Engineering notes

Engineering notes

How sipnab is actually built and debugged: regressions found and fixed, protocol traps, and the reasoning behind decisions that are hard to see from the outside.

Working notes from building sipnab. Each one is a real problem with real numbers — a regression that shipped, a protocol assumption that turned out to be wrong, a gate that passed when it should not have.

These are not release announcements. The changelog covers what changed. These cover why it was hard, and what the evidence actually said.

  • A 27% regression hid behind a gate built to catch it

    sipnab shipped a quarter of its offline throughput away for ten releases. The benchmark gate ran on every one of them and passed. Here is why, and what the fix says about ratchets.

  • The INVITE that arrived before its keys

    A TLS capture that decrypted everything except the one message that mattered, and reported a NAT problem instead. Three bugs, none of them the race everyone assumed.