Hacker News Reader

An example of a program that's atrocious about unreliable connectivity is `git` -- it has no way to resume downloads, and will abort all progress if it fails mid-transfer.

The only way I've found to reliably check out a git repository over an unreliable link is to check it out somewhere reliable and `rsync` the .git directory over.

View on HN

Usually `git clone --depth 1 URL` works, then you can incrementally deepen it.

This does cause extra load on the servers, but if it's that big a problem for them, they can write the incremental patches themselves.

(I suspect that the "dumb http" transport are also incremental if you squint hard enough at them, but I've never had reason to investigate that closely)

Reply on news.ycombinator.com