[AngeloR]

I've been thinking about code reviews a lot.. and I think a big problem with existing code reviews is that for most orgs, code reviews were a way to move planning to the end of the development process instead of keeping it up-front.

I think the foundation of the pyramid (API and Implementation semantics) is often discussed in code reviews when it should have almost no place. By the time you get to the code review not only should these have been ironed out and shared with the whole team. What you should be looking at is simply adherence to the plan, and calling out deviations.

I wrote about a small rat about this: https://xangelo.ca/posts/code-reviews-are-failure/

[gunnarmorling]

> is often discussed in code reviews when it should have almost no place. By the time you get to the code review not only should these have been ironed out and shared with the whole team.

That's a fair point. It's my background in open-source and I suppose the specific ways we work in the projects I'm involved with which made me add these things at the bottom. Oftentimes, contributors come up with PRs for new features, and there was no prior discussion about API design etc. (although we highly recommend to have such discussion before starting with the work on a large PR).

I.e. oftentimes a PR review will be the first time I see how something is implemented, how the API looks like etc. It may be different with more controlled settings in in-house development teams adhering to more formally defined processes.

[AngeloR]

This is 100% a side-effect of GitHub. Open Source used to have the additional barrier of mailing lists to detract from "drive-by PRs". The side effect, of course, is that it made it much harder to get involved. GitHub really brought a lot of new eyes on open source software and lowered the barrier to contribute exacerbating (an argument could be made for creating) the drive-by PR problem.