Why licensing matters and LLVM


Recently I was talking about LLVM with a colleague and the question arose of exactly why the OpenBSD project has an issue with the LLVM license change that occurred a couple of years ago. As I knew only a cursory level reason as to why this occurred, I started to do a little digging into the history and thought it may be of interest. Here is what I came up with.

The Problem

Initially the LLVM project was placed under the NCSA license which is a very open license, much like the license that BSD uses in general. If anyone isn’t aware of the BSD license model please take some time and research it separate from this entry if you want to fully understand it. There are many licensing models out there and if you think that the Apache 2, GNU, and ISC are all basically the same, I would suggest doing some reading, as that is where much of the reasoning behind this issue stems from, the nuances of the licenses. OpenBSD has a very clearly stated policy at https://www.openbsd.org/policy.html that outlines what they consider acceptable licenses for code that goes into their project.

I Thought Apache 2 License is Free?

It turns out that the Apache 2 license is free, but the change to it is what is actually in question. The one called out by OpenBSD in particular is that it was possible to change the license without the consent of the people who are listed as contributors in the NCSA license. In addition, what Theo expanded on was the LLVM team’s apparent “IANAL but let’s do it anyway” attitude that seems to indicate they really did not consult someone as to if it was possible to change the license mid stream as they did. The email in this case it isn’t a piece of software that a single person contributed to, and they would be available to provide consent to the license change, this software had been created by a team of contributors. What happens when someone at a time in the future who is listed on the NCSA license calls them out for not getting their permission to perform the license type change?

What will be the fallout?

At the end of the day the OpenBSD project has been forced to stick with the last NCSA licensed version of LLVM. They generally have forked products in the past when this occurs and then will move forward in one of two ways. Either they will create their own version as in the case of LibreSSL and PF, or they will move to another piece of open source software as they did in the case of going from GCC to Clang. Unfortunately now this has bitten them there may be a tendency to write their own, but that is a huge undertaking. I tend to think they will not go the route they have with LibreSSL just due to the scale of building out a compiler toolchain, but one never knows.

This is how it ended for ipfilter. It occurred midstream in a release cycle, with no packet filter in place that could be used at the time. One had to be written from scratch and within the timelines. This is how PF was born. from the CVS commit message:

Remove ipf. Darren Reed has interpreted his (old, new, whichever) license in a way that makes ipf not free according to the rules we established over 5 years ago, at www.openbsd.org/goals.html (and those same basic rules govern the other *BSD projects too). Specifically, Darren says that modified versions are not permitted. But software which OpenBSD uses and redistributes must be free to all (be they people or companies), for any purpose they wish to use it, including modification, use, peeing on, or even integration into baby mulching machines or atomic bombs to be dropped on Australia. Furthermore, we know of a number of companies using ipf with modification like us, who are now in the same situation, and we hope that some of them will work with us to fill this gap that now exists in OpenBSD (temporarily, we hope)

Until the next blog, don’t talk about it…

Shut up and Hack!