(Vlad has published a Russian translation of this post at http://howtorecover.me/udalenie-nekotorogo-koda. Thanks, Vlad!)
This is another update on our effort to re-license the OpenSSL software. Our previous post in March was about the launch of our effort to reach all contributors, with the hope that they would support this change.
So far, about 40% of the people have responded. For a project that is as old as OpenSSL (including its predecessor, SSLeay, it’s around 20 years) that’s not bad. We’ll be continuing our efforts over the next couple of months to contact everyone.
Of those people responding, the vast majority have been in favor of the license change – less then a dozen objected. This post describes what we’re doing about those and how we came to our conclusions. The goal is to be very transparent and open with our processes.
First, we want to mention that we respect their decision. While it is inconvenient to us, we acknowledge their rights to keep their code under the terms that they originally agreed to. We are asking permission to change the license terms, and it is fully within their rights to say no.
The license website is driven by scripts that are freely available in the license section of our tools repository on GitHub. When we started, we imported every single git commit, looked for anything that resembled an email address, and created tables for each commit, each user we found, and a log that connects users to commits. This did find false positives: sometimes email Message-ID’s showed up, and there were often mentions of folks just as a passing side-comment, or because they were in the context diff. (That script is also in the tools repository, of course.)
The user table also has columns to record approval or rejection of the license change, and comments. Most of the comments have been supportive, and (a bit surprisingly) only one or two were obscene.
whattoremove script finds the users who refused, and all
commits they were named in. We then examined each commit – there were 86 in
all – and filtered out those that were cherry-picks to other releases. We
are planning on only changing the license for the master branch, so
the other releases branches aren’t applicable. There were also some obvious
false positives. At the end of this step, we had 43 commits to review.
We then re-read every single commit, looking for what we could safely ignore as not relevant. We found the following:
- Nine of them showed up only because the person was mentioned in a comment.
- Sixteen of them were changes to the OpenBSD configuration data. The OpenSSL team had completely rewritten this, refactoring all BSD configurations into a common hierarchy, and the config stuff changed a great deal for 1.1.0.
- Seven were not copyrightable as because they were minimal changes (e.g., fixing a typo in a comment).
- One was a false positive.
This left us with 10 commits. Two of them them were about the CryptoDev engine. We are removing that engine, as can be seen in this PR, but we expect to have a replacement soon (for at least Linux and FreeBSD). As for the other commits, we are removing that code, as can be seen in this first PR. and this second PR. Fortunately, none of them are particularly complex.
Copyright, however, is a complex thing – at times it makes debugging look simple. If there is only one way to write something, then that content isn’t copyrightable. If the work is very small, such as a one-line patch, then it isn’t copyrightable. We aren’t lawyers, and have no wish to become such, but those viewpoints are valid. It is not official legal advice, however.
For the next step, we hope that we will be able to “put back” the code that we removed. We are looking to the community for help, and encourage you to create the appropriate GitHub pull requests. We ask the community to look at that second PR and provide fixes.