Stephen Harpster's Weblog
The Alethiometer
Archives
« June 2009
SunMonTueWedThuFriSat
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
    
       
Today
XML
Get OpenSolaris
Search

Links



blogs.sun.com
Weblog
About
Login

 
 

Today's Page Hits: 241

All | General | JDS | OpenSolaris
« Atlanta Tech Days | Main | Beryl »
20070130 Tuesday January 30, 2007
GPLv3?

In the last few months I've seen more and more speculation about the prospect of dual-licensing OpenSolaris under GPLv3. In November Jonathan very publically asked Rich if he would look into it, and everyone knows that we are fully engaged in the GPLv3 process. As Rich has made clear, we're looking into it. No decisions have been made. We've seen discussions in blogs (Rich, Simon) and in the news (CNet, eWeek), but I haven't seen much in the OpenSolaris community itself.


I think that we ("we" being all of you) should be asking ourselves what we think about GPLv3. What would it mean to the community if we dual-licensed? It's now a possibility that we could attach an "assembly exception" to the GPLv3 which would let us mix GPL and CDDL code. This could open up a world of possibilities.


But what are the downsides? What does the community, you, think of the way GPLv3 is taking shape? These are important issues and I urge everyone with an opinion to voice it sooner rather than later.


Technorati Tags: ,


Jan 30 2007, 04:03:59 PM PST Permalink Comments [12]

Comments:

Having prominent Linux developers being up-front against GPLv3, I think that as soon as there is a GPLv3'd OpenSolaris, this will be the "next kernel for GNU", something that could be replacing Linux immediately (at least in theory) and attract a lot of developers (and users like me) who want a "free-as-in-free-speech" system of their choice. Linux pushed FOSS forth in an incredible way, but I am not sure it's the best "free software" OS kernel out there.

Posted by kawazu on January 30, 2007 at 11:09 PM PST #

From my point of view, dual-licensing OpenSolaris under GPLv3 would increase the number of potential contributors quite a lot.

Let's be pragmatic. The target is the project itself, the license is just a tool. If by changing the license we make the OpenSolaris community to keep growing.. why not?

Posted by alo on January 31, 2007 at 01:10 AM PST #

The question is not "what are the downsides?", but rather "what issue does this address?" The downsides are as plain as day (the burden of dual-licensing, the inevitability of a fork as a result). You mention that the possibilities are endless: could you name few?

Posted by Adam Leventhal on January 31, 2007 at 08:30 AM PST #

With regards to "assembly exceptions", it would be much more interesting to license parts of OpenSolaris under a BSD license; as it stands the FreeBSD Project is unable to include dtrace support in the default kernel as we cannot mix the CDDL into it. GPLv3 will not resolve that, and I'd much rather see a discussion over how Sun can help to resolve such situations. As for GPLv3, I'd be far more interested if GPLv3 actually existed. As it is, it stands to be more viral than the last incarnation and therefore I'm even less interested. Like Adam, I'm also very concerned about forks, what it means for code quality and so on, and I would much rather have OpenSolaris stay with the CDDL than go GPLv3 (or a dual-license therewith).

Posted by Ceri Davies on January 31, 2007 at 08:46 AM PST #

If FreeBSD can't inclue CDDL-licensed code, how is it that the FreeBSD project is planning to include ZFS?
Dragonfly BSD doesn't seem to have a problem with the CDDL, either.
In fact:
DTrace is being ported to FreeBSD in such a way that it will be possible to do a source-level build of the FreeBSD operating system without any of the source files covered by the CDDL. This means that it will still be possible for anyone to produce a release of FreeBSD which contains only BSD licensed source files. Of course such a release won't have a compiler or linker because those are covered by the GPL.
[link]

Posted by Jim Thompson on January 31, 2007 at 10:48 AM PST #

(This comment represents my opinion, and mine alone. While it is influenced by my job as an employee of Tadpole, it doesn't necessarily represent any official opinion of Tadpole or General Dynamics.)

I would far, far rather steer away from GPLv3. The viral nature of it, and the desire of some developers to promote their political agendas with software licensing, is likely to lead to a split in the community. (Ask me if you need an explanation of why.)

Commercial developers cannot work with GPLv3 easily -- there are often requirements that certain things be closed. CDDL allows bits that need to be closed, while keeping the open bits open. GPLv3 also breaks certain kinds of applications that might be required -- such as security appliances where upgrade/key management facilities must be controlled.

Please, please, keep the code CDDL, or dual license under BSD.

Dual licensing under BSD would allow fair play with the BSDs, and also allow Solaris code to be introduced into Linux, without encouraging new Solaris code to be introduced under GPLv3.

Posted by Garrett D'Amore on January 31, 2007 at 11:12 AM PST #

Jim, I can assure you that are problems with the DTrace port related to the way that traps are done, the code for which is under the CDDL. In order for DTrace to work at all, this code must be present in the kernel whether or not DTrace proper is enabled, and therefore DTrace must be a compile time option without a license change on that code. In other words, the author of the sentence that you quoted has since found that it isn't possible to do after all. ZFS doesn't have the same problem, and so it isn't an issue; non-BSD licensed code is fine in the tree, it just isn't fine in the GENERIC kernel.

Posted by Ceri Davies on January 31, 2007 at 11:12 AM PST #

As for dual-licensing OpenSolaris under GPLv3, I asked rms about this directly while he was here in Hawaii. He is cautious, but thinks it would be great. How can he think otherwise?
Sun, with a long period of sweat and heavy lifting, capped by a stroke of the pen, could produce the first Free (as in speech) distribution that non-geeks could appreciate.
  • Red Hat isn't Free
  • Fedora isn't Free
  • (k)Ubuntu isn't Free
  • SUSE isn't Free
  • OpenSuse isn't Free
Open Solaris could be (if the entire thing was Free Software).
GPLv3 is going to cause more trouble for the linux kernel (who's advocates seem to want to keep it GPLv2-only) than anyone else. You can't mix GPLv3 code and GPLv2 code, as the licenses are believed to be incompatible. (At least, rms thinks this is true. And yes, GPLv3 isn't in final form.)
Personally, I think a GPLv3 Open Solaris would start to deflate the linux world. The place where (GNU)/linux wins most is on the server, and the marketspace with the most servers is... the Enterprise.
The Enterprise market will likely re-tool around a GPLv3 OpenSolaris in the next several years. Here's why:
  • OpenSolaris is faster than the linux kernel running linux apps.
  • Sun can provide far better world-wide support than any of the Linux support companies (including RedHat and Oracle).
  • Sun provides less expensive support than the linux distro companies.
  • Fortune 1000 companies would rather deal with their own. Red Hat and Novell aren't on that list. Sun (and Oracle) are.
  • Sun has large direct salesforce. Red Hat doesn't. (Oracle does, but I don't think they're in it for the long haul.)
The second place where linux currently 'wins' is in the embedded space, but the profitable segment of that market is CE (consumer electronics). Apple's recent entry there (with AppleTV and the iPhone) means that small companies seeking to bring embedded linux into a CE box have a much more difficult time:
  1. Investors will be shy of anything that competes head-to-head with Apple.
  2. Apple has a large presence in the "consumer" market (think: Apple Stores and trained employees at places like CompUSA)
  3. Apple "gets it" in ways that other companies known to consumers (such as Sony and Microsoft) don't. People enjoy using Apple's existing CE devices, and I think the iPhone will be a revelation.
So if the Enterprise/server marketplace starts to turn (dare I say 'back'?) toward Solaris, and Apple takes the high-margin business away from embedded linux, what does linux have left?
Well, linux has a bunch of obnoxious "open source uber alles" freaks, and they are dead-on trouble for the various distros and companies who use the linux kernel (and other GPL code) in embedded products.
And if Microsoft starts to gain traction in their semi-declared patent war against linux, the Enterprise (and SME) markets will flee. Corporate edict will issue forth from the offices of the General Council, and linux in large businesses will be crushed.
in summary, Linux will survive in the SMB and SOHO markets. Linux will continue as a hobbist movement, but Linux will never win the desktop, and its existing market strength will be weakened by a GPLv3 OpenSolaris

Posted by Jim Thompson on January 31, 2007 at 11:25 AM PST #

Garrett: Tadpole? Tadpole?!? The company you work (RDI) for is a shallow reflection of the company that produced the SPARCbook. (I was the lone US member of the engineering team at Tadpole, Plc.) Ceri, I am aware of this

Posted by Jim Thompson on January 31, 2007 at 11:32 AM PST #

Jim,

As you're aware of the problem, I'm confused why you attempted to pretend that it doesn't exist. I'm also mystified by your comments to Garrett.

Stephen, please take note of the problems I mentioned.

Ceri

Posted by Ceri Davies on January 31, 2007 at 01:44 PM PST #

I have tried to post a detailed comment here but this stupid blog software rates it as SPAM.

Posted by John Birrell on February 04, 2007 at 01:02 AM PST #

http://people.freebsd.org/~jb/comment.html

Posted by John Birrell on February 04, 2007 at 01:22 AM PST #

Post a Comment:

Comments are closed for this entry.