Incremental ordinary user add/remove projection
Reuses bounded projection indexes when the node graph, direct transports, local users and identity state are unchanged. Ambiguous or complex changes continue through the full rebuild.
For invited Trial sysops · Build 6.6.205.0 · October 30, 2026
A focused performance, persistence and maintenance update for existing ARC6.6 Trial installations.
Manual Trial package: This compact ZIP contains the complete tested Setup and recovery payload. It is not the automatic-updater package format.
Trial update channel: Existing x64 Trial installations can use Tools → Check for Updates to upgrade through the complete updater/support package.
Short descriptions are listed newest first. These are ARC6.6 server, Setup and read-only API changes delivered after or with Trial .186.
Reuses bounded projection indexes when the node graph, direct transports, local users and identity state are unchanged. Ambiguous or complex changes continue through the full rebuild.
Compares five retained projection components directly while preserving the previous ordinal equality and rebuild decisions, including separator-collision behavior.
Builds the final warm route union directly from cached nodes and current user relationships instead of allocating a temporary destination collection for every peer.
Restores valid concurrently captured security events in chronological private views. Adds a selected-instance Installed Apps action that copies and verifies stopped-node data without deleting the source or restarting ARC6.
Reuses node reachability across user-only membership changes while invalidating on node, route, identity, session and expiry changes. Scripted upgrades also refresh maintenance registration and preserve migrated security data.
Fresh installations enforce the existing spot-submission and total-command thresholds. Upgrades preserve each sysop's saved Off, Observe or Enforce choices and numeric limits.
Persists unexpired identity evidence, login cooldowns, reconnect escalation, exact-IP quarantine, temporary holds and short-lived duplicate fingerprints in an owned per-instance SQLite checkpoint. Downtime counts toward original expiry.
Keeps cache scans outside the shared snapshot metadata lock and filters band/frequency before copying unrelated rows. Existing worker, row, cache-scan and snapshot bounds remain.
Accepts ordinary spaces around callsign-prefix lists; isolates optional ARC4 export failures from successful FCC/VE refresh; revalidates freshly fetched update offers instead of rejecting them solely for legacy expiry; adds KL9A to the shared credit roster.
Both spot endpoints accept a bounded comma-separated callsignPrefixes list combined with existing exact callsign, source, band, mode and paging controls. No wildcard, regular-expression or write authority is added.
Compresses successful JSON responses over 4096 bytes when the client accepts gzip and compression makes the response smaller. Small responses, errors and existing clients remain compatible.
Adds optional source/precision coordinates to spots, recent reports and the recent spotter directory. Prefers explicit local grid overrides and safe fallbacks without per-request geocoding or address exposure.
Adds a bounded cached directory and exact spotter selection before the matching-report window, preserving reporting suffixes and existing authentication, paging and local-only skimmer delivery.
Binds paging snapshots to the requested receiver, location, quality, mode and band selection, using bounded cache-read chunks so unrelated traffic cannot hide older matching reports.
The read-only API can return bounded accepted skimmer reports that remain memory-only under the inherited committed-history duplicate policy. Database history and peer forwarding are unchanged.
Caps the effective minimum healthy-peer count at the number of enabled outgoing peers, keeps the header aligned with the lightweight sampler, and exposes the persisted CT1BOH validation flag through the read-only API.
SH/NODE separates direct peers from remotely reachable nodes and lists alternate first hops. New/reset configurations reject a new duplicate login by default while preserving the incumbent; existing saved policy is retained. Suggestions: KM3T and K1TTT.
Replaces damaged separator characters in clean Setup and lets advanced maintenance collect support logs from the selected installation's maintenance log directory while retaining path and link safeguards.
Offline differentials matched 2,484 topology snapshots and 11,001 projection equality cases. NC7J-3 canary upgrades and the security checkpoint save/reload cycle passed.
NC7J completed a 14.6-hour .205 soak with 380 origins, 429–430 remote nodes, 32 peers and 146–148 users. Compared with the retained, closely matched .200 sample, CPU was 32.5% lower, allocations 25.2% lower and GC starts 19.9% lower. Projection, route-snapshot and reachability allocations were also substantially lower. Results vary with workload and are not a universal performance guarantee.
Security checkpoint and database writes remained current, all configured listeners stayed healthy, capacity rejections remained zero, no recent persistence errors were found and private memory remained comparable to the mature .200 baseline.
For the optional ProgramData move, upgrade first. Then open Windows Installed Apps, select the exact ARC6 instance, choose Modify and select Move data to ProgramData while the instance and its watchdog are stopped. Review the source, destination and final log before restarting.
Compact manual ZIP SHA-256:
93E5C1AF1FB0D596899B7E7745D398524ACAD699B1E4A28D317FE006686814C1