Repository navigation
feat(net): qualify measured path latency with bounded periodic selection - #110
Merged
Merged
Conversation
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
rldyourmnd
marked this pull request as ready for review
October 6, 2026 03:56
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The default Iroh selector keeps a usable direct path primary even when an allowed relay has substantially lower measured RTT. Selector callbacks also occur only on topology events, so a different ranking alone cannot react to later delay changes.
Add explicit local
path_preference: latencywith 5 ms stickiness and one-second periodic selection, retaining the existing default, strict pins and transport eligibility. Carry a narrow, default-None refresh hook in exact published Iroh 1.3 source with supplier provenance and all upstream license notices. Unchanged selections are not reapplied, and obsolete refresh ticks are skipped after a stall. QUIC, TLS, relay wire and identity behavior remain unchanged. Owned Noq already periodically selects validated paths by RTT.The bounded desktop observer records initial selected transmit paths and actual changes using numeric local IDs, kind and sampled RTT. Ordinary RTT updates do not create log chatter; no extra query is added to input/control readers.
Validation covers faster permitted relay selection, stickiness/extreme values, strict configuration, real actor refresh without topology changes, no refresh for default selectors, no missed-tick burst after an isolated actor stall, endpoint shutdown, and strict single-path UDP proxy behavior. Final strict Clippy, network tests (171 macOS/172 Linux), desktop unit tests (106 macOS/119 Linux with two unchanged platform ignores), optimized agents and the full public CI matrix passed. The scoped two-device opt-in pilot retains stable background receive/control evidence; sustained native typing/visible-response acceptance remains open; this PR does not claim a completed latency/stability wave. The eleven newly vendored CodeQL public-key derivation cases were individually reviewed in the committed security report; coverage remains enabled.