Notes from the packet path.
These pages are what I kept after a measurement disagreed with the dashboard. Kernel counters, a long TCP flow, a path MTU that was not the NIC MTU. I write them down so I do not have to relearn the same surprise.
- What ss -timem is actually showing
- A tenth of a percent is enough
- Keepalive versus the carrier NAT
- The path MTU is not the NIC MTU
- notsent_lowat and the starved reply
- BBR's first ten round trips
- fq on a virtio nic
- ICMP is a different path
Laboratory Notebooks
- Acoustics & DSP Log — Hardware measurement and real-time audio
- Systems & Network Bench — Linux packet paths and transport telemetry
- AI Notes — Local model evaluation and inference budgets
- Field & Bench — Woodworking, hand tools, and workshop notes
- Shelf Notes — Reading log and marginalia