The project has a long history of skewing the performance benchmarks and making wild claims like "60% faster than epoll" and the creator gets very defensive whe

The project has a long history of skewing the performance benchmarks and making wild claims like "60% faster than epoll" and the creator gets very defensive when anyone questions it. There are claims about SQPOLL that simply cannot be reproduced. Several ongoing threads about it on the liburing tracker. There have been privilege escalation problems from the start that seem like they aren't being addressed. It feels as though Facebook wants to have something revolutionary at the cost of quality and they're using the Linux kernel to do so, though I realize that's my own hot take. Again, I've been following this (and writing code for it) since I could get my hands on the dev branches. It's only really good for filesystem I/O in its current form as that's the biggest focus they have for it. They (Facebook) care less about other resource types (e.g. sockets). You can choose to believe me or not, I suppose.