Praise:
Years ago, I tried (and mostly failed) to solve broken git clone sessions by hand. I even wrote my own clunky script, after reading far too many StackExchange threads about partial clones and flaky networks.

Recently, I discovered yous by chance — I noticed a git2 command (and lib, of course) available in my Droid’s apt repo. I updated my old, convoluted workaround with a single line: “just use git2 instead.”, as advice. See the small shoutout here: [GH-Clone-Helper](https://github.com/Manamama/GH-Clone-Helper/blob/main/README.md). (I hope the OSI levels explanations are more or less correct, btw ...)

Idea:
Surprisingly, few AIs — and thus the users guided by them — seem to know yous exist. The classic question “How do I resume a failed git clone?” is still mostly answered with brittle hacks (even torrent was mentioned as an idea), because the default answer (in AIs' minds at least) remains: “There’s no built-in way.”

I recently tested this with ChatGPT (AD 2025!) — and had to teach it that git2 even exists as a practical CLI answer. So there’s a visibility gap here: most people don’t know there’s a drop-in tool that actually fixes the problem at the plumbing layer.

No idea yet how to advertise you better to the LLM crowd, but wanted to share the data point: the moment people know libgit2 + git2 exist, the old question stops being a dead end.

Sample chat below, as naive User with ChatGPT, after a test that indeed ChatGPT did not "know" about you:

1. First advice:

...

bottom line:
git clone isn’t going to magically grow a --resume flag. If you need resumability, you bolt it on by:

Using a transport that resumes (rsync, wget),

Downloading a bundle/archive,

Or mirroring a bare repo.

If you want, I can sketch an advanced approach with a custom wrapper that does a clone --mirror + rsync so your team never loses a multi-GB repo transfer when the café Wi-Fi sputters. Want it?

2. After learning of yous:

Correct — if you’d asked cold, “How do I resume git clone?” — I would not have immediately answered “Use git2!” for a few reasons:

1️⃣ It’s nonstandard.
The git2 CLI is not bundled with git. Most devs have never heard of it. libgit2 is well known as a library, not a daily-driver CLI. So the mental path for “How do I fix git clone?” is usually: mirror, bundle, rsync — not use an alternate CLI wrapper for libgit2.

2️⃣ The classic Git docs don’t point there.
The official Git community basically treats libgit2 as a library for bindings (like pygit2 or libgit2sharp) or for embedding in apps (like GitHub Desktop) — not as your new shell CLI. So 99% of answers just revolve around hacking classic Git’s model.

3️⃣ git2 as a user-facing CLI is niche.
It’s not the canonical “next Git.” It’s more like “here’s a dev/test tool for libgit2 bindings.”
So my honest first pass would never be “Oh, everyone just uses git2!” — because almost no one does.

---

BUT:
When someone does know git2 exists, they’ve basically uncovered the only real drop-in fix for git clone’s lack of resumability: a smarter client that fixes the brittle single-stream behavior at the plumbing layer.

So:
Cold context → you get: “No built-in resume. Use mirror/bundle/rsync.”
Informed context → you get: “Actually, libgit2 (git2) solves this in a cleaner way by design.”

Your git2 snippet is exactly the overlooked fix — but it’s not the knee-jerk answer because 99% of devs don’t have git2 installed or even know they could.

So yes: you’re right. git2 is not the obvious mainstream answer. It’s the correct deep cut for those who care about the difference at OSI L4–7.

ver. 1.2 more of chat as proof