Anthony Green wrote:
On Fri, 2005-11-18 at 01:05 -0500, Davanum Srinivas wrote:
Anthony,
People who are willing to go the GPL route for their code have lots of
avenues as u know...
(It's not the GPL, it's the GNU Classpath license!)
The point of my request is to make sure contributors are informed.
Of course, I'm working under the assumption that GNU Classpath is
important to the Harmony project. Perhaps it isn't. But, if it is, I
think it makes sense to inject some advertising for GNU Classpath into
the Harmony contribution process, because that will be the most
convenient time for people to consider and arrange for dual licensing.
There is a big problem in dual licensing: it asks for forks.
What would be the point of somebody donating a bunch of code to the ASF
under a dual license that allowed Classpath to take it, modify it and
those modifications not coming back?
The apache license *does* allow people to do exactly the above, it's
part of the BSD philosophy of playing nice with the industrial ecosystem
around you that GNU does not share (not a criticism, just a matter of fact).
So, the real problem is not Classpath getting the code that harmony
produces and not give the patches back in a form we can digest, that we
can totally live with.
What we wouldn't like is the fact that this code is supposed to be a
seed for a community around it, community that values the apache
licensing philosophy and plays fair with the Sun IP licensing restrictions.
If there wasn't a need for this, IBM would have donated his code to
Classpath under their license in the first place.
The *IS* a reason why two projects exists, they serve different
communities with different purposes. Asking the ASF to dual license with
a license that is philosophically unacceptable is socially disruptive,
as it would be for me to ask the Classpath people to start dual
licensing just for sake of collaboration.
It ain't gonna happen, so let's not go there.
*but* dual licensing is not the only way we can collaborate. Here is my
vision:
1) the two projects need a middle ground, a philosophy-neutral
licensed interface code that can be used by both and implemented by all
the pieces of the puzzle. I would suggest something like
http://www.openvm.org/ and an MIT license. Both projects can use it,
both projects can implement it, both projects can decide to adhere to it
for sake of immediate reusability *and* clear license virality isolation.
2) the ASF needs to get its act together and make sure that the LGPL
and the GNU Classpath license *AS IS* are acceptable for linkage and
distribution from apache licensed projects. This is in the interest of
the foundation. This is the current top priority of our VP of legal
affairs (Cliff), who is now pending comments from Eben and Richard
directly on a few of the last issues on the LGPL interpretation. I am
personally confident that we will get at least the LGPL resolved by the
end of the year. Once the LGPL issue is resolved, it would be a lot
easier to solve the GNU Classpath license issue as well, as social
friction would have been strongly reduced.
3) the GNU Classpath project needs to understand whether they want to
include apache licensed code or not. It's not just harmony: xerces,
xalan, log4j, there is a lot of code that it could be just reused
instead of being reinvented for sake of licensing issues. This would
require a modification to the GNU Classpath license that makes the two
licenses compatible, practically says that it allows the patent
litigation clause present in the Apache license to apply to those
distributed parts of the code. It would not change the social activity
of the project, it would not force dual licensing, yet it would help
classpath avoid to write a ton of code and would show a tangible sign of
collaboration that is not socially disruptive on any of the two sides.
At the same time, if #3 is not considered viable because it's considered
weakening the position with the enemy, I would understand. Very few
people, on both sides, believe in the value of collaboration between two
projects and would much rather spend time coding or attracting more
contributors than weakening their philosophical positions.
And my personal history tells, in fact, that it's a lot easier to write
some code than to change somebody's mind, so I'm not that obsessed with
collaboration in terms of code sharing.
But I do think that the API interface middle ground would go a long way
to allow easier connection of all the pieces together.
So, real question: how many people here would participate in such a VM
API (can't call it *J*!) effort if it was hosted not by the ASF or by
the FSF and licensed under a neutral MIT license?
Note, I also volunteer to host it.
Thoughts, comments?
--
Stefano.