I’ve published an Internet-Draft proposing standardized “Passive Hot Reload” semantics for servers — allowing configuration/state reloads without interrupting active connections or forcing process restarts.
This isn’t just conceptual: the mechanism is implemented in the FoxyFy Web Server and currently runs in production across 35+ global Points-of-Presence.
The aim is to move away from ad-hoc reload behaviour (SIGHUP chaos, partial reloads, race conditions) toward predictable, interoperable reload semantics.
I’d genuinely value technical feedback:
- Is this solving a real operational pain point?
- How does this compare to existing reload models you rely on (nginx, systemd, HAProxy, etc)?
- Any edge cases, security concerns, or failure modes you’d expect to see addressed?
Happy to go deep on implementation details if useful.
docjojo•1h ago
This isn’t just conceptual: the mechanism is implemented in the FoxyFy Web Server and currently runs in production across 35+ global Points-of-Presence.
Draft: https://datatracker.ietf.org/doc/draft-ahrweiler-hotreload/
Implementation: https://foxyfy.net/
The aim is to move away from ad-hoc reload behaviour (SIGHUP chaos, partial reloads, race conditions) toward predictable, interoperable reload semantics.
I’d genuinely value technical feedback: - Is this solving a real operational pain point? - How does this compare to existing reload models you rely on (nginx, systemd, HAProxy, etc)? - Any edge cases, security concerns, or failure modes you’d expect to see addressed?
Happy to go deep on implementation details if useful.