Archive for April 2009
EPL ~= ASL
Boy, am I on some kind of blog binge this week. It’s amazing how a few weeks without travel helps.
Dana Blackenhorn wrote a piece earlier today discussing Matt Asay’s article on why the Apache license is better than the GPL. In his article, he states the following:
If your company wants to release its own code, and control that code, if open source is mainly a marketing concept to you, then a BSD license such as Apache or Eclipse makes perfect sense.
Dana’s statement makes an error that I have seen repeatedly. Namely, that the EPL is a “BSD-style” license and is therefore similar or equivalent to the Apache license. This is just plain wrong. And it worries me that a long-time open source observer such as Dana would make this mistake.
The EPL is what is sometimes referred to as a “weak copyleft” license. It most certainly is a reciprocal license in the same way that the LGPL and the MPL are, for example. (The European Union describes the EPL as a strong copyleft license, in its paper describing license compatibility with the EUPL.)
In our view, the copyleft provisions of the EPL gives our community the best of both worlds. Yes, changes and modifications to EPL-licensed code need to be contributed back. This helps ensure that everyone involved is incented to make their contributions back to the platform, and encourages community building. But at the same time, because the EPL (a) does not define simply linking to it as creating a derivative work and (b) allows re-licensing of binaries under commercial terms, it encourages commercial adoption.
LGPL Pain
We have been having a debate internally at the Eclipse Foundation whether we need to add the LGPLv2 to the list of licenses that Eclipse projects can use or pre-req. Despite what you may think, this is not an straightforward conversation.
The most obvious question is why don’t we allow LGPL today? The answer might surprise you because the primary issue is not really legal, it’s business. To date, the licenses that we allow Eclipse projects to include as dependencies in Eclipse projects allow for re-licensing the binaries under a commercial license. This is a huge win for our commercial ecosystem because it means that adopters know that they can take code from Eclipse, embed it in their products and make the result available to their customers under a single software license agreement. Of course, there are also some niggling legal issues, but this is the true crux of the matter.
So to be clear: at Eclipse, the test is not simply whether we can ship a piece of code from Eclipse. The code has to also be usable by our commercial ecosystem in their products.
The flip side, as we have heard, is that in places where there really is no alternative to LGPL licensed code we are causing our adopters to go through cruel and unusual steps to construct a functioning system. I need to hear your pain.
By the way, avoiding the LGPL is not an Eclipse-only viewpoint. The Apache Foundation also excludes it.
So here is my question for the community: how big a deal is this?
I would also like to hear from commercial members as to their experiences with LGPL. Do you use it today in your products? Do your customers see the licensing as an issue?
Here are the cases that I’ve been able to dig up from IPzilla. Please let me know if there are more:
- BIRT wanted to use a version of iText which contained LGPL code. After it was rejected I believe Actuate funded the development necessary to move iText off of the LGPL pieces. As I recall, it delayed the ability to output BIRT reports to PDF by about a year. On the other hand, the project lead thinks it was really worth the effort.
- The Open Health Framework project wanted to use Phonetix and a MySQL driver which were LGPL licensed. Their rejection was actually a factor in forking those projects at OpenHealthTools.org. In other words, those projects left Eclipse.
- Apogee (content management) wanted to use SWTCalendar and JMySpell which were both LGPL. They have had a hard time replacing those components so the project has been somewhat stalled as a result.
- SMILA (semantic search) had a number of components (including beanshell) rejected due to LGPL.
- ACTF wanted to use some IDL code necessary to access accessibility functions on Linux platforms.
- OFMP (open financial market platform) had fastutil rejected due to LGPL.
One closing point: I have heard at various times that the Subversive has been impacted by the LGPL rule. That is an urban legend. By far the greater problem is the TMate license for SVNKit, which is really an almost-GPL disguised to look like a BSD license. There is just no easy way to resolve the SVNKit licensing issue.
Licenses Matter
I’ve been thinking about posting on this topic for a while, and recent posts by Greg Stein and Eric Raymond have finally motivated me to get off my butt and git ‘er done.
As I mentioned recently, the EPL is on a bit of a roll at the moment. And that is a very good thing. However, what I find interesting is that – and this is implicit in the words of both Greg and Eric – is that many in the open source community believe that there are only two interesting positions in the licensing debate: GPL or BSD/Apache. A position which I believe is just plain wrong.
Here is the reason: business models are driven by licensing models. And there are many more business models under the sun than those supported by only those two bipolar licensing positions. In particular, “weak copyleft” licenses such as the EPL are great licenses for those who want to build a ubiquitous software platform with a commercial ecosystem. That is because it allows for commercial licensing of products built on top of EPL-licensed code while also requiring modifications to the platform itself be contributed back to the community. This balance is particularly useful for companies and entrepreneurs that want to create industry platforms.
Of the two, I would have to say that Greg is the most wrong, because he bases his argument on the notion that developers should pick their license based on their personal philosophy. And apparently developers only have personal philosophies that fall into either completely permissive or completely free, with nothing in between. He got it particularly wrong with his closing comment:
Middle-of-the-road licenses like MPL, EPL, and CDDL are wishy-washy. They can’t decide to be permissive, or to maintain Freedom. Choose a philosophy.
Companies and (many) developers do not pick licenses based on a philosophy. They pick them based on their desired business model. I am certainly willing to agree that some people are not interested in thinking through the economic results of their choices. I’m not willing to agree that applies to everyone.
I agree with the content of Eric’s post because find Eric’s economic arguments quite persuasive. I do believe that open source software production is more efficient. But I do not expect that commercially-licensed software will disappear for a very long time, if ever. Personally I believe that the eventual steady state is one where open source platforms provide a commons of infrastructure that supports a wide variety of commercially licensed software and content. However, in one of his comments, Eric pointed to the BSD as the “classic choice”. I would assert that the EPL and similar licenses provide equal, if not better, benefits.
I am a big fan of rational choice economists such as Steven Levitt and Tim Hardford. EPL-like licenses send the correct economic signals to rationally incent the behaviour that I want to see: commercially profitable ecosystems built on top of vibrant open source platforms.
Value of Open Source
Last evening I was in Toronto for the “Value of Open Source” panel at the University of Toronto’s School of Computer Science. The panel was one of a series being organized by the Free and Open Source Software Learning Center (FOSSLC).
The evening was almost more of a conversation than a traditional panel, with moderator Andrew Ross letting the audience largely drive the discussion. But it was a lot of fun and the feedback afterward was very positive.
As an aside, Mark Surman of the Mozilla Foundation was there on the panel. Richard Dice of the Perl Foundation was in the audience. It hadn’t really occurred to me before, but the Executive Directors/Presidents of Eclipse, Mozilla and Perl Foundations are all Canadians. Small world, eh?
The most challenging audience question came from Greg Wilson of UT who wondered about the missing gender in open source: women. I’m not sure of the source of his numbers, but as I recall he said that about 1 in 7 CompSci undergrads are women, while only 1 in 200 active open source committers are. I found those numbers quite startling, although I have certainly heard about the issue before. Although I believe that we have more than 1/200 in the Eclipse committer community, I don’t think we have 1/7. (The Eclipse Foundation itself is 7/17 which is probably close to most small software companies.) Ironically, there is really no easy way for us to tell because gender is not one of the things we track in our committer records. I guess the question to our community is: should we? Is there something the Eclipse Foundation could or should be doing on this front?
Thanks to Andrew Ross and the folks at Ingres for organizing and sponsoring the event. Also, thanks to Karen Reid for hosting the evening.
Best Wishes
As you have probably read by now, this morning Bjorn announced his resignation from the Eclipse Foundation. I know I speak for everyone here at the Foundation and in the broader community that we wish him well wherever his path may take him.
Bjorn has made enormous contributions to the Eclipse community. Our development processes and lifecycle reflect his contributor-first philosophy. EclipseCon and Eclipse Summit Europe operate flawlessly and are world-class developer conferences. He built and led a great team (Anne, Karl, Gabe) in Portland. And those are but a few highlights of what he has accomplished while part of the EMO.
On a personal note, I have to thank Bjorn for all of his support over the past five years. I believe it is true to say that without him I wouldn’t be in this job. For that I am forever grateful.
Bjorn is one of those restless geniuses that never stops challenging your assumptions and forcing you to do better than you thought you could. He will be missed here at the Eclipse Foundation. Hopefully he will land a role somewhere within the Eclipse ecosystem so that his contributions to all of us can continue.
Looking forward, Wayne Beaton is going to move over to the Committer Community role and lead Anne and Gabe’s activities in support of our project and committer community. He will still be involved in evangelizing Eclipse, but now with a tighter tie into the projects.
Best wishes Bjorn.
Life at Eclipse 