Sun's ZFS and Linux

Sun's ZFS and Linux

[Tarkan Erimer]

Hi,

On Nov.16, Sun has open sourced the
(http://www.opensolaris.org/os/announcements/#2005-11-16_welcome_to_the_zfs_community_
) ZFS. I know that, It is licensed under CDDL. So, It is not GPL
compatible. In this situation, there is no way for Linux mainline. But
I wonder, is there anybody has a plan to port ZFS for Linux as a
separate patch ?


Cheers!

Re: Sun's ZFS and Linux

[Theodore Ts'o]

On Fri, Nov 18, 2005 at 11:38:16PM +0000, Tarkan Erimer wrote:
> 
> On Nov.16, Sun has open sourced the
> (http://www.opensolaris.org/os/announcements/#2005-11-16_welcome_to_the_zfs_community_
> ) ZFS. I know that, It is licensed under CDDL. So, It is not GPL
> compatible. In this situation, there is no way for Linux mainline. But
> I wonder, is there anybody has a plan to port ZFS for Linux as a
> separate patch ?

That wouldn't be a "port", it would have to be a complete
reimplementation from scratch.  And, of course, of further concern
would be whether or not there are any patents that Sun may have filed
covering ZFS.  If the patents have only been licensed for CDDL
licensed code, then that won't help a GPL'ed covered reimplementation.

					- Ted

Re: Sun's ZFS and Linux

[Tarkan Erimer]

On 11/19/05, Theodore Ts'o <tytso@mit.edu> wrote:
> That wouldn't be a "port", it would have to be a complete
> reimplementation from scratch.  And, of course, of further concern
> would be whether or not there are any patents that Sun may have filed
> covering ZFS.  If the patents have only been licensed for CDDL
> licensed code, then that won't help a GPL'ed covered reimplementation.

Thanks for the explanation. BTW, I wonder something: Is there any
possibility to give GPL an exception to include and/or link to CDDL
code?

Thanks.

Re: Sun's ZFS and Linux

[Nix]

On 20 Nov 2005, Tarkan Erimer yowled:
> On 11/19/05, Theodore Ts'o <tytso@mit.edu> wrote:
>> That wouldn't be a "port", it would have to be a complete
>> reimplementation from scratch.  And, of course, of further concern
>> would be whether or not there are any patents that Sun may have filed
>> covering ZFS.  If the patents have only been licensed for CDDL
>> licensed code, then that won't help a GPL'ed covered reimplementation.
> 
> Thanks for the explanation. BTW, I wonder something: Is there any
> possibility to give GPL an exception to include and/or link to CDDL
> code?

You'd have to get agreement from *all* the kernel's past
contributors. As some of them are dead this is not likely to happen.

(Well, OK, you could isolate their code and rewrite it but this
would be a big and annoying job, so you'd need a very compelling
reason. One extra filesystem isn't likely to be good enough.)

-- 
`Y'know, London's nice at this time of year. If you like your cities
 freezing cold and full of surly gits.' --- David Damerell

Re: Sun's ZFS and Linux

[Rob Landley]

On Sunday 20 November 2005 15:12, Tarkan Erimer wrote:
> Thanks for the explanation. BTW, I wonder something: Is there any
> possibility to give GPL an exception to include and/or link to CDDL
> code?

No, and Sun likes it that way.

The GPL was the first "copyleft" style license which requires that derivative 
works be placed under exactly the same terms as the original work.  If the 
terms of another code are incompatible, they cannot be exactly the same.  

(Specifically, the GPL says in section 2b, "You must cause any work that you 
distribute or publish, that in whole or in part contains or is derived from 
the Program or any part thereof, to be licensed as a whole at no charge to 
all third parties under the terms of this License."  See 
"http://www.gnu.org/copyleft/gpl.html" and 
"http://www.fsf.org/licensing/licenses/index_html#GPLIncompatibleLicenses".)

Sun intentionally designed the CDDL to be incompatible with the GPL.  This was 
a design goal on Sun's part.*  They want to isolate themselves from the 
existing open source community, and make sure that their code cannot be used 
with the most common open source license.**  Why they want to do this has 
been widely speculated about***, but the fact they want an explicit "us vs 
them, no sharing" stance is not in dispute.

Rob

* See http://www.vnunet.com/vnunet/news/2127094/sun-slams-predatory-gpl or
http://news.com.com/Sun+criticizes+popular+open-source+license/2100-7344_3-5656047.html 
or http://www.adtmag.com/article.asp?id=10927 plus Sun's official rationale 
at http://www.sun.com/cddl/CDDL_why_details.html

** According to http://sourceforge.net/softwaremap/trove_list.php?form_cat=13 
there are currently 72,823 projects on sourceforge specifying a license.  Of 
those, 48050 have chosen to license their code under the GPL.  That's 65.98%, 
or about 2/3 of the total.  In politics, this would be flirting with a 
veto-proof majority.  David Wheeler did a detailed analysis at 
http://www.dwheeler.com/essays/gpl-compatible.html

*** see http://lwn.net/Articles/114839/ or http://lwn.net/Articles/159248/ or 
http://www.eweek.com/article2/0,1759,1754155,00.asp or 
http://www.eweek.com/article2/0,1895,1739000,00.asp or
http://searchopensource.techtarget.com/qna/0,289202,sid39_gci1060779,00.html 
or http://www.technewsworld.com/story/40176.html or 
http://www.vnunet.com/vnunet/news/2126648/sun-hits-back-open-source-critics
or...

Re: Sun's ZFS and Linux

[Jeff V. Merkey]

Good description from Rob L. n the CDDL. No, you can't rip off 
everything and put it in Linux just because someone
put it out another another license and made it public.  ZFS will 
contaminate Linux.  There's already plenty of FS's in Linux
as it is, and who cares about some broken piece of crap from Solaris 
anyway.    Putting ZFS into Linux would help
Sun in any event and not Linux.

Jeff

Re: Sun's ZFS and Linux

[Tarkan Erimer]

The whole picture is more clear now. Thanks for
this very informative reply.

Regards

On 11/21/05, Rob Landley <rob@landley.net> wrote:
> On Sunday 20 November 2005 15:12, Tarkan Erimer wrote:
> > Thanks for the explanation. BTW, I wonder something: Is there any
> > possibility to give GPL an exception to include and/or link to CDDL
> > code?
>
> No, and Sun likes it that way.
>
> The GPL was the first "copyleft" style license which requires that derivative
> works be placed under exactly the same terms as the original work.  If the
> terms of another code are incompatible, they cannot be exactly the same.
>
> (Specifically, the GPL says in section 2b, "You must cause any work that you
> distribute or publish, that in whole or in part contains or is derived from
> the Program or any part thereof, to be licensed as a whole at no charge to
> all third parties under the terms of this License."  See
> "http://www.gnu.org/copyleft/gpl.html" and
> "http://www.fsf.org/licensing/licenses/index_html#GPLIncompatibleLicenses".)
>
> Sun intentionally designed the CDDL to be incompatible with the GPL.  This was
> a design goal on Sun's part.*  They want to isolate themselves from the
> existing open source community, and make sure that their code cannot be used
> with the most common open source license.**  Why they want to do this has
> been widely speculated about***, but the fact they want an explicit "us vs
> them, no sharing" stance is not in dispute.
>
> Rob
>
> * See http://www.vnunet.com/vnunet/news/2127094/sun-slams-predatory-gpl or
> http://news.com.com/Sun+criticizes+popular+open-source+license/2100-7344_3-5656047.html
> or http://www.adtmag.com/article.asp?id=10927 plus Sun's official rationale
> at http://www.sun.com/cddl/CDDL_why_details.html
>
> ** According to http://sourceforge.net/softwaremap/trove_list.php?form_cat=13
> there are currently 72,823 projects on sourceforge specifying a license.  Of
> those, 48050 have chosen to license their code under the GPL.  That's 65.98%,
> or about 2/3 of the total.  In politics, this would be flirting with a
> veto-proof majority.  David Wheeler did a detailed analysis at
> http://www.dwheeler.com/essays/gpl-compatible.html
>
> *** see http://lwn.net/Articles/114839/ or http://lwn.net/Articles/159248/ or
> http://www.eweek.com/article2/0,1759,1754155,00.asp or
> http://www.eweek.com/article2/0,1895,1739000,00.asp or
> http://searchopensource.techtarget.com/qna/0,289202,sid39_gci1060779,00.html
> or http://www.technewsworld.com/story/40176.html or
> http://www.vnunet.com/vnunet/news/2126648/sun-hits-back-open-source-critics
> or...
>

Re: Sun's ZFS and Linux

[Bernd Petrovitsch]

On Sat, 2005-11-19 at 12:23 -0500, Theodore Ts'o wrote:
> On Fri, Nov 18, 2005 at 11:38:16PM +0000, Tarkan Erimer wrote:
> > 
> > On Nov.16, Sun has open sourced the
> > (http://www.opensolaris.org/os/announcements/#2005-11-16_welcome_to_the_zfs_community_
> > ) ZFS. I know that, It is licensed under CDDL. So, It is not GPL
> > compatible. In this situation, there is no way for Linux mainline. But
> > I wonder, is there anybody has a plan to port ZFS for Linux as a
> > separate patch ?
> 
> That wouldn't be a "port", it would have to be a complete
> reimplementation from scratch.  And, of course, of further concern
> would be whether or not there are any patents that Sun may have filed
> covering ZFS.  If the patents have only been licensed for CDDL
> licensed code, then that won't help a GPL'ed covered reimplementation.

Hmm, one could thake the zfs as a blurb and write a GPL'ed adapter (as
external patch) to the Kernel (similar to the nvidia ones and their
binary blurb drivers). The ZFS blurb would count as "not derived" since
it is IMHO exactly that.
And now I don't know if it makes sense, could actually work or how much
work it is. Experienced VFS people may have a opinion on this.

	Bernd
-- 
Firmix Software GmbH                   http://www.firmix.at/
mobil: +43 664 4416156                 fax: +43 1 7890849-55
          Embedded Linux Development and Services

Re: Sun's ZFS and Linux

[Bill Davidsen]

Theodore Ts'o wrote:
> On Fri, Nov 18, 2005 at 11:38:16PM +0000, Tarkan Erimer wrote:
> 
>>On Nov.16, Sun has open sourced the
>>(http://www.opensolaris.org/os/announcements/#2005-11-16_welcome_to_the_zfs_community_
>>) ZFS. I know that, It is licensed under CDDL. So, It is not GPL
>>compatible. In this situation, there is no way for Linux mainline. But
>>I wonder, is there anybody has a plan to port ZFS for Linux as a
>>separate patch ?
> 
> 
> That wouldn't be a "port", it would have to be a complete
> reimplementation from scratch.  And, of course, of further concern
> would be whether or not there are any patents that Sun may have filed
> covering ZFS.  If the patents have only been licensed for CDDL
> licensed code, then that won't help a GPL'ed covered reimplementation.
> 
What a great chance to try out userfs.

-- 
    -bill davidsen (davidsen@tmr.com)
"The secret to procrastination is to put things off until the
  last possible moment - but no longer"  -me

Re: Sun's ZFS and Linux

[Theodore Ts'o]

On Mon, Nov 21, 2005 at 03:40:22PM -0500, Bill Davidsen wrote:
> >That wouldn't be a "port", it would have to be a complete
> >reimplementation from scratch.  And, of course, of further concern
> >would be whether or not there are any patents that Sun may have filed
> >covering ZFS.  If the patents have only been licensed for CDDL
> >licensed code, then that won't help a GPL'ed covered reimplementation.
> >
> What a great chance to try out userfs.

Just for yucks, people who are interested in doing might want to first
implement ext2 in userspace --- this would be relatively easy, given
that most of the code to do this is already in libext2fs, and
interface it to userfs.  Next, benchmark ext2 in userspace using
userfs, and compare it to ext2 running in the kernel using the
identical kernel and hardware configuration, and report on the
results.  Try doing this on both a uniprocessor system as well as a
4-way SMP system, and let us know what you find.....  

I think I know, but it would be a very interesting experiment, and
would probably be a great paper to publish at some conference such as
OLS, LCA, LK, etc., especially if were combined with suggestions about
how to improve userfs's performance.

						- Ted