Monday, March 31, 2008
Google blocking AGPL in Google CodeI received an email this morning from a friend who is planning to put his open source project in Google Code, the open source repository hosted by Google.
He is concerned by this thread in Google groups. A developer asked, back in November:
Can I host a project on Google's Project Hosting, if it's licensed under AGPL? It's one of FSF's licenses. I really need AGPL instead of GPL, as my project is a server-side software.Chris DiBona, Google master of open source, answered:
In fact we do not support the AGPL on code.google.com. We are actively trying to fight the proliferation of licenses that are considered open source and the AGPL both has very little market share and has not been certified as being open source by the OSI.Now, that was November 2007... Later in the thread Chris writes:
We have no current plans to do so. If AGPL adoption is high enough to warrant it, we'll revisit. The question is this: Who will propose the AGPL to the OSI? :-)Nice :-) Funambol did it. AGPL is OSI approved. When pushed again, Chris switched from AGPL not being OSI-approved to not being a popular license:
what really matters is that the license is adopted. So far, we haven't seen much of it.So, first AGPL was not good enough for Google because it was not OSI-approved. That limited its popularity... Now it is OSI-approved. Still, it is not popular enough to be accepted in the Google closed open source hosting site?
And, by the way, why should people put their open source code in the hands of someone who likes open source only when it does good to its business (ehm, that could include me, but we are not talking about me, are we ;-) ?
C'mon Chris, give developers the ability of using AGPL for their own projects in Google Code. Your fight for no proliferation of licenses is something I subscribe to, but AGPL is the license of the future, no matter if Google likes it or not. And I can guarantee you it will become even more popular if it is accepted in Google Code...
Posted by Fabrizio at 08:34
Chris DiBona said...
Wait, what? As I said both times :
Popularity + osi = ok for code.google.com ok
You have OSI, which is good, but AGPL is still not very popular. Heck, pass the EPL in popularity and we'll talk about it. Right now on SF.net there at 8 agpl'd licensed packages to EPLs 674. 8. Eight!
That's in base 10, in case you were wondering.
Heck, even the Nokia public license has 19 projects.
In the meantime, there are plenty of other hosts out there that are happy to serve you. Code.google.com isn't the only host in the world.
Comment Posted at 18:35
Tracy Poff said...
Hey, now. Chris said back in November:
"If AGPL adoption is high enough to warrant it, we'll revisit."
You quoted it in this very post! It's a bit disingenuous of you to act like Chris is moving the goalposts.
I have no comment on the merit of your idea since I'm not familiar with the terms of the AGPL. I figure that reducing license proliferation is a noble cause, though, so it doesn't seem unreasonable to want the license to have some adoption before accepting it.
Comment Posted at 21:27
Per Abrahamsen said...
I'd like to add that the point about Google not being the only provider is valid. They don't have to be all things to all people, but can choose to focus. The situation was very different when SourceForge started, at they were pretty much the only game in town. Today, it makes much more sense for a new free hosting provider to try to find a niche, and excel in that.
Comment Posted at 11:03
Bradley M. Kuhn said...
I don't think Chris (DiBona) is being completely public about Google's full reasoning. The AGPL is a license that Google clearly dislikes for its own proprietary reasons — they have a vested corporate interest in impeding its popularity. I believe they are choosing not to allow AGPL'd code on their systems primarily for those reasons, and the “not popular enough” argument is a convenient excuse (now that they've lost their battle inside the OSI to block its approval).
It's disingenuous to compare licenses like the Nokia Public License to a license published by the FSF, which has a 25 year history of publishing licenses that do become popular over time, and also is opposed to license proliferation, and was so even back when OSI was in favor of proliferation. The FSF didn't write the AGPL just put its own stamp on a document which is otherwise nearly the same as other licenses (which is the common MO of license proliferation). AGPL is trying to address a completely new threat to the freedom of software users, and it happens to be a freedom that Google has a corporate interest in impeding.
The place where Chris and I agree wholeheartedly is that you shouldn't use code.google.com to host your FLOSS project. I recommend a host that will let you see all the source code of your hosting software, and build upon it, and not get locked in. SF.net is no better than google in this regard; I'd suggest Gna! or Savannah.
Comment Posted at 11:15
Bradley M. Kuhn said, "I don't think Chris (DiBona) is being completely public about Google's full reasoning."
Google came up with a policy to only support a small amount of licenses (7) from the start:
You don't see Google Code adding new licenses every time peope ask, even if they are non-vanity, OSI-approved licenses, so they are hardly singling out the AGPL.
As a package maintainer for a Linux distribution I fully support Google's stand against license proliferation.
In Gentoo Linux we have 883 different licenses - enough already.