Epic: IMS runtime — DL/I database calls (PCB/SSA/status codes), PSB/DBD, later IMS TM/MFS #4893

Gap: IMS is refused by name: DL/I database calls (CBLTDLI / EXEC DLI: GU, GN, GNP, GHU, ISRT, REPL, DLET with SSAs, PCBs and status codes), the PSB/DBD definitions, and IMS TM / DC (message processing, MFS screens). The scanner already extracts DL/I calls, IMS segment access and IMS definitions (fact channels), but the translator stops at ENTRY 'DLITCBL'. IMS-centric estates (large banks and insurers) are common and are out of reach today.

Wanted: an IMS runtime: hierarchical database navigation with positioning and status codes from IBM's IMS reference, PSB/DBD-driven access, a stated database content model on both harness sides; IMS TM / MFS as a later stage.

Related: #4514, the scanner's IMS fact channels.

Research basis: 2026-10-11 gap research (runtime inventory, language coverage, issue landscape), saved in the orchestrator handoffs (research-2026-10-11-*). Measure-first: the first child issue of this epic is a reach measurement (how often the class appears and in which constructs), then a design note, then slices that each make programs translate or prove. IBM is the reference, GnuCOBOL / our models the instrument; what IBM leaves open is ASSUMED or REFUSED by name (register).

🤖 Generated with Claude Code