The first day that the CAB met, almost two years ago, we talked about all of the things that OpenSolaris needed to do to become successful. Central to that discussion were three principles:
1) everyone needs to work on a common version control system 2) everyone needs to use a common issue reporting system 3) everyone needs to make decisions on the public lists none of which are true for OpenSolaris today. Any one of those would be far more effective at increasing participation than any change in license terms. Personally, I think discussion of GPLv3, right now, is absurd. I cannot think of anything more harmful for a community than filling it up with a bunch of chattering hens that have yet to contribute anything. Real development is forced off-list because such mindless discussions burn time and destroy productivity. I wouldn't have added this message to the fire, but some people think my silence means consent, when I really just think it is too stupid for words to adequately express. I wrote the Apache License 2.0. I participated in the MPL 1.1 process. I twice negotiated license changes with the FSF over the issue of compatibility, and in both cases the FSF made up a new excuse for their stance after publication. I wrote http://www.apache.org/licenses/GPL-compatibility.html with an understanding that the FSF would follow up in due time, and supposedly the GPLv3 effort will do just that eventually. I didn't expect them to take this long, but I am not surprised. I am far too familiar with RMS making random changes to FSF policies after everyone else has agreed to terms. The comments on their website about license compatibility come from RMS, not FSF lawyers. Even if the ultimate result of GPLv3 is a really cool license, it will still be a long time before we know how it works in action. Dual-licensing, OTOH, is a game that closed development organizations use to retain business licensing advantages over a code base that they know (or think they know) will not be forked successfully. It always involves the GPL and a closed source license, in tandem, because the GPL gets the free press of open source while at the same time avoiding the loss of closed-product licensees for those customers that fear the GPL). Oh, and it usually involves a code base that is so fugly that real open source developers would rather rewrite it from scratch than attempt a fork. I am not aware of any open development project that uses dual licensing. Open, as in, the real development decisions are made in public with the involvement of folks outside the primary sponsor (Mozilla does not qualify). Dual-licensing requires copyright assignment and a generally suspicious review of any incoming code contributions, which effectively neuters both the social and speed benefits of open development. Dual-licensing GPLv3 with CDDL is easily the dumbest idea I've heard in a long time. There is no community to be gained there. Mozilla certainly didn't gain any community support by doing so, unless you count press people who have never written a line of code in their lifetimes. All they did was satisfy one person's ego. If, on the other hand, Sun wants to go back and re-market OpenSolaris as yet another VC-backed closed development organization that uses open source for free press while doing all real development in-house, then by all means follow the MySQL, RedHat, JBoss, and assorted other examples and release as dual GPL / private-terms. You won't have a community, but maybe you'll get lucky on the IPO. [oops, maybe a bit late for that model too.] It's as if you guys are still reading slashdot and thinking that those trolls are an open source community. A real community is a long-term group with a sense of communal understanding and shared social relationship that has been built through working towards a common goal. Communities, by their very nature, are wary of transients and shun trolls. If you want a community, then earn it. Simon said:
I'm very familiar with the situation at Harmony and it doesn't worry me much as a precedent. The situation there was of an Apache project wanting an FSF project under the GPL with an exception generally considered ideal for the technology involved to drop the GPL (or rather dual-license under BSD) so Apache could take all their code.
Wrong. The situation was that an Apache project wanted to build an Apache-licensed JVM in cooperation with the Classpath developers, which would require their willingness to do work under the Apache license terms. Hence, the "Classpath exception" did not apply and could not be used by Apache. If what Harmony wanted was to simply take their code, it could have done so (without modifications) in accordance with the exception. But Apache doesn't give a rip about their code -- what we wanted (and what we always want) are the developers to join our community. If they don't want to, that's fine -- Harmony just wanted to make an open invitation. This was back when Sun was saying it was impossible to do an open source JVM.
We used the same Classpath exception for Sun Java last year and it received rapturous acclaim from the Free software community, and the fully-open Java ME community has already seen 4 platform ports I believe.
Rapturous acclaim? You have to stop believing your own marketing material. If Java had been opened under the BSD or Apache license, there would be a thousand new developers working on it the next week. As it is, the only reason the Java community was positive about it is because any open source license is far better than the previous terms, and there was some hope that it would also mean Sun would stop jerking the rest of the community around in regards to JSR implementations. Unfortunately, Sun is still jerking us around with their TCK agreements, still violating the JSPA terms, and still making a mockery of our prior agreements with Rob Gingell. But you know that already. Sun's use of the Classpath exception is a bizarre case of ignoring what the text of the exception actually says, including the fact that the exception isn't even a valid copyright license, and claiming hold of the "free software community" because Sun knows damn well that there is no community there to compete with Sun. The FSF is happy because they know the exception is nothing more than a covenant not to sue that defaults to the plain old GPLv2. Apache is happy because anything would be better than the prior status quo. Sun is happy because the free software foundation hasn't found out about the TCK licenses yet. Apache doesn't work the same way. When the ASF sees that a license doesn't allow open development under non-viral terms, we don't allow our projects to use it because that would be misleading our customers (the general public for which we are a nonprofit charity). People expect Apache products to have non-viral terms for redistribution. Sun doesn't have that issue because Sun's real customers will get a different license, and the GPLv2 is sufficiently viral to prevent existing licensees from jumping ship. Sun is not effected by the invalidity of the classpath exception because Sun is the copyright owner. The only real impact is on non-GPL communities that could have participated in fixing the bugs in Java, but probably won't bother at this point beyond porting to new platforms. From a legal and community perspective, CDDL would have been a better choice for Java, but I can understand the business concerns that would lead one to abandon the existing community in favor of profits and the usual anti-IBM/anti-MS propaganda. After all, it was just two years ago that you accused the other companies doing that of raping the third world of its IP. OpenSolaris, however, is a completely different world. This community, and all of its heritage, is based on the open innovation and sharing that is characteristic of the BSD licenses. If a license change is being considered, it would make far more sense to change everything to the BSD terms (with maybe a few extra words inserted to make the copyright license complete and ensure that a patent license is implied). That would be an invitation to a huge existing developer community which happens to be suffering from lack of leadership and shares the same principles as the existing OpenSolaris developers. It would also shut up the trolls that claim CDDL is incompatible with Linux. Otherwise, stick with the CDDL. GPLv3 cannot be evaluated seriously until it is actually approved by the FSF and published in final form. Even then, dual-licensing wouldn't make any sense, but at least we'd have an idea of the actual impact of a switch from CDDL to GPLv3. ....Roy _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org