Using git as a database seems appealing due to its built-in version history and ease of use, but it often falls short in practical applications. For example, Cargo, the Rust package manager, initially relied on a git-based index from crates.io. As the index expanded, users faced slowdowns, especially in continuous integration (CI) environments where full index downloads were a waste. The introduction of a sparse HTTP protocol allowed Cargo to fetch only necessary metadata, drastically improving performance and reducing the overhead of downloading large repositories. Homebrew encountered similar issues with shallow clones, leading to substantial download sizes and slow updates. After moving to JSON for tap updates in version 4.0.0, Homebrew reduced update times significantly. CocoaPods faced severe delays due to its expanding specs repository, ultimately shifting to a CDN for podspecs, resulting in nearly instant installations and substantial disk space savings. Meanwhile, Nixpkgs has avoided client-side issues by fetching expressions as tarballs, though its massive repository has strained GitHub's infrastructure. Vcpkg, Microsoft's C++ package manager, remains tied to git, struggling with shallow clones and requiring full history for version resolution. Workarounds are cumbersome and don’t address the underlying architecture. In contrast, Go modules transformed dependency management with a module proxy that improved fetching times from 18 minutes to 12 seconds. This shift also enhanced security by reducing reliance on version control systems. Across various tools, the limitations of using git as a database reveal a broader trend: the need for more efficient, HTTP-based solutions.