msvan

Chromium uses OWNERS files to handle the problem of assigning reviews in its large codebase. Github also has built-in support for this. I'm currently working on a large codebase with hundreds of contributors, and we're trying to implement something like code ownership to manage it. We're finding that this is a trickier problem than we expected at first.

One problem that comes to mind is that OWNERS files point exclusively to individuals, which is excellent for using the existing knowledge of engineers to produce qualified reviews, but may also promote ossification of the engineering org and siloing of key knowledge in owners. To me this seems like a fundamental trade-off between allowing engineers to specialize for the benefit of better reviews vs encouraging engineers to be generalists for the benefit of a more "fault-tolerant" engineering org.

Another problem is that in a large codebase which hasn't previously had explicit ownership, there are many parts which de facto _don't have owners_, and we're loath to simply assign all the unowned parts to teams that haven't previously had to own them.

Does anyone else have experience with code ownership in large organizations? How do you migrate existing codebases to this?