frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: GitDelivr: A free CDN for Git clones built on Cloudflare Workers and R2

https://gitdelivr.net/
1•emirb•1h ago
Hey HN...I built a free CDN to solve a real problem of slow or expensive git traffic. You just prefix a repo URL:

    git clone --depth=1 https://gitdelivr.net/github.com/torvalds/linux
and you get it served from Cloudflare edge servers closest to you, cached in Cloudflare's object storage (R2).

The idea came from reading that GNOME had to redirect git clone traffic to GitHub mirrors because their GitLab bandwidth bill [1]. As a long time open source contributor and FOSS mirror hoster myself, it seemed surprising that this is a thing in 2026. We have jsDelivr for npm, and CDNs for Docker images and ML, but GNOME out of all projects has to move to use GitHub as a mirror.

Technically Git makes this surprisingly natural to cache as the same clone request produces the same packfile, so in my Workers code I hash the request body and use that as the cache key from R2. Refs get a short TTL of 60s.

The nice thing is you don't have to trust the proxy (I know the first question might be how do we trust you?) because git itself verifies every object hash client-side. And the source code is available on GitHub [2]. It's very light and turned out to be a weekend side project to help not just GNOME but anyone either facing a) egress bills b) not willing to move to GitHub or GitLab SaaS due to costs. It's 2026 so nobody should pay for egress if we have nice things like Cloudflare Workers + R2 :)

There's also support for Git LFS. Total cost to run is about $5/month.

I tested it with a dozen of different Git remotes: GitHub and GitLab of course but also, Codeberg, Gitea, and a few self-hosted instances.

This is meant to be used on public repos only, so private repos don't work but you can take the source code and adjust it and deploy it yourself.

If nobody uses it, it was a still fun project to see how far I can push Cloudflare's offerings. (long-time customer, but haven't really played around with Workers)

[1]: https://www.phoronix.com/news/GNOME-GitHub-GitLab-Redirect [2] https://github.com/emirb/gitdelivr.net

Comments

handstitched•1h ago
> You don't have to trust us. Git itself verifies every object by hash on the client side. If we flip a byte, git fsck rejects the entire pack.

If I were to run 'git clone https://gitdelivr.net/$repoUrl` then I would also be getting the Git repository metadata through GitDelivr. You could return any valid git repo, eg. just add one commit on top of the real main with a malicious buildscript. I dont see how this security model works at all?

emirb•1h ago
You're right, that line was overstated in the homepage. I guess security model as with any public mirror in FOSS world?

Git hash checks protect object integrity, so GitDelivr can't silently corrupt a packfile without Git noticing, but that still doesn't make it fully trusted.

Yes the code could still lie about refs/HEAD and serve a different but internally valid history, and Git would accept it. The endgame here is if this is something Cloudflare would pick up (or any other big player) to offer it for free, then you'd trust it because it's a big name (and not a new domain bought yesterday after a weekend project heh)

So the accurate security model is a) GitDelivr preserves Git object integrity b) it does not by itself guarantee authenticity of refs in a way you can verify it c) in that sense it's as close to using any other HTTPS Git mirror or CDN?

re-thc•1h ago
How does this compare to GitHub? Are there benchmarks?

If the problem is performance how much has it improved?

emirb•1h ago
Yes it has improved, mostly thanks to Cloudflare network. From my machine (gigabit fiber), for git clone https://github.com/torvalds/linux --depth=1 --no-checkout a warmed clone through gitdelivr.net took ~5s versus ~18 directly from GitHub (Github seems to cap clones at around 30 MB/s or 250 mbit/s) so about 3x faster in that case.

How to vibe-code a real product in 5 hours

https://www.theaithinker.com/p/how-to-vibe-code-a-real-web-app-in
1•faikadam•58s ago•0 comments

AWS outage in the Middle East: Data center "hit by objects"

https://www.heise.de/en/news/AWS-outage-in-the-Middle-East-Data-center-hit-by-objects-11194454.html
1•tosh•2m ago•0 comments

Show HN: Vibma – let agents create professional design system in Figma, directly

https://github.com/ufira-ai/Vibma
1•ufira•3m ago•0 comments

Sneak peek at the redesigned Stack Overflow

https://stackoverflow.blog/2026/02/25/your-sneak-peek-at-the-redesigned-stack-overflow/
1•jnord•3m ago•0 comments

AMD Am386 released March 2, 1991

https://dfarq.homeip.net/amd-am386-released-march-2-1991/
2•jnord•3m ago•0 comments

Show HN: A visual sitemap tool to see your website structure in seconds

3•epic_ai•4m ago•2 comments

Free Dev Playgrounds

https://labex.io/playgrounds
1•huhuhang•6m ago•0 comments

Three Modes of Cognition

https://kk.org/thetechnium/three-modes-of-cognition/
1•herbertl•7m ago•0 comments

The Looming AI Clownpocalypse

https://honnibal.dev/blog/clownpocalypse
1•birdculture•8m ago•0 comments

Show HN: EasyClaw – One-click installer for OpenClaw AI agent

https://github.com/ybgwon96/easyclaw
1•Singularity26•8m ago•1 comments

Show HN: REP – build-once, with secure runtime env variables for front end apps

https://rep-protocol.dev/
1•olamide226•12m ago•0 comments

Qwen 3.5: 9B, 4B, 2B, 0.8B

https://huggingface.co/collections/Qwen/qwen35
1•tosh•17m ago•0 comments

Deflationary spiral if AI sparks unemployment and only benefits a small elite

https://www.businessinsider.com/ai-layoffs-deflation-us-economy-job-market-outlook-citi-2026-2
3•robtherobber•17m ago•0 comments

Anthropic and Alignment (Ben Thompson)

https://stratechery.com/2026/anthropic-and-alignment/
1•toomanybits•18m ago•0 comments

UK launches consultation asking for views on under-16s social media ban

https://www.bbc.com/news/articles/cvg3vjkx9d7o
3•breve•18m ago•0 comments

Tipping points and ecosystem collapse are the real geopolitical risk

https://news.mongabay.com/2026/02/tipping-points-and-ecosystem-collapse-are-the-real-geopolitical...
1•PaulHoule•18m ago•0 comments

Memgraph Production Telemetry Stack with AWS, ClickHouse, and Grafana

https://memgraph.com/blog/memgraph-production-telemetry-stack-aws-clickhouse-grafana
2•taubek•19m ago•0 comments

Apt Graph Colouring

https://github.com/RyanGibb/apt-graph-colouring
1•ryangibb•20m ago•0 comments

Global emissions from digital technologies were largely obscured

https://www.nature.com/articles/s44458-025-00022-6
3•doener•23m ago•0 comments

Show HN: AgentKeeper – cognitive persistence layer for AI agents

1•thinklanceai•23m ago•0 comments

FX (of Phenoelit) has passed away

https://defcon.social/@thedarktangent/116157827849844661
2•eqvinox•23m ago•1 comments

Show HN: RippleMessenger – A Blockchain-Based Messenger

https://github.com/ripplemessenger/RippleMessengerClient
1•ripplemessenger•24m ago•0 comments

C64: Putting Sprite Multiplexing to Work

https://bumbershootsoft.wordpress.com/2026/02/28/c64-putting-sprite-multiplexing-to-work/
1•ibobev•25m ago•0 comments

Day Roguelike 2026: Day 1

https://www.gridbugs.org/7drl2026-day1/
1•ibobev•26m ago•0 comments

Dial-Up over VoIP with a Commodore Modem from 1985

https://www.gridbugs.org/dial-up-over-voip-with-a-commodore-modem-from-1985/
2•ibobev•26m ago•0 comments

Infrastructure Agents Guide – Design and operate AI agents for infra safely

https://blog.cloudgeni.ai/why-we-open-sourced-our-infrastructure-agent-architecture/
1•davletdz•26m ago•2 comments

Pydantic Monty: A Minimal Python Sandbox for AI Agents

https://pydantic.dev/articles/pydantic-monty
1•jonbaer•26m ago•0 comments

Would appreciate feedback on some light analysis on YC HR tech startups

https://substack.com/home/post/p-186734772
2•aarchitboora•27m ago•0 comments

Show HN: Sift – 10-12 HN discussions daily, ranked by practitioner depth

https://sift.1mb.dev/
1•vnykmshr•27m ago•0 comments

Wearable AI and the Risk of Human Manipulation

https://venturebeat.com/technology/what-if-the-real-risk-of-ai-isnt-deepfakes-but-daily-whispers
2•bonkerbits•27m ago•0 comments