open design at large scale by solomon hykes
TRANSCRIPT
Open design at large scale
It’s getting people to agree on something
The real value of Docker is not technology
How active is Docker?
4854 PRs(Oct 2014, sub-projects not included)
4854 PRsTotal Merged Average
days openMedian
days openDocker, inc 2410 1995 (82%) 7.2 1
Community 2444 1615 (66%) 9.4 1
(Oct 2014, sub-projects not included)
• Differences are noticeable but reasonable: initial maintainers are by definition employees. “community” includes the long tail of low-quality patch bombs.
• We want to trend towards zero difference• The median pull request has been open 1 day which is insane!
The Human Stack (c) Damien Duportal
Project maintainers
Open Design API
ContributorsGoogle Red Hat Microsoft Docker,
inc.IBM etc.
DGABadvice
The Human Stack (c) Damien Duportal
Open Design API
ContributorsGoogle Red Hat Microsoft Docker,
inc.IBM etc.
Project maintainers DGABadvice
Project maintainers
daemon.go registry.go
BDFL
DGAB
CONTRIBUTORS
libcontainer/
How do we scale this to 10,000 contributors and dozens of sub-projects?
Proposed improvement: PR 9137
BDFL
chief architect chief operator chief maintainer
operators core maintainersarchitects
subsystem subsystem subsystem subsystem
subsystemsubsystemsubsystemsubsystem
(yes, governance changes happen in PRs)
less importantDGABProject maintainers
9137https://github.com/docker/docker/pull/9137
Please give feedback!
Human organizations are programmable.
Bugs can be fixed, scale bottlenecks removed, inefficient code refactored
We can define a new model for open-source
at large scale
Thank you!