Post

I’ve been thinking about this too, and because I don’t want to write my own database, I’ve been exploring Git as a generic lazy CAS. Modern features get you really close, and for data that isn’t deeply nested the performance is well within an order of magnitude of other databases.

github.comGit EntsMostly-friendly guardians of your (Git) trees. Git Ents has 16 repositories available. Follow their code on GitHub.

cyber professional@permadeath.com

build seems to be an area ripe for total disruption in this space. imagine a Bazel-like where all components are distributed, mesh-based global caching, worktree oriented, agentic integration - that would've been an intimidating years long buildout. now it seems approachable for a small team

Joey Carpinelli

@jdc.pub

13 Sept

I’ve been really surprised with how far you can take Git’s data model past traditional text tracking. I hope to share more in October, after my (unrelated) Monktoberfest talk. For now, I stand by pinned reposts in git-ents. Unpinned repos are experimental slop unfit for human consumption.

Joey Carpinelli

@jdc.pub

More than anything, I think we’d benefit a lot from storing ALL project information in one data model. I think provenance and signed attestations will be more important as time goes by, and having one VCS makes that easier for compatibility and distribution.

13/9/26, 4:52 am0 reposts1 like

No replies yet.