Git has no built-in way to browse and display files from a remote repo without cloning. Its closest concept is git ls-remote, but this shows only the hashes for references like HEAD or master, and not the files inside.

I wrote a server to expose a git repo in a new way. Watch the short demo:

How to try it yourself

You can get the code at begriffs/gitftp. It’s currently a proof of concept. Once I’ve added some more features I’ll run a public server to host the project code using the project itself.

The server is written in C and requires only libgit2. It’s small and portable.

Why do it this way?

The standard solution is to use a web interface like GitHub, GitLab, cgit, stagit, klaus, GitWeb, etc. However these interfaces are fairly rigid, and don’t connect well with external tools. While some of these sites also provide RESTful APIs, the clients available to consume those APIs are limited. Also desktop clients for these proprietary services are often big Electron apps.

By serving a repo behind an FTP interface, we get these benefits:

• Web browser supported but not required
• Minimized network traffic, sending just the file data itself
• Supported on all platforms, with dozens of clients already written
• Support for both the command line and GUI

GitFTP reads from a git repo’s internal database and exposes the trees and blobs as a filesystem. It reads from the master branch, so each new connection sees the newest code. Any single connection sees the files in a consistent state, unchanging even if new commits happen during the duration of the connection. The FTP welcome message identifies the SHA being served.

TODOs

This is a proof of concept. I’m putting it out there to gauge general interest. If we want to continue working on it, there are plenty of features to add, like providing a way to browse code at different commits, supporting SFTP so files cannot be changed by a man in the middle, etc etc. See the project issues for more.