[coreboot] [RFC] Developer Guidelines: Policy for Gerrit review score of -2

[Peter Stuge]

Paul Menzel wrote:
> this week the Gerrit review score -2 was discussed again in
> #coreboot at irc.freenode.net.
> 
> Several times, during the last two years, a review score of -2 was
> given and dealing with it caused a lot of discussion and friction.
..
> To formalize that a little bit

I think the last thing we need is more formality. I think we need
to understand why -2 has caused problems, and then address that.

Adding a time limit to -2 scores is not helpful, then we can just as
well remove the -2 score completely, which I hope noone thinks is a
good idea.

A review-driven workflow is always slower than a commit-at-will
free-for-all bazaar, and that's a highly desirable feature, because
it increases code quality.

If someone is in a hurry with some change then it is their own
responsibility to invest resources (time, money, or both) into
getting this change reviewed. If they are clever, they invest more
in working carefully with the community *before* creating that
change, because that is *always* a net win.

Put another way: If a reviewer learns about a change for the first
time when the change is pushed, then the person pushing the change
has already made the mistake of not communicating before starting the
work.

If a reviewer already knows about a change when it is pushed, then
the reviewer is able to review this change MUCH faster, and as a
bonus the change is much more likely to be exactly right on the first
try.

Handling this pre-push pre-coding communication well is a distinct
skill from doing the actual coding, and I have seen many many
projects where amazing developers simply aren't able or willing
to do it, and in those cases the solution isn't formalities but
instead mentoring.

I'm not saying that there is no problem. I'm saying that formalities
most likely is not an effective solution.


//Peter