So git-remote-iroh works, but it's basically a toy demoware. I mean, you can push and pull over iroh, which is cool, but it's not this hub-breaking new thing that changes development workflows and defeats corporate hegemony over open-source software. The server mode is still a centralized hub.
I've been thinking this week about where to take it next, if I should at all. Where my head is at:
All repository clones could be remote peers that you can pull from. It should be easy to do this from a background process that spawns and stays running in the background. Could be on-demand, like how podman and gpg-agent do.
Because it's pull-only, it should be safe to serve this up from a clone with a working tree.
Hubs allow pushes, but don't attempt to force a single history. They are a synchronization point for blobs, tracking who has what blobs. All remote refs. What this does, is remove the need for everyone being online all the time.
Who decides what is the true timeline then? Up to a peer that decides to act as a publisher, and who follows it. What's nice about this, is you can have a primary and following replicas. You could even have multiple such primaries, possibly with different policies.
#git #iroh #p2p #nohubsjustspokes
1/
