Hello LLVM community,
The review of LP0001 "Introduce a new LLVM process to resolve contentious decisions” ran through June 12, 2020 in this thread on llvm-dev. The discussion raised a few points (including some typos in the proposal) but the feedback was overall very positive. I discussed with the review manager :-), and we decided to accept the proposal.
Following up on some points that were raised in the review:
People seem to agree that this is a significant problem that needs to be solved. There are concerns and questions about the details, but everyone seems willing to take the “let’s start somewhere, then learn and iterate” approach over time.
The most controversial aspect was the question of where proposal discussions should take place: on llvm-dev, a Discourse forum, or on a new mailing list. The community is currently divided on this matter. We concluded that the discussion should stay on llvm-dev right now.
A couple of people raised the question of how review managers are selected, and one raised the concern about potential conflict of interest between the proposal author and the people selected to mediate the review. We believe this is a serious concern, but think that a “trust but verify” approach is a reasonable starting point. If this is problematic in practice, we can introduce a more refined approach in the future.
On the second point, the rationale for the decision is that the center of gravity of the community is currently on the llvm-dev mailing list. Introducing a new mailing list or moving the discussion to Discourse right now would fragment the community. While the review managers agree with the inclusivity problems with using llvm-dev, we believe that solving it should be handled as a follow-on discussion. We specifically recommend that the potential move to Discourse be considered in light of this: people interested in such a move start an RFC about it, and we use the proposal process to decide this if necessary.
Thank you all for participating in this discussion, your time and engagement in this thread, and your passion for helping improve the LLVM community and its processes.