Skip to content

v0.4.1 — experimental cached-peer race

Pre-release
Pre-release

Choose a tag to compare

@TheCryptoDonkey TheCryptoDonkey released this 24 Oct 12:57
1dc2ae6

Experimental library prerelease from 1dc2ae664de46e358e463a1561541ea339d7dc6e, following immutable v0.4.0. Includes PR #22 and the release/CI follow-through in PR #24.

When no safe peer wait remains, opt-in zeroBudgetPeerRace: true starts normal whole-segment HTTP immediately and races only an already advertised cached schema-3 object. Only a complete verified winner reaches the player. Stopping the swarm leaves pending HTTP running; loader cancellation stops both arms. The default stays false. Progressive media bypasses the race, and nonempty HTTP progress locks delivery to HTTP.

Multi-holder origin tails retain their measured per-load recovery reserve and absolute deadline. Bounded, URL-free counters distinguish range failures, timeouts, aborts and wrong-sized bodies. Accepted-contributor holder recovery is shared by ordinary peer-first loads and cached races; unsuccessful or origin-only work receives no peer credit.

Downloads include an installable package tarball, standalone browser bundle, matching WASM, Apple core XCFramework/Swift bindings and Android AAR, with three checksum manifests and build metadata carrying this tagged source. The native Chrome cached-race receipt is now retained in CI and proves both winners, one whole HTTP request per load and reuse of the same connection with synthetic loopback data.

This is not field qualification. Physical devices, public NAT, sustained quality and net hosting benefit remain open. Accepted peer bytes alone do not prove net savings because a losing HTTP request may already have transferred bytes. No public service or pilot was deployed. The npm registry remains unpublished. RelaySwarmKit v0.6.0 retains its published v0.3.0 core pin; these standalone native core assets do not upgrade Kit or add a native WebRTC application.

Validation: all 14 PR checks, 14 tag-pipeline jobs and both merge-source realtime jobs passed. The final hosted relay check and attachment job stayed unassigned for over twenty minutes, so their queued remainder was cancelled. The same audit and public-relay journeys passed locally (zero vulnerabilities; both hash-verified journeys), and the CI-built, CI-verified native archives were uploaded without replacing assets. The tag workflow records that cancelled remainder; it is not an all-green tag run.