[kixelated]

Problem

moq-clock's items exported under the bare prefix: now() and origin() were moq_now and moq_origin, which read like moq-net's origin types. moq has since moved moq-clock into kio::clock (moq-dev/moq#7027), and rs/moq-clock is gone from moq's main.

Approach

• The exporter's recipes pin moq at 947c91d0, which has kio::clock, and export kio's clock in moq-clock's place: the rule names its items moq_clock_*.
• The pin is the exporter's own moq_rev in ffi/justfile; just ts's stays where it is (see Decisions needed). ts's moq-pin takes the revision.
• 947c91d0 is the last commit before fix(net): take a reserved group without asking its cursor again moq#7032, past which moq-tokio's Send check overflows the recursion limit that just ffi moq denies.
• Two clock renames stay, since the rule still clashes a function with its type: clock::wall is moq_clock_now_wall and clock::interval is moq_clock_interval_every.
• moq-mux's media Clock held moq_clock, so its now and wall clashed with the clock's. It is renamed to moq_mux_clock, beside moq-mux's other mux_* renames.
• The 288 commits of moq the pin moves over change what the export refuses (moq-net, moq-tokio's QUIC driver, moq-mux, hang, kio), so the config's excludes follow just ffi census. moq.c, moq.cpp, and moq.kt follow moq-net's renames (path::Error, origin::Consumer::request).

Checks: just ffi moq (C, C++, Kotlin against the export) passes; the bench's export.c, export.cpp, and export.kt build against the new header.

Impact

• C names of moq's export, recorded in ffi/MIGRATE.md: moq_now is moq_clock_now (C++ moq::clock_now, Kotlin clockNow), and each of moq-clock's items moves under moq_clock_. moq-mux's Clock is moq_mux_clock. moq-net's upstream renames follow.
• No change to the exporter's own API.

Alternatives

• Bump the one pin both just ts and just ffi share. moq dropped moq-wasm in chore!: remove WASM support moq#7475, which picked the browser's randomness for moq-net on wasm32, so just ts moq and just ts census fail to build getrandom and uuid past it.
• Pin moq's main: just ffi moq fails on the recursion limit lint above.
• Rename kio's clock::now and clock::wall instead of moq-mux's Clock: moq_clock_now is the clock's natural name, and moq-mux's type is the one that borrows the namespace.

Follow-ups

• Waiting: moq-tokio within the recursion limit: the exporter's pin moves past fix(net): take a reserved group without asking its cursor again moq#7032 once moq fixes it.
• Translate moq-net without moq-wasm: the ts pin catches up and the two share one pin again.
• moq's WASM removal also affects the model epic's premise of comparing against moq's WASM build; recorded in the new ts quest for the maintainer.

Decisions needed

1. The split pin. The exporter pins moq ahead of the translator.
   • A (recommended, implemented): keep the split until Translate moq-net without moq-wasm lands.
   • B: hold this PR until the ts quest lands, then bump the one pin.
   • C: fix moq-net's wasm32 randomness for the translator in this PR.
2. moq-mux's Clock C name.
   • A (recommended, implemented): moq_mux_clock, with TIMESCALE as moq_mux_clock_fixed_timescale.
   • B: moq_media_clock.
   • C: keep moq_clock for moq-mux and rename kio's clashing functions instead.

Review

Round 1: quest review, Codex gpt-6.1-sol at medium effort, on head ee324f6d: no findings. Head 48fd8e12 adds only a conflict-free merge of main.

just ci passes on 48fd8e12, and just ffi moq passes on ee324f6d.

🤖 Generated with Claude Code

[kixelated]

Maintainer decision, 2026-11-01: hold this PR until rust2 has one moq pin again, once TypeScript can move to moq's main (quest ts-moq-pin).

(written by Claude Opus 5.5)