but found other work arounds for 10 years ago that turned out to be as good or better and don't really need it any more.

Re:ZFS recap

[Galactic Dominator]

Keep telling yourself that, it doesn't hurt so much then. I'll just be over with my transparent file system compression and other goodies if you need some more consoling.

Re:ZFS recap

[Anonymous Coward]

We've had that since Stacker. Try again.

Re:ZFS recap

[Mysticalfruit]

Exactly... I've currently got a 27TB raid50 running on OpenSolaris. Add more storage? Snap on another cabinet of disk (live) and add the space to the pool (also live)

As for this being the first native ZFS port to linux... I must be living in a time warp...

http://github.com/behlendorf/zfs/wiki has been ported for a while. All it's missing is the posix layer, which they claim is on the way...

Re:ZFS recap

[RedK]

Exactly... I've currently got a 27TB raid50 running on OpenSolaris. Add more storage? Snap on another cabinet of disk (live) and add the space to the pool (also live)

And what exactly is impressive about that ? We do that all the time with LVM/VxFS on HP-UX (Present storage through FC LUN discovery, pvcreate on the new "disk", vgextend the volume group, lvextend the logical volumes as needed, VxFS supports online resizing, up or down!). Heck, if you want to add in the mirroring capabilities of ZFS, know that LVM also supports that, using the -m switch and setting your VGs on strict allocation so that your extents are always picked from different PVs.

And before that, we did it using VxVM/VxFS on Solaris (Plexes for mirroring, online growth, etc..). You're proving the other short-sighted comment true with your own, there is much in ZFS that we did find solutions for many years ago. In fact, I'm willing to bet ZFS just used the ideas spread forth by these volume managers and just went from there, because ZFS does really look like a volume manager at its core (zpools are basically disk groups (VxVM) or volume groups (LVM)).

Of course it's different. On top of zpools, the ZFS slices themselves look more like Netapp's QTree concept than LVM's LVs or VxVM's volumes. One big volume (the zpool) with different small slices (zfs) that share the underlying storage using a quota based strategy. You can overbook storage so to make sure there is no waste (something VxVM/LVM lacks).

Then there's all that new stuff like snapshots and compression which just aren't there in the older volume managers and of course, RAIDZ, which is not supported at all in LVM unless you use 3rd party tools to create a soft RAID5 or 6 and then pvcreate that device to add to your VG.

Re:ZFS recap

[TheRaven64]

Really? Let's take a look at some ZFS features:

Block-level checksums? I didn't want these 20 years ago, because disks were much coarser then - they either worked or failed completely. Small corruptions are much more common now. This does eliminate some problems with mirroring (i.e. the two disks return different values, or the RAID-5 read fails the checksum, but you can't tell which volume has the error), so maybe some people did want it. What was the work-around that we had 10 years ago? Scattering par2 files all over the disk and manually verifying periodically?

Copy-on-write? A few people at Berkeley wanted that 20 years ago and created LFS, but for most people the fragmentation that it introduced was problematic because it meant that seek times on mechanical disks killed performance. With modern SSDs, or hybrid systems, CoW now gives the best performance because it uses linear writes but sequential seeks, which are the best case for Flash. We did want the undeleting and snapshotting support that CoW gives though, and we hacked in snapshots at the block layer, but that was ugly and the performance didn't scale well. In contrast, ZFS snapshots are O(1).

Top to bottom transactional I/O? We really wanted that, and were willing to pay companies like IBM and Oracle large amounts of money for databases that provided it because our filesystems didn't.

Reliable replication? We had things like RAID-1, but without block-level checksums a mirrored RAID can't tell which disk is returning errors. We had RAID-5, but it suffered from the write hole. We hacked around the write hole with battery-backed RAID controllers, but they were an expensive hardware solution to a software problem. RAID-Z and RAID-Z2 provide the advantages of RAID-5/6 without the limitations.

Remote replication and backup? We hacked something like this together with rsync and snapshots, but it works at the file level so wastes a lot of space if you want to be able to roll back changes. In contrast, zfs send and receive let you stream the filesystem I/O commands to a remote machine or a file, allowing you to replay a set of changes and have a backup server that can be kept in sync with the main system and can easily restore old versions of files.

Re:ZFS recap

[Blakey Rat]

I'll just be over with my transparent file system compression

That's been in NTFS for ages... at least a decade.

and other goodies

If transparent file system compression is the most impressive goodie, then... I'll pass.

Re:ZFS recap

[cblack]

But we're all slashdotters here,...

I'm not!

Lack of professionalism?

[Anonymous Coward]

This sounds great.

However, some of the authors comments in the Phoronix thread (such as "FUSE is crap") -- and the fact that he's announcing this on Phoronix instead of via some other technical channel temper my enthusiasm just a bit...

Re:Lack of professionalism?

[Beelzebud]

Yeah his comments on that forum are extremely childish. Too many "dudes" for me to even take seriously.

Re:Lack of professionalism?

[atari2600]

Chill dude.

Re:Lack of professionalism?

[Anonymous Coward]

Yeah his comments on that forum are extremely childish. Too many "dudes" for me to even take seriously.

Yeah, well, that's just, like, your opinion, man.

Useless without dedupe

[Zlurg]

Just ask freenas and openfiler and, heck, even opensolaris
Oh, and this:
23212 [root@place]/mnt/Scratch: zfs get utf8only
bad property list: invalid property 'utf8only'
Know why? Cuz you don't get any choice in the matter outside Solaris proper.
Without dedupe and without the ability to use non-UTF8 in filenames (try creating a directory called Télépopmusik in a ZFS pool), ZFS is a future-former. I hear they name filesystems after murderers these days just for the street cred.
You did not read it here first, but I'll take the heat.

Re:Useless without dedupe

[Anonymous Coward]

1. Dedup is already part of ZFS - just not in an official release yet. You can get it in the dev builds of OpenSolaris. However, because Oracle has canceled OpenSolaris, you will not likely see an official dedup-enabled ZFS until Solaris 11.

My understanding, though, is that Sun has some storage solutions in production /w ZFS dedup that you can buy today.

Considering that no other open source file system offers this feature, it seems like a stupid criticism.

2. Those characters can be represented in UTF-8 just fine, and I just created a folder with that exact name. No problems.

Re:Useless without dedupe

[chgros]

non-UTF8 in filenames
UTF8 can represent all of Unicode.
I'm not aware of any non-unicode characters.

Re:Useless without dedupe

[TheRaven64]

Note, however, that UTF-8 is optimised for storing the subset of unicode that is common in the US and, to a lesser extent, Europe. For other subsets of unicode, it can be much less space-efficient than UTF-16, requiring three bytes per character instead of two. Of course, the only time this will ever be a problem is if you have a filesystem containing a few billion files with long Japanese filenames and no data. For most of us, the space wasted by UTF-8 in the filename is going to be tiny compared to the space used by the contents of the file.

Re:Useless without dedupe

[corychristison]

Considering that no other open source file system offers this feature, it seems like a stupid criticism.

There are a couple in development although I wouldn't use in a production environment just yet:
1) SDFS - http://code.google.com/p/opendedup/
2) LessFS - http://www.lessfs.com/

I think they are both based on FUSE but I'm not entirely sure, I haven't delved that deep into them I was just poking around one random night.

In Windows 2003

[Sycraft-fu]

It was deployed to desktops, and on by default, in Windows Vista/7. It does copy on write and maintains old snapshots of files automatically. On the server side, there is some more management of this if you like. This snapshotting feature is also used by backup utilities to do hot backups. Ghost and TrueImage can image a running system using it. They can snapshot the state for backup and new data can be committed while they work, without messing with anything. Works great. That is also independent of the maintaining of old versions so you can shut that down if you like and still do snapshots for backups.

Re:In Windows 2003

[shutdown -p now]

Ghost and TrueImage can image a running system using it.

Having worked on True Image, I think I need to qualify that. I don't know about the recent versions (though I doubt it would have changed), but back in the day it was only using VSS API to notify applications that a snapshot for backup purposes is taking place, so that they can bring themselves to consistent state before it happens (this is most useful for database servers). However, actual snapshotting code in True Image is entirely its own in-house developed kernel driver.

Can I remove a disk from it yet?

[Daffy Duck]

http://www.opensolaris.org/jive/thread.jspa?threadID=131604
http://www.opensolaris.org/jive/thread.jspa?messageID=270957

Long story short: disk pools in ZFS can only grow, so don't make any mistakes unless you can afford to do a full dump and restore. Sun had been "working on" this for years. Anyone heard any news lately?

Re:Can I remove a disk from it yet?

[diegocg]

The ZFS design makes this very difficult. Btrfs, on the other hand, has supported this feature for a long time, thanks to a nice design feature called backrefs.

Re:Can I remove a disk from it yet?

[catmistake]

Sun has had this tutorial out for a couple years now.

Re:Can I remove a disk from it yet?

[Cato]

Btrfs backrefs also make it easier to do an fsck (something that ZFS doesn't have) - there's a comment from the btrfs designer in the penultimate paragraph of http://www.osnews.com/story/22423/Should_ZFS_Have_a_fsck_Tool_

not holding my breath, but...

[CAIMLAS]

I'm not holding my breath, but god damn...

If this ever happens, my wildest dreams will have come true. Forget the threesome with now-hideous-and-leathery-old-but-once-hot porn stars from my youth! Not having to deal with any more Solaris or FreeBSD for a 'modern filesystems' would be incredible.

Like I said, not holding my breath (or even breathing heavily!). If it happens it'll happen only at the fringes, and poorly.

I can't wait

[Anonymous Coward]

I don't know about you, but I've gotten nothing but top-notch bug-free, reliable, and maintainable code from Indian companies...

Re:I can't wait

[Anonymous Coward]

i really hope that was sarcasm! i bet its written in java too, lol.

i'll be sticking to btrfs thank you, rather than a "bolt it on as a module as we've still fucked the licensing up" filesystem.

i can't see how this is any better than the fuse version. if it isn't mainstream it won't get much support from userland tools etc.

wtf?

[sofar]

154 comments and nobody mentions btrfs? Seriously, we don't need zfs anymore... And most definately not some half-assed probably-patent-encumbered version that will never make it upstream - mark my words: it will be unmaintained within 10 minor kernel versions.

Re:wtf?

[Renegade88]

Obviously other people don't share your high regard for the current state of BtrFS in relation to ZFS. I wonder why that is?

Re:wtf?

[inode_buddha]

Stick around a while, this might get interesting. I plan to see what happens. btrfs has no encumberances, you are correct about that; but OTOH ZFS is production-tested and proven - and lots of ppl are drooling for it (along with dtrace). The whole legal morass (licensing) is definitely *going* to tie things up for a while though, IMHO. From a historical perspective, you should also remember that Solaris draws from SunOS which SunOS in turn tried to merge the best of SysV and BSD. Which coincidentally Linux also tries to do. So I think it's a good match.

Re:wtf?

[sofar]

Name one linux distribution that ships zfs as part of the default installation, and installs by default on it.

Re: BTRFS - state of affairs

[internet_everyone]

Its easy to DOS the BTRFS and kill the system. A normal user can do it. Just create a couple of big files. Then, copy a large number of small files. You will get ENOSPC at around 90%. Then, delete the large files. Even with most of the space being free, it won't allow you to create new files. All new requests will get ENOSPC. Bug is at: https://bugzilla.kernel.org/show_bug.cgi?id=16508 BTRFS is a mess. Go read their mailing lists.

Native ZFS has already been ported to Linux

[Anonymous Coward]

And the source is available now: http://github.com/behlendorf/zfs/wiki

Has nobody seen this?

Re:Native ZFS has already been ported to Linux

[trasz]

That doesn't include ZPL layer. Basically, it's not even a filesystem, in that you can't put any files on it.

ZFS on Linux

[smash]

Given that Oracle were pushing BTRFS, they also push linux, and they now own Sun (and thus ZFS), suing over ZFS development on Linux would kinda be shooting themselves in the foot.

I can't see how oracle being able to use ZFS on linux and then easily upgrade their customers to Solaris + ZFS as needs change would be a bad thing for them?

What about patents?

[yyxx]

Oracle probably has a ton of patents around ZFS as well, so merely complying with the license is probably not enough. Just look at Java...

Re:What about patents?

[trasz]

CDDL license grants the user the right to use patented techniques. In other words, it guarantees that Oracle can't sue people using ZFS code.

That's not the GPL's fault

[symbolset]

That's not the GPL's fault. It's the fault of the IP lawyers who are dicing permissions exceedingly fine. The GPL is designed to guarantee certain freedoms at the cost of others. It does its job very well, and is well architected with a lot of forethought considering we're only on version three after 21 years. At least one of those two revisions can be blamed not on the faults of the license but on the changing legal and IP environment.

Believe it or not once upon a time if you wrote some code somebody found interesting you just sent it to them. No patents. No copyrights. No approvals from management or legal. You just sent it, happy that someone else might benefit from not redoing the work you'd done once already. The idea of profiting from the derivatives they might make, or the derivatives of the derivatives, was simply not an idea that would occur to a normal person. If you had suggested such a thing at that time we'd have thought it hilarious.

And now I have to point to the onion on my belt, which was the fashion in my day.

Re:That's not the GPL's fault

[TheRaven64]

Yes, it is the GPL's fault. The CDDL is a per-file license. It places absolutely no restrictions on what other code can be combined with it in other files. You can mix CDDL'd code with code under any other license, Free or proprietary, without the CDDL having any problems. The GPL is the one that introduces a set of conditions which restrict what you can link the code with.

The CDDL, for example, has some quite strict patent retaliation clauses. If you assert patent claims against any contributor to a CDDL codebase, then you immediately lose the rights to the entire CDDL codebase. The GPL has no such retaliation clause, and it contains a specific requirement that it can not be linked against any code that is under a license with additional conditions. This means that the GPL prevents you from incorporating CDDL code, even though it is a license which promotes Free Software - something that GPL zealots keep telling me was the aim of the GPL.

And that is why the claim in the summary that they are confident that Oracle will not sue is meaningless. Of course Oracle will not sue - they would not have standing to do so. Porting ZFS to Linux would not be violating the CDDL in any way. Distributing the result along with the Linux kernel, however, would be violating the GPL, so any Linux kernel developer would have standing to sue. Given the hostility to the CDDL among Linux developers, this is not a risk that any distribution is likely to take, meaning that a ZFS port done in this way is likely to remain an obscure niche product.

Re:That's not the GPL's fault

[Znork]

Yes, it is the GPL's fault. The CDDL is a per-file license. It places absolutely no restrictions on what other code can be combined with it in other files.

As the CDDL is deliberately GPL incompatible, had there not been any other issues, one can assume that Sun would have added 'may not be distributed together with GPL licensed code'. The CDDL/GPL incompatibility was on purpose, it was a feature asked for by Solaris engineers. Had the Linux kernel been BSD licensed, the CDDL would have been made incompatible with the BSD license.

Generally, fault implies some form of control over the issue. Under the circumstances, the only party with any control in this case would have been Sun, and as they would have redesigned the license until it was not compatible, it's quite obvious where any 'fault' should be assigned.

And unless the Oracle buyout has changed some attitudes within Sun for the better (heh), it's also quite naive of KQ Infotech to believe that Sun/Oracle would not go after them for violating the point of the license, as opposed to the actual text of the license (assuming any wider distribution). Standing is hardly a necessary prerequisite for a company of Oracles size to grind a small company into dust in the courts (and both Oracle and Sun would have standing as kernel contributors to sue any distributor of ZFS+Linux kernel combo).

Personally I can't say I consider it either a big loss or much to complain about. ZFS was a huge (HUGE) deal for Solaris, considering the painfully anemic storage stack it had in disksuite+ufs, but for any OS with a more modern volume management and file system stack it merely boiled down to a few nice features and some drawbacks, depending on your underlying storage architecture (SAN capabilities, etc).

Useful but needs the zpool recovery feature

[Cato]

ZFS is nice but it doesn't have an fsck, partly because of its design philosophy. I think this is an omission and http://www.osnews.com/story/22423/Should_ZFS_Have_a_fsck_Tool_ explains why (links to many examples where ZFS won't open a zpool after an abrupt shutdown, possibly due to ZFS bugs or disks that lie about flushing cache to disk).

There is a new zpool recovery feature - discussed in http://www.c0t0d0s0.org/archives/6071-No,-ZFS-really-doesnt-need-a-fsck.html - the command is "zpool clear -F data" which is a very specific sort of fsck - it just unwinds the last few transactions, enabling you to have a valid zpool but losing a few recent updates, which is usually better than a complete restore from backups. The feature is mentioned at http://www.c0t0d0s0.org/archives/6067-PSARC-2009479-zpool-recovery-a.html (PSARC 2009/479) and is available since Opensolaris build b128 (ref: http://opensolaris.org/jive/thread.jspa?threadID=127689&tstart=0 )

For those who want a ZFS based NAS, have a look at NexentaStor (which has a proprietary GUI, free for up to 12 TB of disk) or Nexenta Core (just the OS without GUI, and open source) - NexentaStore has some nice features to get you started quickly, or you can use the open source napp.it GUI with the Nexenta Core. Nexenta uses a very recent zpool version (v24) and is based on OpenSolaris build b134 so it includes the above zpool reocvery feature.

Nexenta generally will move to using the Illumos fork of OpenSolaris when that's stable, so it should have a future as long as NetApp don't sue them. If they do get sued you could move to a more community-based distro based on Illumos.

Key question is whether the ZFS on Linux port will be updated to the Solaris b134 code to include this feature. Without it, you are in for some painful recovery using zdb (filesystem debugger) - but in any case you need up to date backups of your entire zpool. FreeBSD does have ZFS but using a much older zpool version without this feature - from my point of view, it's best to use the latest Solaris ZFS code to get the best stability, despite the limitations of Solaris hardware support.

ZFS recovery is an interesting topic given commodity hardware - see http://opensolaris.org/jive/message.jspa?messageID=292794 - there are other failure modes not addressed buy this zpool recovery feature.

In corporate environments ....

[Anonymous Coward]

.... you rarely have to make a volume smaller.

I know that is not an excuse, but frankly I personally could not care less since that is a feature that would not be used at all...

Re:In corporate environments ....

[Just Some Guy]

.... you rarely have to make a volume smaller.

Try to add a new device to one of the pools on a system. Intend to write:

# zpool add mediapool /dev/newdisk

...but accidentally pull this up from your shell history:

# zpool add bigassraid /dev/newdisk

Panic when you realize that you just told the system to load-balance operations between that 64-drive RAID and a 160GB PATA drive you were screwing around with, and that you can't remove the new drive without backing up and restoring the whole RAID.

I love ZFS, but I despise the fact that you can't undo operations that seem like they should be very undoable.

Can we stop this bullshit once and for all please?

[Anonymous Coward]

Software has at the very least 2 actors that interact with it, the developer of the software and the user of the software.

So when you say "The BSD license is less restrictive than the GPL", who are you talking about? Which interests are you defending or promoting?

If you are an user, BSD software is a pain in the neck, since often you can't do anything about troubleshooting a problem except dutifully waiting for the company providing it to fix the problem (so for users BSD licensing can, and often is, akin to closed source software).

If you are a developer, GPL is a pain in the neck, since you can't distribute your work without granting the same rights you received when you used GPLEd software as part of your project (even then you can profit from your software without giving the source away if you actually don't distribute anything, a situation quite prevalent in the Internet if you have not noticed).

So your absolutis nonsensical statment clearly requires more context, I am sick an tired of people stating this nonsense when the situation is far more nouanced than they would like it to be.

insert first mention of porn ..

[Anonymous Coward]

gfgfg gdsdsrhbjdfa ;)

Sun and their licenses

[Anonymous Coward]

Does anybody know, how former Sun strategists think today of their license game?

cb

It has cost them Apple

[Ilgaz]

ZFS fits amazingly well to Apple's future and their style of computing. Right after this Netapp thing, they switched on "panic mode" and got rid of it. If they had a brain, 10.6 could be running HFSZ now.

Funny is, they will eventually switch to something which has ZFS features, don't forget "be nice to Apple if you want to deal with Disney/Pixar" factor. Some suits will really wonder if it was really worth it when World's trend leader uses something instead of their patents.

They could license it for cheap without hurting their main line of business. It is not like soon World will switch to Xserve blades you know. Apple is a company who can even license "Arial" from MS, a cheap mock up of Helvetica. SJobs isn't exactly rms.

Nothing gets a Linux sysadmin an erection faster

[Anonymous Coward]

...than hearing there's yet another FS coming to Linux.

ZFS is useless on phones, & laptops too

[toby]

If you look at the major features of ZFS, none of them apply to phones:
1) pooled storage management
2) integrity + self healing
3) snapshots
etc

To get the real payoffs with ZFS you need to have device redundancy. Also, a ZFS stack likes a lot of RAM. There's no real reason to run it on a phone or even a laptop (for most people). There are plenty of suitable filesystems for these situations.

ZFS works in 32bit x86 Solaris 10 & OpenSolari

[toby]

But it's aimed at the bigger system/server market, which does tend to be 64 bit. The 32 bit end is not going to get as much love.

Also see what TheRaven64 wrote above.

Rejoicing.

[Leolo]

... there was much of it. Very much of it!

You mean free by Stallman's own definition

[judeancodersfront]

I am also capable of making a politically loaded definition of freedom that fits my goals but then I would feel like a dirty car salesman who writes his own definition of one year warranty that is out of line with the common meaning.

Not that unstable really?

[judeancodersfront]

I don't want a file system that is described as not that unstable really. I'll take just plain stable thank you.