I really don’t think that would scale at all. A reasonably popular community could have tons of simultaneous posts, and if everyone needs to sync before posting, that would suck. You could probably avoid the worst of it by having posts use uuids, but you’re going to have IO issues at scale. Also, would you need the full repo cloned? That can get big, and you generally only care about recent posts.

Also, if you’re doing the UUID thing, you’d have sort everything every time locally. That’s fine if you only have a few thousand posts, but if you get into millions or billions, it’ll get bad, especially if you’re dealing with files.

Databases solve these problems really well. Even a simple SQLite dB would be much better than a filesystem, like orders of magnitude better.