- waku/waku_mix/protocol.nim: drop the magic-2 cover-traffic fallback
and the hardcoded 10s epoch. Source cover-traffic totalSlots and
epochDuration from spamProtectionConfig when RLN is on (so cover
emission can't outpace proof minting), and from named waku
constants (WakuCoverTrafficTotalSlots=40, WakuCoverTrafficEpochDuration=60s,
~10 emissions/min/node) when RLN is disabled. Single ConstantRate
CoverTraffic.new call site at the end of the block; the if/else
only sets up spam protection. Addresses PR review comment on
protocol.nim line 102.
- waku.nimble: bump mix-rln plugin to 8ec5dc24 (latest on
feat/cover-traffic-epoch-support: messageId guard + drift-corrected
epoch timer) and pin nim-lsquic to #6d2bc489 (v0.2.0) so libp2p
1.15.3's certificate_ffi keeps finding EVP_PKEY in lsquic_ffi.nim.
- waku/common/option_shims: explain in a header comment that the file
exists because libp2p 1.15.3 dropped Option[T] overloads of
valueOr/withValue from libp2p/utility; can be removed once those are
restored upstream. Addresses PR review comment.
- simulations/mixnet/setup_credentials: drop the unused
SpammerUserMessageLimit constant (the "Higher" comment was wrong
since 3 < DefaultUserMessageLimit=4, and zerokit blocks proof-gen
past the per-user limit anyway, so it could never simulate a
spammer). Addresses PR review comment.
- simulations/mixnet/run_chat_mix{,1}.sh: pass --rln-user-message-limit=4
so the chat client's RLN budget matches the keystores baked at
limit=4, otherwise cover-traffic totalSlots vs RLN-budget mismatch
jams the sim.
Sim verified end-to-end:
- RLN-on (default sim config): PASS, ≥2 proof-verified per node,
cover-traffic metrics non-zero.
- RLN-off (mix nodes only, ad-hoc config): cover-traffic emits at
~13/min/node from the new waku defaults (target ~10/min); the
no-RLN code path is exercised correctly.
- Integrate ConstantRateCoverTraffic from libp2p mix module with default
totalSlots = userMessageLimit (or 2) and 10s epoch
- Add --mix-user-message-limit and --mix-disable-spam-protection CLI flags
with corresponding MixConfBuilder accessors and MixConf fields
- Wrap mixRlnSpamProtection construction so it is skipped when spam
protection is disabled, with a nil guard in setupSpamProtectionCallbacks
- Add waku/common/option_shims.nim restoring valueOr/withValue templates
for std/options (removed upstream by results), and import it across
modules that relied on the old behavior
- Sink chat2mix logs to textlines (stdout) instead of textlines[file] to
work around a chronicles compile-time macro-eval bug under Nim 2.2.4
- Rename ExtendedKademliaDiscoveryParams -> ExtendedServiceDiscoveryParams
to match the kad_disco -> service_discovery rename in nim-libp2p
- Bump nim-libp2p to e1bbda4f6 (PR #2243 "cover traffic with constant
rate") and mix-rln-spam-protection-plugin to 153d0c0 (PR #5 cover
traffic epoch change support); both pre-libp2p_mix-extraction
- Add simulations/mixnet/check_cover_traffic.sh for monitoring
mix_cover_* / mix_slot_* metrics, plus per-node cover-traffic configs
Combines five dep-and-build changes that all flow from the libp2p v2.0.0
upgrade and the move to the extracted libp2p_mix / mix-rln plugin stack:
waku.nimble:
* libp2p: ff8d51857 -> c43199378 (release/v2.0.0 tip; sha-pinned until
vacp2p cuts a v2.0.0 tag).
* Drop the bare `zlib < 0.2` cap — no longer needed by the upgraded
libp2p.
* websock: bare ">= 0.4.0" — replaces the d4cd68b URL+SHA workaround
that pinned through a libp2p commit-specific websock SHA.
* nim-json-rpc: switch to chaitanyaprem/nim-json-rpc#f05fad25 — relaxes
websock cap to allow >=0.4.0. TODO: revert to status-im/nim-json-rpc
once status-im/nim-json-rpc#277 merges and a tag is cut.
* lsquic: bare ">= 0.4.1" (drops URL form).
* Add mix-rln-spam-protection-plugin pin (23b278b4) and nim-libp2p-mix
pin (50c4ab4f — PR #14 HEAD); the plugin pins the same libp2p_mix
SHA so the diamond dep collapses to a single source.
waku/factory/waku.nim:
* Explicit HPService.setup(switch) / AutonatService.setup(switch)
calls. libp2p v2.0.0's Service lifecycle refactor (libp2p#2462)
removed switch.start's auto-setup loop, so any caller that assigns
directly to switch.services (we do) is responsible for calling
setup() themselves. Without it, AutonatService.addressMapper stays
nil and peerInfo.expandAddrs SIGSEGVs during start(). Wrapped in
try/except for ServiceSetupError so a setup failure surfaces as a
logged error rather than a crash.
Build / scripts:
* scripts/build_rln_mix.sh removed and Makefile simplified — librln
is now a single shared archive built from zerokit's `stateless`
features (no separate librln_mix archive).
* simulations/mixnet/build_setup.sh + setup_credentials.nim updated
to use librln_v2.0.2.a directly and run RLN keystore setup before
nodes start.
Validated:
* Cold local-cache nimble setup --localdeps -y.
* wakunode2 and chat2mix link cleanly.
* Mixnet roundtrip sim: [PASS] bob received message from alice.
* RLN proof generation + verification on every in-path mix node:
5 gen_called == 5 verified, 0 SPAM_PROOF_* errors.
* update rendezvous to work with WakuPeeRecord and use libp2p updated version
* split rendezvous client and service implementation
* mount rendezvous client by default
