Humphrey: On Code Review

When I used to work for Mozilla one of problems with reviews was to discover who at a particular moment can really review the code as opposed to just enforcing a code style. But people who do the valuable reviews spend more time on that and their queues of patches to review grow so one have to wait days or even weeks before getting a review. And with a review queue there is temptation to start rubber-stamping patches as opposed to doing real reviews.

Another drawback of multiple reviewers is that as the code for their component growths the reviewers would have only vague understanding of the codebase they can review. They simply do not follows all the development. For example, before the Tracemonkey JIT for JavaScript was removed from the source tree, nobody really new all the details of how it worked and implications of changes.

I think for Mozilla it would be worth to try to implement the Linux model when all patches for a particular component has to go via one person. Then there is a chance that at least one person does understand the code.