Repository navigation
release: NFX 0.1.2, nfxd serves browsers through NATs with Nostr signaling - #92
SovereignTechnology wants to merge 1 commit into
Conversation
…aling nfxd's mesh endpoint (#82, #86) is a full-ICE WebRTC peer that meets browsers in the rooms of the renditions it holds through ordinary Nostr relays (NFX-10 §6), with no tracker and no TURN. With it come --stun (#74), seeders that hold only some renditions (#77), the one-rendition packaging default (#76), --iroh-bind, --iroh-external and --bridge-public for seeders behind a NAT or a DNAT (#47, #48), a fetch deadline that grows with the transfer (#70), and rustix 1.1.6 (#88). The crates nfxd, nfx-node and nfx-media are 0.1.2 (the release workflow reads the version from nfxd's and nfx-media's manifests and requires the tag to name it), with the workspace and desktop dependency pins and both Cargo.lock files to match: five version lines, nothing else. The release notes docs/nfx/releases/v0.1.2.md say what changed, what operators change, what was tested (the NAT test's four cells, the Nostr-only browser and nfxd runs), what is left out (payments, nfxd fetching from browsers, nfxd behind a symmetric NAT, fetch throughput), and how to verify the signed notice. The README's status paragraph says 0.1.2. No locked path moves: the money crates do not depend on these three, locked-compiled.txt is unchanged, and release.yml is unchanged.
|
No actionable comments were generated in the recent review. Walkthrough
Merge Risk:
|
|
All three crate versions, their workspace/desktop pins, and the five lockfile package entries are consistently bumped to 0.1.2 ( Risks
Looks fine to merge |
NFX 0.1.2:
nfxdserves browsers directly, through NATs, with Nostr relays as the only signaling path (its mesh endpoint, #82 and #86, NFX-10 §6), plus--stun(#74), seeders that hold only some renditions (#77), the one-rendition packaging default (#76),--iroh-bind/--iroh-externaland--bridge-publicfor seeders behind a NAT or a DNAT (#47, #48), a fetch deadline that grows with the transfer (#70), and rustix 1.1.6 (#88).nfxd,nfx-nodeandnfx-mediaare 0.1.2 (the release workflow reads the version from nfxd's and nfx-media's manifests and requires the tag to match), with the workspace pin ofnfx-nodeand the desktop pin ofnfxdto match. The same files release: NFX 0.1.1, the relay searches a video's title #27 touched for 0.1.1.version = "0.1.1"→"0.1.2"of one of those three crates:nfx-media,nfx-nodeandnfxdincrates/Cargo.lock, andnfx-nodeandnfxdincrates/desktop/Cargo.lock. No checksum, dependency or other package line changes.nfx-payandnfx-pay-wiredo not depend on these three crates, socrates/ci/locked-compiled.txtis unchanged and needs no re-pin;release.ymlandcrates/ci/locked.sha256are untouched.docs/nfx/releases/v0.1.2.md: what changed; what operators change (three defaults moved: one rendition fromnfx-package, up to 60 s waiting for an iroh relay where 0.1.1 waited 10 s, SIGTERM as a clean shutdown; and a scoped NFX-04 relay cannot be a mesh relay); what was tested (the NAT test's four cells from run 42790887367, the Nostr-only browser andnfxdruns, the Rust and web tests); what it does not do (payments;nfxdfetching from browsers, Hole-punch browsers and NFX clients behind NAT over Nostr (no TURN; STUN from NFX runners) #71;nfxdbehind a symmetric NAT; fetch throughput, nfxd fetches at ~17–26 Mbit/s where the nutflix-pear iroh helper reaches 420–500 on the same links #21); and the signed-notice verify lines.After merge: tag
v0.1.2(annotated, asv0.1.1) on the merged commit, which builds a DRAFT release; signing the notice and publishing stay with the maintainer.Note: main's
checkat 8202c7a failed once on a port race innfxd'sa_seeder_waits_for_an_iroh_relay_that_comes_up_late_and_then_seeds_through_it("Address already in use"). It is not touched here; if it recurs, the job is re-run.Merge pick: squash (one commit).
Summary by CodeRabbit