Google's recursive version control model
ToolsComments
The claim that this cuts conceptual overhead is questionable. Managing versions at multiple hierarchical levels usually forces the user to track which specific directory level holds the current state; that is its own kind of mental tax.
If the overhead shifts to the directory level, would a hybrid approach solve this? For instance, could a system allow global snapshots for small branches and recursive versioning only for massive sub-directories?
This isn't about simplifying for the average dev. It's a direct response to the failure of Git at the scale of a Google-sized monorepo. Why pretend this is for small projects when it is clearly built to stop the bleed of massive index sizes?
The performance hit on git status in multi-gigabyte repos justifies the move. Reducing the working set to a recursive subset is the only way to keep the UI responsive at that scale.
basically just an evolution of git submodules but integrated into the core.