SQLite is now in use
Maintained native SQLite, legacy SQL CE conversion, verified recovery copies and compact self-contained installers.
After 15 years, development is moving forward
AR-Cluster 6.6
Modernization and interoperability
The familiar DX cluster, with maintained SQLite storage, safer installation and recovery, stronger login rules, useful diagnostics and an optional read-only API. Trial .205 reduces PC9X topology processing work, strengthens learned-security checkpoint reliability and adds a guarded Installed Apps path for moving one stopped instance to ProgramData.
Original author: Terry Gerdes, AB5K (SK). Recovery and continued development: Max George, NG7M.
With protocol and security collaboration from Dirk Koopman, G1TLH, and the DXSpider team, together with our testing and collaboration contributors.
October 30, 2026 · Trial .205
What is new since .200
Trial .205 is available from the public Trial download page and Tools → Check for Updates → Trial. Existing connection, filter, security and read-only API settings are preserved.
- Lower PC9X topology overhead: four focused changes reuse bounded node reachability, build route unions without temporary per-peer collections, compare retained projection components directly and incrementally reconcile ordinary remote-user additions and removals.
- Exact fallback behavior: structural, route, transport, local-user, identity, expiry and ambiguous changes continue through the full projection path. Offline differential checks matched 2,484 service snapshots and 11,001 equality cases.
- More reliable learned-security saves: valid concurrent events are retained even when callers acquire the admission gate in a different timestamp order. Identity evidence, cooldowns, quarantine and temporary holds keep their original deadlines.
- Guarded ProgramData maintenance: Windows Installed Apps → Modify can move one selected stopped instance after upgrade or repair. Setup verifies identity, ownership, copied data, databases, recovery files and maintenance registration before switching the binding.
During an overnight NC7J soak, a mature .205 sample with a closely matched graph, more users and higher filtering work measured 30.5% lower CPU, 25.3% lower allocations and 21.4% fewer garbage-collection starts than the retained .200 sample. Workloads differ, so these are field observations rather than a universal performance guarantee. Review all changes · Read the sysop notes.
Community & release
AR-Cluster community on Groups.io — sysops and users welcome.
Trial .205 is available now. Use Tools → Check for Updates → Trial or the public Trial download page. Stable remains unpublished.
October 30, 2026 · through Trial .205
Substantial progress since .160
Implemented features, with operator and wider field acceptance kept separate. Read the update details →
Safer user logins
Node identities are reserved for peers; duplicate user reconnect policy and network fullcall limits are separate from abuse controls.
More useful diagnostics
Direct-peer session history, observed user IPs, clearer timeout messages, copyable details and quieter history-save logging.
Protected API and browser view
The optional token-authenticated API supports server-side consumers. The separate AR-WebCluster provides a convenient live browser view.
HamQTH in Security Monitor
Look up a selected valid caller without blocking the monitor. Also: backups, toolbar choices, map neighborhoods and clearer Setup targets.
Familiar commands, useful fixes
Listener-default filter restore, dot-decimal frequency display and clearer login/SSID explanations. Existing saved filters remain your choice.
Research with N3RD and NG7M
Why was this skimmer spot delivered?
Explore real cached evidence and user delivery decisions in AR-Cluster and DXSpider.
Inside SKIMVALID
See which earlier reports established Valid status, how the cache was scanned, and why the incoming spot was delivered.
Inside Qualitee
Follow complete saved collections, receiver selection, frequency voting and remembered Q values through actual user deliveries.
SKIMVALID versus Qualitee
Compare shared evidence, filtering, timing and delivery for similarly configured users. Separate observations explain behavior without ranking decoding accuracy.
Recovery scope
The familiar cluster, with years of work brought forward.
Max George, NG7M, continues Terry Gerdes AB5K’s work with recovered, maintainable source. The change record goes well beyond a rebuild: it documents corrections to peer traffic, history, filtering, callbooks, sockets, background work and operator tools.
ARC6.6 combines CW/RTTY and FT4/FT8 RBN feeds, WCY solar reports, longer RBN spotter callsigns, and FT SNR and grid filtering. Expanded BandMode.dat definitions recognize 2200 m, 630 m and the wider global 60 m spotting range. ARC6.6 also negotiates PC9X between supported ARC6.6 peers, alongside DXSpider and CC-Cluster interoperability. Legacy PCxx, ARC6 ARx and ARC4 paths remain available.
Comparison tests establish parity in the areas reviewed against the latest preserved deployment. Broader release acceptance continues; planned distribution and update work is listed separately below and in the future roadmap.
Defects and measured corrections
Faster history, steadier server operation
Fixture measurements describe specific tested paths. Trial CPU observations are early field results, not a guaranteed speed multiplier.
Cold DX history queries
When the recent cache lacked enough matching spots, SHOW/DX fell back to SQL CE and could scan history slowly. A date index reduced a 20-result human query from 307 ms to 18 ms, and an FT query from 2,538 ms to 23 ms in an isolated copied database. Current builds now use SQLite; these historical measurements explain the earlier correction, not a SQLite performance promise. Restrictive callsign searches can still cost more.
Large history verification
The upgrade verifier previously held three complete DX tables while checking before and after data. It now streams rows and retains the same preservation checks. A 20,000-row synthetic fixture took 836 ms instead of 1,090 ms on the development PC; this is a local directional result, not a promised installation time.
Connections and high-volume feeds
Socket send queues now handle partial writes and bound pending output. Reconnect cleanup rejects work from obsolete sessions. Receive-only RBN connections no longer get blank application heartbeats. Separate CW/RTTY and FT logging switches keep high-volume feed records from overwhelming ordinary Telnet diagnostics.
PC9X topology processing
Trial .205 adds four focused PC92 optimizations for reachability, route unions, component comparison and ordinary remote-user changes while retaining full rebuilds for complex changes. Offline differential checks matched 2,484 service snapshots and 11,001 equality cases. A closely matched overnight NC7J comparison measured lower CPU, allocations and garbage collection; results vary with network and user workload.
FT reports remain an explicit choice.
SET DX FILTER skimmer selects CW/RTTY reports. SET DX FILTER ftskimmer opts into FT4/FT8. Use SET DX FILTER skimmer OR ftskimmer for both, or SET DX FILTER NOT skimmer OR ftskimmer for human spots plus FT. Saved filters are preserved during upgrade.
Implementation status
Recovery workstreams
Work is checked against preserved behavior, with recovered features, intentional improvements and acceptance gaps recorded separately.
Everyday operation
Command and history behavior, filtering, callbook processing and selected operator-interface features have been recovered or improved.
Peer interoperability
Connection handling, reconnects, topology and duplicate suppression are being exercised with AR-Cluster, CC-Cluster and DX Spider.
FT-mode support
Direct FT4/FT8 reception, signed SNR and user opt-in are implemented alongside CW/RTTY. Broader operational acceptance continues.
WCY solar reports
Accepted hourly WCY reports can be delivered live and recalled with SH/WCY. Users can set WCY output and history count independently of WWV; peer forwarding retains PC73 and capable ARx paths.
Windows data paths
After upgrade or repair, Windows Installed Apps → Modify can move one selected stopped instance to its owned ProgramData location. Setup verifies identity, copied content, databases and recovery state before switching the binding; the original data is retained and the instance remains stopped.
Extensive testing, before a public release.
Testing spans spot processing, filtering, duplicate suppression, connection lifecycles and peer interoperability. Completed checks establish their tested scope; they do not replace final operator, installation and release validation.
Trial protection controls
Understand traffic before enforcing limits
Protection controls are available in the Trial, with separate modes for each policy.
Spot and connection controls
Shared spot limits, sender-topology checks, IPv4/IPv6 policies, login and reconnect penalties, command limits and session controls have separate Observe and Enforce modes. Reported origin IPs retain their provenance.
Review decisions and policies
The Security Monitor shows totals, recent identity evidence and cooldown models. Security Lists explains CIDR coverage and validates blocked, Registered and Trusted entries. These lists do not authenticate callsigns.
Read the protection operating guide → View scope and acceptance limits →
Recent desktop and data work
See the connections. Understand the evidence.
Current Trial builds provide combined clean installation and upgrade, clearer connection details, grouped Setup and maintained reference sources.
Country groups and reported users
Switch between alphabetical and CTY Country Prefix views, keep RBN feeds separate, and expand peers for their reported users. Unknown prefix or membership remains explicit.
Live Network Map
A modeless graph shows loops, software evidence, reported user counts and known initiation direction. Zoom, pan, reposition nodes and inspect provenance. It shows this node’s knowledge, not a complete worldwide map.
Maintained sources, preserved policies
HTTPS CTY, SCP, LoTW, directory and Most Wanted downloads are validated and converted for the existing caches. Local additions, custom sources and working data survive failed updates.
Delivered and next
Install, maintain and qualify the next release
Separate installed behavior from future development and download availability.
Combined Setup
Clean install, upgrade, repair and retained-data reinstall with stable instance ownership and recovery. Manual kits now use a compact self-contained ZIP.
Broader qualification
Longer peer/storage soak, more Windows hosts and specific UI/recovery cases continue. An accepted test on one node is not blanket acceptance.
Web administration
The API remains read-only. Full authenticated web configuration and optional bundled web hosting are future work. Manual-entry HamQTH lookup is already available in Security Monitor.
Selected changes
What has changed since ARC6
A short selection from the full change record. The catalogue documents the implementation, evidence and remaining limits.
PC9X peer networks
ARC6.6 peers negotiate PC9X with each other alongside DXSpider and CC-Cluster sessions. Origin-owned PC92 user/IP topology and ordinary PC61 spots coexist with legacy ARC6 and ARC4 protocols.
More useful reports
One server offers CW/RTTY and FT4/FT8 feeds, WCY solar reports and expanded 2200 m, 630 m and 60 m band recognition. FT delivery remains a user choice.
Spot handling across protocols
Modern PC61 spots use the established validation and duplicate pipeline. RBN/skimmer spots, including every spotter ending in -#, stay with local users and are never forwarded to cluster peers.
Faster history and safer updates
Indexed cold SHOW/DX queries, streaming upgrade checks and trace-guided PC92 optimizations reduce work in measured paths. The catalogue separates fixture results from early field observations.
Connections and diagnostics
Bounded socket queues, stale-session rejection and clearer peer and feed diagnostics address failures seen during reconnects and heavy traffic.
Safer instance paths
Trial .205 exposes the verified stopped-instance ProgramData move through Windows Installed Apps → Modify. Recent upgrade checkpoints preserve matching application/data recovery points alongside the original baseline.
Explore the full change record →
Expandable entries, including the updates since .160, covering user benefits, protocols, performance, data safety, threading and operator tools.
Protocol review
Interoperability status
PC9X, legacy PCxx and native ARx paths are evaluated separately. Legacy identity advertisement remains the default; manual builds also offer a per-peer experimental ARC6.6 identity option. Wider compatibility remains under review.
| Server family | Work exercised | Acceptance boundary |
|---|---|---|
| AR-Cluster | Mutual ARC6.6 PC9X negotiation and user/IP topology exchange; legacy ARx, PCxx and ARC4 compatibility. | Broader mixed-version mesh and longer-run acceptance continue. |
| DX Spider | Live PC9X negotiation, PC92 topology and spot handling. | More peer implementations, message paths and topology soak remain to be exercised. |
| CC-Cluster | Live PC9X negotiation and ordinary traffic; legacy frames remain valid on the session. | Wider CC-Cluster mesh routing remains a field-acceptance gate. |
These are scoped observations, not blanket compatibility or release-readiness claims. No endorsement by other cluster authors is implied.
Technical notes
Implementation detail
Expand a topic for recovery details and the remaining acceptance boundaries.
Discuss interoperabilitySource recovery & behavioral comparison
The recovery brings preserved source into a maintainable build and compares behavior with the latest preserved deployment. Targeted fixtures distinguish recovered behavior, inherited limitations and intentional improvements.
The server targets .NET Framework 4.8 (runtime 4.8.1 supported) and x64 using Visual Studio 2026. A later major release could explore modern .NET, cross-platform hosting, Windows service operation and web-based administration. See the future major-release roadmap. A compiled application or matching interface alone is not treated as complete feature parity.
Commands, history & user filters
Work includes command and history recovery, filter composition, wildcard behavior and local-user spot delivery. Comparison fixtures exercise accepted and rejected results as well as familiar command behavior.
Personal output filtering remains distinct from peer acceptance and forwarding policies.
Callbook processing
Current FCC/VE callbook processing has been restored, with comparisons covering lookup results, data refresh and cache consistency. Testing also exercises lookups during dataset updates and recovery from interrupted publication.
Peer connections & duplicate handling
Testing covers fragmented login prompts, initialization, reconnects, topology cleanup and incoming PC11 processing. Matching spots arriving through different peers have exercised the shared duplicate cache and local-user delivery path.
Version-advertisement checks preserve existing protocol identifiers. Remaining topology and connection-direction questions are part of continued acceptance testing across PCxx, PC9X and ARx paths.
Multithreaded processing & server hardening
Hardening addresses coordination between asynchronous socket callbacks, background workers and connection shutdown. Ordered send queues handle partial writes, protect queued buffers from caller changes and bound pending output through per-connection and shared admission limits.
Peer processing rejects queued work from obsolete sessions after disconnect or reconnect, with synchronization around topology changes. Cache refresh and expiration fixes protect newer records from concurrent cleanup. Cleanup paths account for late callbacks and release pending resources as operations finish.
Controlled concurrency fixtures and sustained local workloads exercise overlapping input, slow receivers, reconnects, filter changes and resource cleanup. These checks support the completed corrections; broader production-load and release acceptance continue.
Operator interface, persistence & release preparation
Selected menus, lookup actions, connection-list filtering and disconnect handling have been recovered. Configuration migration, persistence and restart behavior have also been investigated.
Isolated GUI diagnosis and private upgrade/rollback checks have been completed within their tested scope. Complete operator acceptance, clean-machine provisioning and general distribution remain open. The separate Windows client has not received full release acceptance.
Project ownership
Authorship & continuity
Original author
Terry Gerdes AB5K (SK)
The author of AR-Cluster, whose work remains the foundation of this project.
Recovery & continued development
Max George NG7M
Max has received explicit permission from a member of the Gerdes family to continue Terry’s AR-Cluster legacy.
Testing & collaboration
Our contributors
NG7M, N3RD, K1TTT, NY4I, G1TLH, EA3CV, K7CO, KM3T, WC2L, K7AR, W3YQ, K3LR, HA8TKS, KL9A.
Thank you for testing, operational feedback, protocol guidance and collaboration. These are the contributors acknowledged in the application’s Help/About screen.
Protocol & security collaboration
Dirk Koopman G1TLH & the DXSpider team
ARC6.6’s PC9X protocol support and security implementation could not have been completed without the outstanding work of Dirk and the DXSpider developers and contributors. Their shared protocol knowledge, implementation and security work provided the foundation for this effort.
Sharing this work across DX cluster systems helps strengthen security and improve interoperability throughout the worldwide DX cluster network. We gratefully acknowledge the entire team and their continuing contribution to amateur radio.
Review & coordination
Comments and protocol feedback
Comments, protocol guidance, compatibility observations and coordinated testing are welcome. Please share protocol expectations, interoperability concerns or suggestions for coordinated test cases.
AR-Cluster has a Groups.io community group
The AR-Cluster group on Groups.io welcomes both sysops and users. Share setup questions, operating experiences, bug reports and ideas for ARC6.6, and learn from others across the AR-Cluster community.
Comments & feedback
Max George NG7M
[email protected]Recovery, modernization
and interoperability coordination.
Installation and recovery
One Setup for a new node or an existing instance
The combined Trial installer supports clean installation, upgrade, repair, retained-data reinstall and per-instance removal. Each node retains its own identity and data ownership. Setup validates copied data before publication and retains recovery information if an operation is interrupted.
Start with the installation guide and recovery instructions. Trial availability and acceptance limits are documented with each download; Stable remains a separate channel. Windows x64, .NET Framework 4.8 or 4.8.1 and Windows PowerShell 5.1 Desktop are required.
Recent reports and suggestions from Al K7AR, Dave KM3T, Dave K1TTT, Will WC2L and NG7M helped improve installer recovery, connection details and RBN setup. HA8TKS regional-settings feedback helped strengthen locale handling while preserving filter text. Implementation and preparation: NG7M.