Discussion on GitLab's roles and permissions model

Overview

We had a recent internal discussion on the proposed Operator Role, which evolved into a higher-level discussion of how GitLab should consider building roles and permissions out in the future.

Things we won't do anytime soon

We won't pursue things that optimize for configuration over convention:

• Fully custom roles.
• Widespread or requisite feature-by-feature permissioning.

Questions to answer

• Should we make substantive changes to our existing role system? (e.g. replace it with a different hierarchy, or eliminate a hierarchical role model altogether?)
• When should a new role be introduced?
• When should we build feature-level permissioning controls?