November 5, 2025
Packets, puns, and prior art
An eBPF Loophole: Using XDP for Egress Traffic
Startup claims 10x speed, commenters yell 'prior art' and iOS crashes
TLDR: Loophole Labs claims it made Linux’s fast packet tech work on outgoing traffic with 10x speed and no kernel changes. Commenters applaud the engineering but debate originality, citing a 2022 post, while an iOS link crash adds comic relief—important because faster data movement can make cloud apps and migrations smoother.
A wild networking flex just dropped: Loophole Labs says they found a way to make Linux’s fastest packet engine, XDP (a tool that turbocharges how computers handle network data), work for outgoing traffic too. Translation: they claim 10x speed over today’s methods, no kernel changes, and plug-and-play with Docker/Kubernetes. Cue the comment fireworks. The founder jumps in with “happy to answer questions,” but the top drama comes from a heavyweight linking prior art: “You can also use XDP for outgoing packets”, posted back in 2022. That set off the classic “is this new or just a better spin?” battle—some cheer the scale and polish, others ask for receipts, benchmarks, and edge-case proof. The crowd also loved the ironic pun: Loophole Labs found a loophole; the meme machine turned “line-rate” into “line-rage.” One commenter couldn’t even load the page on iOS, turning the thread into a mini roast of mobile browsers. Meanwhile, a practical voice asked if the shift from veth to netkit affects this trick, nudging the convo toward real-world deployability. Verdict from the peanut gallery: impressive engineering, spicy claims, and a heated novelty debate—just the way Hacker News likes it.
Key Points
- •Loophole Labs describes a method to use XDP for egress traffic by exploiting the Linux kernel’s packet-direction determination.
- •The approach achieves around 10x better performance than existing egress solutions and requires no kernel modifications.
- •It is compatible with existing Docker and Kubernetes workloads and aims for line-rate processing at high throughput.
- •XDP traditionally operates only on ingress, while TC is commonly used for egress but suffers performance limitations.
- •Reported TC throughput is about 21Gbps on egress and 23Gbps on ingress due to late execution in the networking stack and sk_buff overhead.