Hi Victor,
Thank you, this is a useful write-up!
Le 22/09/2017 à 16:26, Victor Stinner a écrit :
I started to list "responsabilities" (is it the correct word?) of a core developer.
First of all, I like how Mariatta summarized a promotion (in an oral discussion that we had). Becoming a core developer doesn't give *power*, but *responsabilities*. (Sorry, I'm not sure about the exact wording, maybe Mariatta can correct me here ;-))
Well, it gives both, which is the only way things can work sanely. If you give power without responsabilities, you are creating a jungle; if you give responsabilities without the power to exert them properly, you are creating frustration and a dysfunctional environment.
I identified the following CPython core developers responsabilities:
- Long term commitement. We someone lands a big chunk of code, we need someone to maintain it for at least the next 2 years. Maybe for the next 10 years. I think that some people sign with their blood to maintain crappy code for their own life, but it's better to not elaborate this part ;-)
Unfortunately we can't evaluate that in advance. Even the person being promoted often does not known whether they'll still be there in 5 or 10 years. Hopefully that's on their horizon, but many factors can interfere.
I, personally, can only think of a couple of cases where a person being promoted core developer vanished a few months after that. It's not a big deal in the grand scheme of things, though it *is* frustrating to spend your time mentoring and promoting someone (which also engages your own responsability, since you're the one vouching that they'll be up to the task) only to see that person produce little to no work as a core developer.
- Review patches and pull requests. While we don't require not expect newcomers to review, we expect that core developers dedicate a part of their time on reviews.
Yes, I believe this is the most important part of being a core developer. What it means is that core developers care about the quality of the whole code base (and also the non-code parts), not only their own contributions to it.
- Know the CPython workflow. Be aware of the pre-commit and post-commits CIs. How ideas are discussed. It's not only about writing and pushing patches.
This part is also required from regular contributors, at least the experienced ones.
- For C developer: know CPython specific issues like reference leaks and the garbage collector. We expect that a core developer write code with no reference leak, right? ;-)
This is no different from regular contributors posting patches with C code in them.
- Good quality patches: proposed changes are good (or almost good) at the first iteration. I'm not sure about this point, but I know a few other developers have this requiurement to promote someone.
Or, if the code isn't good at the first iteration, the author is able to figure it out by themselves and doesn't rush merge it. Of course, nobody is perfect, which is why non-trivial code written by core developers ideally goes through a review phase anyway. But a general sense of what is "in good state for review/merging" vs. "just a draft I'm working on" is indeed, IMHO, preferrable.
- Pushing core means becoming responsible for this code. For regressions, backward compatibility, security, etc.
Yes, this is the whole "know the project's lifecycle" thing. It also includes knowing what to backport or not.
- Something else?
Two things I would add:
Know to be nice and respectful to the others, at least to the extent they're nice and respectful to yourself :-) We don't have a rock-star (or "bro", "wizard", "ninja", whatever the hyperbole of the day is) culture here.
Show a bit of humility towards existing work and try to understand the decisions behind something before deciding to change it all. That said, given Python's current position on the technical evolution and adoption curve, we get less and less proposals for sweeping changes (perhaps not enough, actually, since even when rejected, they help challenge the statu quo).
Regards
Antoine.