I taught a bucket to speak Git

Tigris Data is a startup that builds storage infrastructure (object storage + a global database). This blog post describes "ObjGit", an open-source project that lets developers use standard Git commands to push/pull data directly to/from Tigris's object storage (S3-compatible buckets), bypassing a traditional Git server. The key insight: Git's internal object model (blobs, trees, commits) maps naturally onto key-value object storage. By implementing a custom Git transport, ObjGit treats an S3 bucket as if it were a Git remote repository — no separate Git server process needed. This is useful for teams who want Git-like versioning for large files or datasets that are already stored in object storage, or who want cheaper/easier Git hosting without managing a dedicated Git server. The post explains the technical architecture: how Git's packfiles and references are translated to S3 API calls, and the tradeoffs (e.g., no built-in access control beyond S3's IAM policies, no merge request UI).