[masarray]

Why

ARSAS and IEDScout field-capture comparison for F650 proves that bounded 64-page MMS GetNameList leaves eveGGIO1, CSWI1–16 beyond the initial namespace; successful DataSet directories already prove their ST/MX membership, but type GVA was only run against initial name-list LN roots. This left Discovery 24 live points vs 67 from Open SCL (8 DataSets / 124 static members and 6/6 RCB ingress matched).

Architecture

Add a pure LiveIedDataSetTypeProbePlanner in engine that admits only live-domain, successful and internally consistent FC=ST/MX DataSet directory members. It extends the association-scoped directory with exact verified member points before the existing, single, negotiated-window LN→DO→leaf type-probe ladder. No second sweep, no per-leaf guess, no cyclic MMS reads. Default limits: 48 extra LN roots and 2048 distinct positional member hints. Full original positional DataSet directories are unchanged; repeated named values are indexed once.

New backwards-compatible overload LiveIedVariableTypeProbeExecutor.ProbeSmartAsync(session, discovery, options, ct) opts into this behavior. The legacy directory overload remains untouched; ARSAS must explicitly adopt the new one after verifying engine CI. Public synthetic tests cover late LN hints, domain/FC spoofing, malformed hints, deduplication, limits and idempotent model expansion.

Delivery/validation

Exact parent 352c81e6a798635c6addcee0683235ca87ad416d (current ARSAS engine pin). No customer IID/PCAP uploaded. Draft, separate from #153/#154 and control/reporting changes. CI and actual F650 parity + MMS budget remain mandatory; do not merge this branch to public main before all gates.

[masarray]

2026-10-22 staging→main audit triage. Keep this PR OPEN (unique/divergent functionality), do not cherry-pick or close as duplicate.

SHA 43f8935eb636967359028a0d9df293aac1812d99. Evidence: This branch has a DISTINCT LiveIedDataSetTypeProbePlanner.cs plus matching Tests.cs, both ABSENT from staging. It also modifies the already advanced LiveIedVariableTypeHierarchy and MmsClientSession.SmartDiscovery. Needed F650 namespace >64-page / DataSet-first late LN evidence cannot be declared equivalent from existing staging tests. Preserve for a separate no-overwrite reconciliation and MMS-budget gate.

Current staging SHA 44ac7554f43f2ba06763f3895242cb569b2e6a1c (CI https://github.com/masarray/ARIEC61850/actions/runs/39782663272 = 1,262/1,262 PASS), main promotion DRAFT PR #192. Reconcile on staging only, with new exact SHA and regression proof. No consumer repin or field qualification claimed. Tracking stays in #144 and #192; no duplicate issue.

[masarray]

2026-10-22 #155 adaptively integrated into newer staging (NOT verbatim original branch merge or Git ancestor). Source HEAD 43f8935eb636967359028a0d9df293aac1812d99 holds distinct DataSet-first functionality. Reviewer-approved surgical integration PR #193 was merged into staging at d086ed5b03aeb45c91ad37cefbc8b5b5627a2718, copying the original LiveIedDataSetTypeProbePlanner.cs and LiveIedDataSetTypeProbePlannerTests.cs Git blob IDs exactly, while adding only a new opt-in ProbeSmartAsync(session, MmsDiscoveryResult) overload and two bounded tuning options to latest engine sources. Did NOT overwrite later negotiated-window RCB/control/SMV/identity code. CI #39787085722 checked out refs/pull/193/merge at aff9968: 1,264/1,264 passed, 0 errors/warnings.

Closing original #155 as functionally superseded without merging it. Original source branch, commits and field hypothesis preserved. GE F650 real-I​​ED #162 not closed: opt-in adoption and physical MMS budget/type parity still unproven. main/ARSAS engine lock untouched. Reopen if specific missing behavior confirmed.