# Sibling-binary presence census — TARGET1 / TARGET2 — 2026-10-01

Question: does the TARGET2 routine (or TARGET1) exist in any sibling DOS 6.22
binary that could (a) act as an external caller, or (b) prove shared-library
dead code carried by the linker?

Method: `tools/xscan_target.js` — full sliding-window alignment of the pinned
target windows (target2.bin 110 B, target1.bin 66 B, from the canonical
Stage30DT V4 r4 load image) against each candidate binary, reporting the best
byte-alignment score and longest exact substring (>=8 B). All binaries scanned
raw; none carry the EXEPACK marker ("Packed file is corrupt" absent), so raw
scanning is sound.

## Result matrix

| candidate                        | size     | T2 best   | T1 best  |
|----------------------------------|----------|-----------|----------|
| DRVSPACE.BIN (6.22 driver)       |  66,294  | 11/110    |  7/66    |
| DBLSPACE.BIN (6.00 driver)       |  51,214  | 11/110    |   —      |
| SCANDISK.EXE (6.22)              | 124,262  | 11/110    |  7/66    |
| DEFRAG.EXE (6.22)                |  79,065  | 11/110    |  6/66    |
| SETUP.EXE (6.22)                 |  72,842  |  8/110    |  7/66    |
| MSBACKUP.EXE stub (6.22)         |   5,506  | 17/110    | 10/66    |
| SMARTMON.EXE (6.22)              |  28,672  | 15/110    | 15/66    |
| DBLSPACE.SYS (6.00 stub)         |     339  |  6/110    |   —      |
| drv622.load (self, sanity)       | 299,648  | 110/110   |   —      |
| DBLSPACE.EXE 6.00 upgrade (sanity)| 274,484 | 98/110    |   —      |

All sibling scores are noise-level; both targets are absent from every sibling
tested, including both resident drivers. The two sanity rows confirm the
scanner recovers the known homologs (110/110 self, 98/110 in the DOS 6.00
upgrade variant).

## Forensic consequence

- Driver-caller hypothesis: falsified at the binary level (no shared routine
  in either DRVSPACE.BIN or DBLSPACE.BIN).
- Shared dead-library hypothesis: falsified across the scanned sibling set.
  Survives only if TARGET2 is DRVSPACE-unique code (consistent with a
  compression-UI-specific routine — the UI exists in no sibling).
- Combined with dsh's 288-materialization census and the Press-C livelock:
  consistent with TARGET2 being a UI action handler reachable only through the
  console dispatch that the v86 environment never releases.