Hi,
While thinking about how to get more contributors onboard, I identified that one bottleneck is building trust. Currently, a vote to promote a contributor as a core dev requires the approval of almost all active core developers, and this list is quite large (50 people? more?). It takes a lot of time until a contributor is known by enough people to start the process to promote them as a core dev.
My idea is to create subteams which would have less permission than "everything": restrict changes to specific directories. It seems like in practice, we already have such subteams: Documentation, IDLE and asyncio have a dedicated Component in the bug tracker, their own directories, a group of people focused on these components, and even... a dedicated mailing list!
The restriction would be the ability to merge a pull request. Since miss-ilington recently got the power of merging a PR, it makes me think that it's doable to allow "non core" people to merge a change under certain conditions. For example, let's say that a contributor called "Alice" is part of a subteam. If Alice approves a change in the review, added the "approved" label and the CI pass: a bot can merge the change, since Alice is allowed to merge a PR modifying specific directories.If Alice doesn't have the power, the bot may notice Alice that she lacks permission and the bot may remove the label.
IMHO it's better to have two steps to merge: approval in the review and a label, sometimes a change looks good but should not be merged yet, or you don't want to take the responsibility to merge it yourself.
What about current core developers? No change for them, they keep their "super power" to modify everything.
If a member of a subteam shows interest to do more than only working in a subteam, the usual promotion process with a vote on python-committers can be followed.
My expectation is that it will be faster to promote a contributor into a subteam. It's easier to trust someone if they is only allowed to modify some directories. In my experience, people with the same interest find their way to meet other people working on the same topic. The trust is built naturally in a component.
You may see a parallel with the Linux kernel hierarchy and Linux subsystems, but the proposed organization is different. In Python, the tradition is that everybody works in the same repository. I don't want to change that.
What do you think? Do you like the idea of subteams? Is it feasible (the label and the bot things)?
I identified 3 obvious subteams:
- Documentation
- IDLE
- asyncio
Maybe you see more candidates?
Victor
On Thu, Apr 26, 2018 at 10:12 AM Victor Stinner <vstinner@redhat.com> wrote: [..]
I identified 3 obvious subteams:
- Documentation
- IDLE
- asyncio
Sorry, asyncio isn't an obvious choice for me. There are not so many low-hanging fruits left in asyncio except improvements to its documentation. I'm a firm -1 to allow people to merge without Andrew's or my review at this point, almost no PRs are fine when they are submitted (including our own). There's a lot of complexity in asyncio which isn't immediately evident to people who are not working with its internals on a daily basis.
Now, people who report and submit asyncio PRs seem to do that just fine without subteams. Although it's rare to see people contributing more than once, but that's not an asyncio-specific pattern, I see it in every big and complex project I happen to contribute to. Even having a dedicated asyncio mailing list doesn't help to get people to contribute to asyncio more frequently.
Don't get me wrong, Andrew and I would certainly welcome any help we can get, but I'd be against running a public experiment with asyncio to see if 2 of us can handle the management of the new sub-teams idea. Unfortunately 2 of us just don't have capacity for that.
Please pick another project for your idea. Maybe we should try it for documentation first, where we have a lot of core devs who can help with PR reviews and management of "subteams".
Yury
Ok, maybe asyncio is not a good candidate to experiment. I know that asyncio internals are complex, and asynchronous programming is hard.
Sure, the risk of regression in the Documentation is lower :-) But it doesn't mean that we should accept any change in the doc. I already saw people proposing to fix the doc, whereas they misunderstood something and the doc was plain right :-)
Victor
2018-04-26 16:31 GMT+02:00 Yury Selivanov <yselivanov.ml@gmail.com>:
On Thu, Apr 26, 2018 at 10:12 AM Victor Stinner <vstinner@redhat.com> wrote: [..]
I identified 3 obvious subteams:
- Documentation
- IDLE
- asyncio
Sorry, asyncio isn't an obvious choice for me. There are not so many low-hanging fruits left in asyncio except improvements to its documentation. I'm a firm -1 to allow people to merge without Andrew's or my review at this point, almost no PRs are fine when they are submitted (including our own). There's a lot of complexity in asyncio which isn't immediately evident to people who are not working with its internals on a daily basis.
Now, people who report and submit asyncio PRs seem to do that just fine without subteams. Although it's rare to see people contributing more than once, but that's not an asyncio-specific pattern, I see it in every big and complex project I happen to contribute to. Even having a dedicated asyncio mailing list doesn't help to get people to contribute to asyncio more frequently.
Don't get me wrong, Andrew and I would certainly welcome any help we can get, but I'd be against running a public experiment with asyncio to see if 2 of us can handle the management of the new sub-teams idea. Unfortunately 2 of us just don't have capacity for that.
Please pick another project for your idea. Maybe we should try it for documentation first, where we have a lot of core devs who can help with PR reviews and management of "subteams".
Yury
On Thu, 26 Apr 2018 at 07:41 Victor Stinner <vstinner@redhat.com> wrote:
Ok, maybe asyncio is not a good candidate to experiment. I know that asyncio internals are complex, and asynchronous programming is hard.
Sure, the risk of regression in the Documentation is lower :-) But it doesn't mean that we should accept any change in the doc. I already saw people proposing to fix the doc, whereas they misunderstood something and the doc was plain right :-)
While I have no issue with the subteam concept (e.g. we have the import-team on GitHub that automatically get asked for PR reviews for relevant files), doing what you're asking will require some coding which is always hard to get people to do. ;)
The other option is we follow our own traditional practice of granting people commit rights for subsets of the code base and trust them to not overstep their comfort zones. I don't see why we can't do the same for documentation. If we trust them enough to change our docs then we should trust them enough to not touch C code unless they have the appropriate experience.
Victor
2018-04-26 16:31 GMT+02:00 Yury Selivanov <yselivanov.ml@gmail.com>:
On Thu, Apr 26, 2018 at 10:12 AM Victor Stinner <vstinner@redhat.com> wrote: [..]
I identified 3 obvious subteams:
- Documentation
- IDLE
- asyncio
Sorry, asyncio isn't an obvious choice for me. There are not so many low-hanging fruits left in asyncio except improvements to its documentation. I'm a firm -1 to allow people to merge without Andrew's or my review at this point, almost no PRs are fine when they are submitted (including our own). There's a lot of complexity in asyncio which isn't immediately evident to people who are not working with its internals on a daily basis.
Now, people who report and submit asyncio PRs seem to do that just fine without subteams. Although it's rare to see people contributing more than once, but that's not an asyncio-specific pattern, I see it in every big and complex project I happen to contribute to. Even having a dedicated asyncio mailing list doesn't help to get people to contribute to asyncio more frequently.
Don't get me wrong, Andrew and I would certainly welcome any help we can get, but I'd be against running a public experiment with asyncio to see if 2 of us can handle the management of the new sub-teams idea. Unfortunately 2 of us just don't have capacity for that.
Please pick another project for your idea. Maybe we should try it for documentation first, where we have a lot of core devs who can help with PR reviews and management of "subteams".
Yury
python-committers mailing list python-committers@python.org https://mail.python.org/mailman/listinfo/python-committers Code of Conduct: https://www.python.org/psf/codeofconduct/
Agree with Yuri. We have not big amount of non-committer contributions into asyncio, and every non-trivial change requires very careful review.
On Thu, Apr 26, 2018, 17:32 Yury Selivanov <yselivanov.ml@gmail.com> wrote:
On Thu, Apr 26, 2018 at 10:12 AM Victor Stinner <vstinner@redhat.com> wrote: [..]
I identified 3 obvious subteams:
- Documentation
- IDLE
- asyncio
Sorry, asyncio isn't an obvious choice for me. There are not so many low-hanging fruits left in asyncio except improvements to its documentation. I'm a firm -1 to allow people to merge without Andrew's or my review at this point, almost no PRs are fine when they are submitted (including our own). There's a lot of complexity in asyncio which isn't immediately evident to people who are not working with its internals on a daily basis.
Now, people who report and submit asyncio PRs seem to do that just fine without subteams. Although it's rare to see people contributing more than once, but that's not an asyncio-specific pattern, I see it in every big and complex project I happen to contribute to. Even having a dedicated asyncio mailing list doesn't help to get people to contribute to asyncio more frequently.
Don't get me wrong, Andrew and I would certainly welcome any help we can get, but I'd be against running a public experiment with asyncio to see if 2 of us can handle the management of the new sub-teams idea. Unfortunately 2 of us just don't have capacity for that.
Please pick another project for your idea. Maybe we should try it for documentation first, where we have a lot of core devs who can help with PR reviews and management of "subteams".
Yury
python-committers mailing list python-committers@python.org https://mail.python.org/mailman/listinfo/python-committers Code of Conduct: https://www.python.org/psf/codeofconduct/
-- Thanks, Andrew Svetlov
participants (4)
-
Andrew Svetlov -
Brett Cannon -
Victor Stinner -
Yury Selivanov