MACM — the interactive operator dialogue during SINTRAN system generation¶

Binary analysed: D:\ND\BPUN\MACM-1718L.BPUN — the standalone copy, which is the one loaded in Ghidra. It is not byte-identical to the copy on the L distribution floppy (standalone base 0o77120 / 19,273 words vs the floppy's 0o76203 / 19,738 words). Every address in this document is a word address in Ghidra's ram: space for the standalone build. Ghidra's hexdump byte offset is always exactly 2 × word address, which is how data addresses below were derived.

An earlier note claimed the standalone build "lacks the MSTYP strings". That is wrong. [VERIFIED] All of PLEASE DEFINE THE DISC TYPE (MSTYP) !, ENTER MSTYP: ', REMOVABLE OR FIXED (R/F): ', )REDEF, )HENT, 22! and 10,0$ are present in D:\ND\BPUN\MACM-1718L.BPUN. Ghidra's auto-analysis had simply not defined them as strings, so list_strings did not report them. They were found with search_bytes / get_hexdump (7-bit ASCII, high bit clear — the 8-bit form cd d3 d4 d9 d0 for "MSTYP" returns 0 matches).

Marking convention used throughout: [VERIFIED] = quoted bytes or disassembly at a named address. [INFERRED] = reasoning shown. COULD NOT DETERMINE = stated plainly.

Companion tool: E:\Dev\Ronny\NDInsight\tools\boot-floppy\tools\nd_sixbit_decode.py (decodes the ND 30-bit packed 6-bit symbol names; validated against 12 anchors).

1. Command dispatch table¶

[VERIFIED] MACM's command/symbol table is a flat array of 3-word entries:

    word 0 : packed name, bits 29..16 (top 2 bits are entry flags)
    word 1 : packed name, bits 15..0
    word 2 : value — for a command, the handler's word address

Proof by three independent anchors, all of which land on the same mod 3 lattice (0x872A, 0x8760, 0x8763 are all ≡ 0x872A mod 3):

address	words	decoded name	value	pre-existing Ghidra function at that value
`ram:872a`	`1214 4146 9257`	`REDEF`	`ram:9257`	(was unnamed)
`ram:8760`	`0020 5394 9913`	`HENT`	`ram:9913`	`cmd_HENT_restore_core_image` ✔
`ram:8763`	`001c a14d 990b`	`GJEM`	`ram:990b`	`cmd_GJEM_save_core_image` ✔
`ram:8766`	`0001 2090 98ce`	`RBP`	`ram:98ce`	`cmd_RBP_reset_breakpoints` ✔
`ram:8703`	`00e4 53c6 8917`	`9EOF`	`ram:8917`	—

Three of the five values are functions a previous analyst named independently. That is what makes the table format certain rather than plausible.

[VERIFIED] )REDEF is a real command-table entry, name REDEF (0x12144146) at ram:872a, handler ram:9257. Disassembly at ram:925c: LDA *0x92ac ; COPY SA,DB ; JPL 0x92ae ; JPL 0x92af ; SAT 1 ; SAA 0 ; …

[VERIFIED] )HENT is a real command-table entry, name HENT (0x00205394) at ram:8760, handler ram:9913 = the already-identified cmd_HENT_restore_core_image.

COULD NOT DETERMINE — the full table extents. The lattice continues in both directions well past ram:8600 (it is MAC's whole permanent symbol table: opcodes, pseudo-ops and commands share it), and I did not dump all of it. The window ram:86e0–ram:87b6 was dumped and is available in the Ghidra program.

COULD NOT DETERMINE — the 11 three-word slots at ram:872d–ram:874d. These sit immediately after REDEF on the same lattice and their value words (872f, 8732, 8735, 8738, 873b, 873e, 8741, 8744, 8747, 874a, 874d) are the destinations of the )9BYTT parameter copy (§6.5). But their name words do not decode to sensible 6-bit symbols: (0000,0415) (aa3d,033d) (0000,0030) (0000,0000) (0000,0008) (0000,0011) (0000,0348) (0000,0d04) (0000,0010) (0000,0d4d) (0000,3909). I am not going to invent an interpretation. Raw words recorded here for whoever closes it.

2. Every prompt and its accepted input¶

2.1 String inventory (operator-facing, MSTYP dialogue)¶

All [VERIFIED] by hexdump. ND text here is 7-bit ASCII, terminated by an apostrophe ' (0x27); a NUL byte pads to the word boundary.

address	text
`ram:934a`	`\r\nGIVE DISK TYPE AS ONE OF THE FOLLOWING OCTAL NUMBERS:\r\n\r\n 0: DISC-14MB … 24: SCSI\r\n\r\nDISK TYPE: '`
`ram:9701`	`\r\nINITIALIZED FOR: '`
`ram:970b`	`REMOVABLE\r\n'`
`ram:9711`	`FIXED\r\n'`
`ram:9817`	`\r\nPLEASE DEFINE THE DISC TYPE (MSTYP) !\r\n\r\nMSTYP SINTRAN DEVICE NAME\r\n\r\n'`
`ram:983c`	`REMOVABLE OR FIXED (R/F): '`
`ram:984b`	`\r\nENTER MSTYP: '`
`ram:9853`	`\r\n\r\n'` (blank lines)
`ram:9856`	`\r\nREMEMBER THE MACM COMMANDS:\r\n\r\n)REDEF => REDEFINE DISC TYPE\r\n)HENT => GET SINTRAN FROM SAVE-AREA\r\n22! => START SINTRAN\r\n10,0$ => LOAD SINTRAN FROM DISKETTE\r\n\r\nTYPE ANY MACM COMMAND:\r\n\r\n'`
`ram:9597..95aa`	20-word pointer table to the 20 `MSTYP SINTRAN DEVICE NAME` lines (`ram:95ab` … `ram:96fc`)

Other operator-facing strings elsewhere in the image (from list_strings, all [VERIFIED]): ILL. ADDRESS, WHAT?, RANGE EXCEEDED, ILL. INSTRUCTION, ILL. MNEMONIC, ALREADY DEFINED, POSSIBLE FAULT, TABLE FULL, ERROR, ILL. BRF UNIT INITIATION (ram:8431–ram:84e5); UDEF ENTRY, )FILL MISSING, ILL. EXPRESSION, OPTION MISSING, SYMBOL NOT DEFINED, EXT DEFINED, EXT IN ADDRESS ARITHMETIC, \r\n% **** ERROR AT:; REGISTERS: P X T A D L S B, LEVEL, PAGE TABLES: (ram:998d); UNDEF. CORELOAD NUMBER, UNDEF CORELOAD NO:, UNDEF SYMBOLS; WRITE, READ, ERROR, STATUS= (ram:9fb7–ram:9fc2); DISASSEMBLER ERROR, CHECKSUM ERROR, \r\nREAD ERROR, \r\nFEED ERROR.

2.2 Prompt: `DISK TYPE:` — routine `disktype_menu_prompt_and_parse` @ `ram:94ad`¶

[VERIFIED] disassembly:

ram:94ad: 58 19   LDX *0x94c6         ; X := 0x934A  ("GIVE DISK TYPE …")
ram:94ae: ba 19   JPL I *0x94c7       ; call 0x915C  (print string at X)
ram:94af: ba 19   JPL I *0x94c8       ; call 0x9CC3  (read an octal number)
ram:94b0: a8 fd   JMP *0x94ad         ; <- error return: re-prompt
ram:94b1: b1 fc   JAN *0x94ad         ; negative value -> re-prompt
ram:94b2: f3 15   SAX 0x15            ; X := 21
ram:94b3: c6 3d   SKP DA,LST,SX       ; skip if A < 21
ram:94b4: a8 f9   JMP *0x94ad         ; out of range -> re-prompt
ram:94b5: cc 6f   COPY SA, DX
ram:94b6: cc 2f   RADD SA, DX         ; X := 2*answer
ram:94b7: 2e 12   LDD ,X I *0x94c9    ; A,D := table[0x9483 + 2*answer]
ram:94b8: 09 87   STA -0x79,B         ; M[0x8342] := A   (MSTYP)
ram:94b9: cc 4d   COPY SD, DA
ram:94ba: 09 80   STA -0x80,B         ; M[0x833B] := D   (SINTRAN disc-type code)
ram:94bb: a8 08   JMP *0x94c3

Input accepted: an octal number, read by the routine at ram:9CC3. ram:94af is a skip return call: the error path returns to 94b0 (re-prompt), the success path to 94b1. [INFERRED from the JMP *0x94ad at the return address followed by validation code — the standard ND-100 skip-return idiom.]
Valid range: 0 to 0o24 inclusive [VERIFIED] — SAX 0x15 (21 decimal) then SKP DA,LST,SX, i.e. answer < 21. Plus JAN rejects negatives.
Bad input: silently re-prompts from the top, reprinting the whole menu. There is no error message on this path. [VERIFIED — every rejection path is JMP *0x94ad.]
COULD NOT DETERMINE — the terminator character. It is consumed inside the number reader at ram:9CC3, which I did not disassemble.

2.3 Prompt: `ENTER MSTYP:` — routine `mstyp_menu_prompt_and_parse` @ `ram:94cc`¶

[VERIFIED] ram:94cc–ram:94f7 is a straight-line sequence of 21 LDX *lit ; JPL I *0x952b pairs that print, in order: ram:9853 (blank lines), ram:9817 (the PLEASE DEFINE THE DISC TYPE (MSTYP) ! header), then the 19 name lines pointed to by ram:952d…ram:953f, then ram:984b (ENTER MSTYP:).

[VERIFIED] The menu prints 19 lines, not 20. The display list at ram:952d–ram:953f is 95ab, 95b2, 95bb, 95cc, 95e6, 960a, 9624, 962e, 963f, 9648, 9659, 966f, 9678, 968b, 969e, 96b1, 96cd, 96e9, 96fc — it skips ram:95b0, which is the string ?\r\n', the MSTYP 1 entry. The 20-entry lookup table at ram:9597 does include it.

Validation, [VERIFIED]:

ram:94f8: ba 49   JPL I *0x9541       ; call 0x9CC3 (read octal number)
ram:94f9: a8 d3   JMP *0x94cc         ; <- error return: re-prompt
ram:94fa: b1 d2   JAN *0x94cc         ; negative -> re-prompt
ram:94fb: f2 14   SAX 0x14            ; X := 20
ram:94fc: c6 35   SKP DA,LST,SX       ; skip if A < 20
ram:94fd: a8 cf   JMP *0x94cc         ; >= 0o24 -> re-prompt
ram:94fe: f2 01   SAX 0x01            ; X := 1
ram:94ff: c4 35   SKP …,SX            ; compare A with 1
ram:9500: a8 cc   JMP *0x94cc         ; MSTYP == 1 -> re-prompt
ram:9501: 09 87   STA -0x79,B         ; M[0x8342] := MSTYP

Input accepted: an octal number, 0 to 0o23, excluding 1.
MSTYP 1 is explicitly rejected [VERIFIED — the SAX 1 / SKP / JMP re-prompt at ram:94fe–ram:9500], which is exactly the entry whose name string is ? and whose parameter record (ram:9734) is all zeros.
Storage: ram:8342 [VERIFIED — STA -0x79,B with B = 0x83bb, confirmed by LDX *0x9589 at ram:9549 loading the literal 0x8342 directly].
Bad input re-prompts with the full menu; no error message.

2.4 Prompt: `REMOVABLE OR FIXED (R/F):`¶

[VERIFIED] disassembly at ram:9510:

ram:9510: 58 35   LDX *0x9545         ; X := 0x983C ("REMOVABLE OR FIXED (R/F): ")
ram:9511: ba 1a   JPL I *0x952b       ; print
ram:9512: bb 46   JPL I 0x46,B        ; read one character -> A
ram:9513: f5 ae   AAA -0x52           ; A - 'R'
ram:9514: b2 04   JAZ *0x9518         ; 'R' -> A == 0  (removable)
ram:9515: f5 0c   AAA 0x0c            ; net A - 'F'
ram:9516: b3 fa   JAF *0x9510         ; not 'F' -> re-prompt
ram:9517: f1 01   SAA 0x1             ; 'F' -> A := 1  (fixed)
ram:9518: 0a 2b   STA I *0x9543       ; M[0x98B8] := 0 or 1
ram:9519: bb 46   JPL I 0x46,B        ; read (consume terminator)

Input accepted: exactly one character, R or F, upper case only. [VERIFIED — AAA -0x52 is -'R'; -0x52 + 0x0c = -0x46 is -'F'. Lower case r = 0x72 / f = 0x66 would not match either test.]
Anything else re-prompts (JAF *0x9510), with no error message. [VERIFIED]
The answer is stored as an index 0/1 at ram:98b8, used both to pick the library mark and to decide whether to set the FIXED bit (§6.4).
This question is asked only for MSTYP 2 and MSTYP 6 — see §6.3.

2.5 Prompt: `DISK TYPE:` / `ENTER MSTYP:` selection logic¶

[VERIFIED] at ram:94bf:

ram:94bf: cc 65   COPY SL, DA         ; A := return link
ram:94c0: 0a 0a   STA I *0x94ca       ; M[0x98B9] := return link
ram:94c1: 49 87   LDA -0x79,B         ; A := M[0x8342]  (MSTYP)
ram:94c2: b2 eb   JAZ *0x94ad         ; MSTYP == 0 -> ask "GIVE DISK TYPE…"
ram:94c3: fa 90   BSKP 0x2, 0, *0x94c5
ram:94c4: aa 07   JMP I *0x94cb       ; -> 0x9502 (MSTYP already known, skip menu)
ram:94c5: a8 07   JMP *0x94cc         ; -> ask "ENTER MSTYP:"

So the two prompts are alternatives selected by a flag and by whether MSTYP is already non-zero. The return address is parked in ram:98b9 and jumped back to at ram:9587 (LDA I *0x9596 ; COPY SA, DP). [VERIFIED]

3. `10,0$` decoded¶

3.1 Where the string lives¶

[VERIFIED] 10,0$ => LOAD SINTRAN FROM DISKETTE occurs at byte offset 0x1312A (mid-word, word ram:9895), inside the single help string that starts at ram:9856:

130ac  0d 0a 52 45 4d 45 4d 42 45 52 20 54 48 45 20 4d 41 43 4d 20 43 4f 4d 4d 41 4e 44 53 3a
       "\r\nREMEMBER THE MACM COMMANDS:"
130ce  ")REDEF => REDEFINE DISC TYPE"
130ea  ")HENT  => GET SINTRAN FROM SAVE-AREA"
13111  "22!    => START SINTRAN"
1312a  "10,0$  => LOAD SINTRAN FROM DISKETTE"
13152  "TYPE ANY MACM COMMAND:"

3.2 What it is¶

[VERIFIED] 10,0$ is NOT a MACM command.

The evidence is a clean contrast within the very same help text:

)REDEF does have a command-table entry — ram:872a, handler ram:9257.
)HENT does have a command-table entry — ram:8760, handler ram:9913.
10,0$ and 22! have no command-table entry. The command table stores names as 6-bit packed symbols; ,, $ and ! are outside the symbol character set MACM's scanner builds names from, and no table value decodes to anything resembling either string.

[INFERRED] 10,0$ is therefore a command typed at the ND-100 console (MOPC / the microprogram operator communication), not at MACM's ) prompt. MACM prints it purely as a crib sheet for the operator, alongside 22!, so the operator knows what to type after leaving MACM. Reasoning: MACM contains the string but contains no parser for it; the two entries in the list that are MACM commands are both marked with the ) prefix and both exist in the table, and the two that are not marked with ) are both absent from the table.

3.3 What the two numbers are¶

COULD NOT DETERMINE from the MACM binary.

I did not find any code in MACM-1718L.BPUN that parses 10,0$, so I cannot verify from this binary what 10 and 0 mean, nor which character is the executor. Specifically:

I could not confirm whether $ is a load-executor and ! a start-executor — MACM implements neither.
I could not confirm whether 10 is a device number, a load-address, or an ALD (Automatic Load Descriptor) value.
Naming the diskette device number reached by 10,0$ requires the MOPC microcode or the console-emulator, neither of which is this binary.

I will not guess. The honest state is: 10,0$ is a console command whose semantics are outside MACM-1718L.BPUN. To close it, disassemble the ND-100 MOPC microcode or the panel/console handler, not MACM.

3.4 Relation to `)9READ`, `)HENT`, `)9GET` and the BPUN loader¶

[VERIFIED] )HENT = cmd_HENT_restore_core_image at ram:9913 — pulls the SINTRAN core image back from MACM's own save area. Table entry ram:8760.
[VERIFIED, from the pre-existing Ghidra naming] cmd_9READ_load_binary_tape at ram:b474, cmd_9SAVE_coreimage_to_disc at ram:a019, cmd_9GET_disc_to_coreimage at ram:a01c, cmd_GJEM_save_core_image at ram:990b (independently re-confirmed here via its table entry ram:8763).
[INFERRED] 10,0$ is the bootstrap step that gets MACM-plus-SINTRAN into memory in the first place, i.e. it happens before MACM's own commands are available; )9READ / )HENT / )9GET are what MACM offers once it is running. This ordering follows from the help text itself, which lists 10,0$ as "LOAD SINTRAN FROM DISKETTE" and 22! as "START SINTRAN" — i.e. load, then start — and from )HENT being described as "GET SINTRAN FROM SAVE-AREA", a different source.

No IOX instruction reachable from any 10,0$ handler was traced, because no such handler exists in this binary.

4. `22!` — MOPC vs MACM¶

[VERIFIED] the literal 22! => START SINTRAN is at byte 0x13111 (word ram:9888), inside the same help string at ram:9856.

[VERIFIED] 22! is not a MACM command — same argument as §3.2: no entry in the 3-word command table, whose format and contents are proven by REDEF, HENT, GJEM, RBP and 9EOF.

[INFERRED, consistent with the earlier agent's finding] 22! is a MOPC console command meaning "start the program at address 22 octal", handled by the ND-100 microcode/console monitor. I can corroborate the shape of the claim from MACM (it is printed as a reminder and never parsed) but I cannot verify the ! semantics from this binary, because MACM does not implement !.

Summary of the four items MACM lists¶

item	MACM command?	evidence
`)REDEF`	YES	table entry `ram:872a` → handler `ram:9257` [VERIFIED]
`)HENT`	YES	table entry `ram:8760` → handler `ram:9913` [VERIFIED]
`22!`	NO — console/MOPC	string only, no table entry [VERIFIED negative]; MOPC meaning [INFERRED]
`10,0$`	NO — console/MOPC	string only, no table entry [VERIFIED negative]; meaning COULD NOT DETERMINE

5. The full generation dialogue, in order¶

Step status is marked per line. Steps 1–2 are outside this binary.

#	who	what appears / what is typed	status
1	operator	`10,0$` at the ND-100 console — loads MACM+SINTRAN from the diskette	[INFERRED] from the help text; no code in MACM
2	operator	`22!` at the console — starts the loaded program	[INFERRED] likewise
3	MACM	if `M[0x8342]` (MSTYP) is 0 → prints `GIVE DISK TYPE AS ONE OF THE FOLLOWING OCTAL NUMBERS:` and the 21-line list `0: DISC-14MB` … `24: SCSI`, then `DISK TYPE:`	[VERIFIED] `ram:94c2`, `ram:94ad`, string `ram:934a`
4	operator	an octal number `0`–`24`	[VERIFIED] range check `ram:94b2`–`94b4`
4a	MACM	out of range or negative → silently reprints the whole menu	[VERIFIED] `JMP *0x94ad`
5	MACM	translates it via the table at `ram:9483` into MSTYP (`ram:8342`) and a disc-type code (`ram:833b`)	[VERIFIED] `ram:94b7`–`94ba`
3′	MACM	alternative path: prints `PLEASE DEFINE THE DISC TYPE (MSTYP) !`, the header `MSTYP SINTRAN DEVICE NAME`, the 19 name lines, then `ENTER MSTYP:`	[VERIFIED] `ram:94cc`–`94f7`
4′	operator	an octal MSTYP, `0`–`23`, not `1`	[VERIFIED] `ram:94fa`–`9500`
6	MACM	looks up `mstyp_record_ptr_table[MSTYP]` and copies 9 geometry words into the `)9BYTT` block at `ram:8343`–`ram:834b`	[VERIFIED] `ram:9503`–`950b`
7	MACM	if record word 0 == 1 (only MSTYP 2 and 6): prints `REMOVABLE OR FIXED (R/F):`	[VERIFIED] `ram:950e`–`9511`
8	operator	one character, `R` or `F`	[VERIFIED] `ram:9513`–`9517`
8a	MACM	any other character → reprints the question	[VERIFIED] `JAF *0x9510`
9	MACM	if `F`: sets bit 15 (0o100000) in `ram:8347, 8348, 834a, 834b`; and if `M[0x833b]==0o20`, rewrites it to 7	[VERIFIED] `ram:951c`–`9528`
10	MACM	installs the library mark: fetches the packed 30-bit symbol via `record[10 + R/F index]` and writes it, with value −1, into the symbol slot at `0x5389`/`0x538b`	[VERIFIED] `ram:9561`–`9569`
11	MACM	prints blank lines, `INITIALIZED FOR:`, blank lines, then the chosen `MSTYP SINTRAN DEVICE NAME` line (skipping its leading number word)	[VERIFIED] `ram:956a`–`9575`
12	MACM	for MSTYP 2 or 6 only, prints `REMOVABLE` or `FIXED` based on the sign of `M[0x8348]`	[VERIFIED] `ram:9576`–`9583`
13	MACM	prints `REMEMBER THE MACM COMMANDS:` … `TYPE ANY MACM COMMAND:` and returns to the command loop via `ram:98b9`	[VERIFIED] `ram:9584`–`9587`
14	operator	`)REDEF` to go round again, `)HENT` / `)9GET` / `)9READ` to bring in the image, `)9SAVE` / `)GJEM` to write it out	[VERIFIED] table entries; exact ordering of a real session [INFERRED]

Note on step 11's exact rendering: the pointer sequence printed is ram:9853 ("\r\n\r\n"), ram:9701 ("\r\nINITIALIZED FOR: "), ram:9853 again, then the device-name line. The pointers are [VERIFIED]; the resulting screen layout is [INFERRED] from them.

6. MSTYP — the tables and the complete mapping¶

This is the section the whole exercise was for. Every row below is [VERIFIED] from data in the image plus the code that indexes it, and the mapping is cross-checked 21/21 against the two independent name tables (§6.6).

6.1 The `DISK TYPE` menu (21 entries, octal 0–24) — string `ram:934a`¶

 0: DISC-14MB                     11: DISC-70MB
 1: DISC-21MB                     12: DISC-74MB   (DISC-36MB-C (BUTTERFLY))
 2: DISC-23MB                     13: DISC-75MB
 3: DISC-28MB                     14: DISC-140MB  (DISC-2-70MB)
 4: DISC-30MB   (DISC-60MB/DISC-90MB)   15: DISC-2-75MB
 5: DISC-33MB                     16: DISC-288MB-R (DISC-225MB-R/DISC-3-75MB/DISC-4-70MB-R)
 6: DISC-38MB                     17: DISC-288MB-F (DISC-4-70MB-F)
 7: DISC-45MB                     20: DISC-450MB-F (DISC-2-225MB-F/DISC-6-70MB-F)
10: DISC-66MB                     21: DISC-288MB-E (DISC-4-70MB-E)
                                  22: DISC-450MB-N (DISC-2-225MB-N/DISC-6-70MB-N)
                                  23: DISC-288MB-N (DISC-4-70MB-N)
                                  24: SCSI

6.2 The translation table `ram:9483` — DISK TYPE → MSTYP¶

21 entries × 2 words. Read at ram:94b7 with X = 2 × answer. LDD loads A from the first word, D from the second; A → MSTYP (ram:8342), D → disc-type code (ram:833b).

DISK TYPE (oct)	menu name	words	→ MSTYP (oct)	→ `ram:833b`
0	DISC-14MB	`0008 0008`	10	8
1	DISC-21MB	`0008 0009`	10	9
2	DISC-23MB	`000a 000a`	12	10
3	DISC-28MB	`000c 000c`	14	12
4	DISC-30MB	`0006 0010`	6	16
5	DISC-33MB	`0003 0002`	3	2
6	DISC-38MB	`0004 0011`	4	17
7	DISC-45MB	`0009 000b`	11	11
10	DISC-66MB	`0003 0003`	3	3
11	DISC-70MB	`0004 0012`	4	18
12	DISC-74MB	`000b 000d`	13	13
13	DISC-75MB	`0004 0013`	4	19
14	DISC-140MB	`000d 0014`	15	20
15	DISC-2-75MB	`0007 0016`	7	22
16	DISC-288MB-R	`0005 0017`	5	23
17	DISC-288MB-F	`000e 0019`	16	25
20	DISC-450MB-F	`0010 001a`	20	26
21	DISC-288MB-E	`000f 000f`	17	15
22	DISC-450MB-N	`0011 001c`	21	28
23	DISC-288MB-N	`0012 001d`	22	29
24	SCSI	`0013 001e`	23	30

COULD NOT DETERMINE what ram:833b means beyond "a per-disc-type code that the FIXED answer remaps from 0o20 to 7" (ram:9524–9528). Note that the only entry with 833b == 16 is DISK TYPE 4 / MSTYP 6, which is exactly one of the two MSTYPs that ask R/F — a satisfying internal consistency, but the field's purpose is still open.

6.3 The MSTYP record table¶

Pointer table ram:9715, 20 words, indexed directly by MSTYP. [VERIFIED] ram:9503–9505: LDA -0x79,B ; LDX *0x9542 (=0x9715) ; RADD SA,DX ; LDX 0,X.

Record layout (11 words, or 12 when there is an R/F variant):

word	meaning	evidence
0	1 ⇒ ask REMOVABLE/FIXED; 0 ⇒ don't	`ram:950e` `LDA 0,X` / `ram:950f` `JAZ *0x9546` [VERIFIED]
1	SINTRAN device number	copied to `ram:8343` by `LDF 1,X ; STF -0x78,B` [VERIFIED]
2–3	`0o30000`, `0o30000` (constant in every record)	[VERIFIED]
4	`0o100` (constant in every record)	[VERIFIED]
5–6	geometry pair; bit 15 set on FIXED	copied to `ram:8347/8348` [VERIFIED]
7	`0o175777` (constant in every record)	copied to `ram:8349` [VERIFIED]
8–9	geometry pair; bit 15 set on FIXED	copied to `ram:834a/834b` [VERIFIED]
10	pointer to packed 30-bit library-mark symbol (removable / default)	`ram:9561`–`9566` [VERIFIED]
11	same, FIXED variant — present only when word 0 == 1	selected by `ram:98b8` = 1 [VERIFIED]

COULD NOT DETERMINE the individual meaning of words 2–9 (geometry: almost certainly sectors/track, words/sector, cylinder counts and similar). Only three distinct parameter sets exist:

set	used by	w2..w9 (octal)
DRUM	MSTYP 0	`30000 30000 100 4000 4000 175777 2000 40`
Winchester-A	MSTYP 2	`30000 30000 100 2000 2000 175777 1000 10`
standard	MSTYP 3 – 23	`30000 30000 100 400 400 175777 200 2`

6.4 The library marks¶

[VERIFIED] Eight 2-word packed symbols at ram:9807–ram:9816, decoded with nd_sixbit_decode.py:

address	words	symbol
`ram:9807`	`0011 254d`	`DRUM`
`ram:9809`	`0000 0000`	(none — MSTYP 1's placeholder)
`ram:980b`	`1214 d3d6`	`REMOV`
`ram:980d`	`0625 8144`	`FIXED`
`ram:980f`	`0213 2e38`	`BD288`
`ram:9811`	`0210 6258`	`BDFIX`
`ram:9813`	`17e0 9383`	`W8INC`
`ram:9815`	`130c 14c9`	`SCASI`

All six marks the brief asked about are accounted for, plus DRUM and the null entry. [VERIFIED] installation code:

ram:9561: 5a 29   LDX I *0x958a       ; X := M[0x98B7]  (record base address)
ram:9562: 4a 29   LDA I *0x958b       ; A := M[0x98B8]  (0 = removable, 1 = fixed)
ram:9563: cc 2f   RADD SA, DX         ; X := base + R/F index
ram:9564: f7 0a   AAX 0xa             ; X := base + index + 10
ram:9565: 5c 00   LDX 0x0,X           ; X := record[10 + index]  = mark pointer
ram:9566: 2c 00   LDD 0x0,X           ; A,D := the packed 30-bit name
ram:9567: 22 25   STD I *0x958c       ; store name at 0x5389/0x538A
ram:9568: f1 ff   SAA -0x1
ram:9569: 0a 24   STA I *0x958d       ; store value -1 at 0x538B  (mark = TRUE)

The value −1 is what makes it a set library mark. Exactly one mark is installed per generation run.

[VERIFIED] the FIXED flag (ram:951c–9523): when the answer is F, bit 15 (0o100000) is set in four words — ram:8347, ram:8348, ram:834a, ram:834b — by two LDD / BSET 0xf,DA / BSET 0xf,DD / STD pairs. The earlier "FR=100000 marks fixed disc" note is therefore correct but incomplete: it is four parameter words, not one, and it is in addition to selecting the FIXED / BDFIX mark. ram:8348 is the word later tested (ram:957c LDA I *0x9592 ; JAN) to decide whether to print FIXED or REMOVABLE.

6.5 Where the parameters land¶

[VERIFIED] with B = 0x83bb (derived from LDX *0x9589 at ram:9549 loading the literal 0x8342, which is B-0x79):

address	source	note
`ram:833b`	disc-type table word 1	`0o20` → 7 when FIXED
`ram:8342`	MSTYP	the operator's answer
`ram:8343`	record w1	device number
`ram:8344`–`8346`	record w2, w3, w4
`ram:8347`, `8348`	record w5, w6	+bit15 when FIXED
`ram:8349`	record w7
`ram:834a`, `834b`	record w8, w9	+bit15 when FIXED

On the earlier claim that the )9BYTT parameter storage is at ram:834c: REFUTED as stated. [VERIFIED] The block written by the MSTYP dialogue is ram:8342–ram:834b (ten consecutive words), plus ram:833b. ram:834c is one word past the end of it. The routine at ram:9546–ram:955f then copies those same ten words plus ram:833b out to eleven stride-3 slots starting at ram:872f (B = 0x872d), i.e. into the value fields of eleven consecutive symbol-table entries — which is how the SINTRAN source gets to reference them.

6.6 THE MSTYP MAPPING — the deliverable¶

MSTYP (oct)	record @	SINTRAN device name (`ram:9597[i]`)	device no. (oct)	library mark(s)	R/F asked?
0	`ram:9729`	`0 DRUM`	0o540	`DRUM`	no
1	`ram:9734`	`?` (rejected on input)	0	none	no
2	`ram:973f`	`2 DISC-10MB-1`	0o500	`REMOV` / `FIXED`	YES
3	`ram:974b`	`3 DISC-33MB-1, DISC-66MB-1`	0o1540	`BD288`	no
4	`ram:9756`	`4 DISC-38MB-1, DISC-70MB-1, DISC-75MB-1`	0o1540	`BD288`	no
5	`ram:9761`	`5 DISC-288MB-1-R, DISC-3-75MB-1, DISC-225MB-1-R`	0o1540	`BD288`	no
6	`ram:976c`	`6 DISC-30MB-1, DISC-60MB-1, DISC-90MB-1`	0o1540	`BD288` / `BDFIX`	YES
7	`ram:9778`	`7 DISC-2-75MB-1`	0o1540	`BD288`	no
10	`ram:9783`	`10 DISC-21MB-1, DISC-14MB-1`	0o500	`W8INC`	no
11	`ram:978e`	`11 DISC-45MB-1`	0o500	`W8INC`	no
12	`ram:9799`	`12 DISC-23MB-1 DISC-16MB-1`	0o500	`W8INC`	no
13	`ram:97a4`	`13 DISC-74MB-1 DISC-36-C (BUTTERFLY)`	0o500	`W8INC`	no
14	`ram:97af`	`14 DISC-28MB-1`	0o500	`W8INC`	no
15	`ram:97ba`	`15 DISC-140MB-1-F, DISC-2-70MB-1-F`	0o1540	`BD288`	no
16	`ram:97c5`	`16 DISC-288MB-1-F, DISC-4-70MB-1-F`	0o1540	`BD288`	no
17	`ram:97d0`	`17 DISC-288MB-1-E, DISC-4-70MB-1-E`	0o1540	`BD288`	no
20	`ram:97db`	`20 DISC-450MB-1-F, DISC-2-225MB-1-F, DISC-6-70MB-1-F`	0o1540	`BD288`	no
21	`ram:97e6`	`21 DISC-450MB-1-N, DISC-2-225MB-1-N, DISC-6-70MB-1-N`	0o1540	`BD288`	no
22	`ram:97f1`	`22 DISC-288MB-1-N, DISC-4-70MB-1-N`	0o1540	`BD288`	no
23	`ram:97fc`	`23 SCSI`	0o144300	`SCASI`	no

Cross-check: every one of the 21 rows of §6.2 lands on an MSTYP whose device-name line in this table lists the disc the operator picked. All 21 agree. For example DISK TYPE 5 (DISC-33MB) → MSTYP 3, whose line reads DISC-33MB-1, DISC-66MB-1; DISK TYPE 10 (DISC-66MB) → also MSTYP 3. DISK TYPE 24 (SCSI) → MSTYP 0o23, device 0o144300, mark SCASI. That 21-for-21 agreement between two tables that are consumed by different code paths is what makes this mapping certain.

6.7 The SCSI device number 0o144300 as a bit pattern¶

[VERIFIED] the raw word: ram:97fd = c8c0 = 0o144300 = 1100 1000 1100 0000. It occurs exactly once in the whole image (search_bytes c8 c0 → 1 match).

[VERIFIED, structural] 0o144300 cannot be an IOX device address. The ND-100 IOX instruction carries an 11-bit device field (max 0o3777); 0o144300 needs 16 bits. A device number this large is only reachable through IOXT, which takes the number in the T register.

COULD NOT DETERMINE any further decomposition of 0o144300 into sub-fields. I found no code in MACM that manipulates it — it is a constant that MACM merely hands to the SINTRAN assembly as a symbol value. One observation worth recording without over-reading it: bit 15 (0o100000) is already set in 0o144300, and bit 15 is also the FIXED flag applied to other parameter words — but MSTYP 0o23 has record word 0 == 0, so the FIXED path never runs for SCSI and the two never interact. Whether the coincidence is meaningful: COULD NOT DETERMINE.

7. Does MACM speak SCSI?¶

Short answer: NO. MACM-1718L.BPUN (standalone D:\ND\BPUN build) is not SCSI-aware. It performs blind block I/O to the SMD disc controller and trusts the MSTYP the operator typed.

7.1 No IOXT anywhere — the decisive negative¶

[VERIFIED] search_bytes d3 0d (the ND-100 IOXT opcode, 0o150415 = 0xD30D) over the entire program returns 0 matches.

Combined with §6.7 — 0o144300 needs 16 bits and IOX only carries 11 — this is conclusive: MACM cannot address device 0o144300 at all. It has no instruction in it that is capable of doing so.

(Caveat, stated for honesty: the numeric value of the IOXT opcode, 0o150415, is ND-100 ISA knowledge I brought to the search, not something I re-derived from this binary. The search itself and its zero result are verified.)

7.2 What MACM actually drives: the SMD controller, device group 0o1540¶

[VERIFIED] disassembly at ram:a46c — Ghidra's own ND-100 device tables label the registers:

ram:a46c: 08 14   STA *0xa480
ram:a46d: f1 00   SAA 0x0
ram:a46e: f2 12   SAT 0x12
ram:a46f: c3 b0   RDIV ST
ram:a470: d9 08   SHA 0x8
ram:a471: cc 0d   RADD SD, DA
ram:a472: eb 63   IOX 0x0363          ; SMD1_LOAD_BLOCK_ADDR   (= IOX 0o1543)
ram:a473: f1 00   SAA 0x0
ram:a474: f8 fd   BSET 0xf, DA
ram:a475: eb 65   IOX 0x0365          ; SMD1_LOAD_CONTROL      (= IOX 0o1545)
ram:a476: 48 0a   LDA *0xa480
ram:a477: eb 63   IOX 0x0363          ; SMD1_LOAD_BLOCK_ADDR
ram:a478: f4 02   AAB 0x2
ram:a479: 50 08   LDT *0xa481
ram:a47a: f1 04   SAA 0x4
ram:a47b: fc dd   BSTA 0xb, DA
ram:a47c: eb 65   IOX 0x0365          ; SMD1_LOAD_CONTROL
ram:a47d: cc 62   EXIT

This is a plain "load block address, load control word, go" sequence — a dumb block driver, no command descriptor, no protocol phase handling.

[VERIFIED] occurrence counts across the image: * eb 65 (IOX 0o1545, SMD control) — 20+ sites, including ram:a0a2, a340, a475, a47c, a485, a49b, a4f3, a4fa, a505, a51c, a78c, a7eb, a931, a9c8, a9cd, aa16, aa1b, ac74, ac7e, ac81. * eb 63 (IOX 0o1543, SMD block address) — present, e.g. ram:a472, a477. * e9 40 (IOX 0o500, the Winchester device) — 4 sites: ram:af09, af0b, c2d9, c2db.

So )9SAVE / )9GET and the core-image engine at ram:9d27 / ram:9d4b reach the disc through IOX to device group 0o1540 (SMD) and 0o500 (Winchester) — the same two device numbers that appear as record word 1 for MSTYP 3–23 and MSTYP 2/10–14 respectively. [INFERRED] that is not a coincidence: MACM can only save to the disc types it can itself drive, and SCSI is not one of them.

7.3 No SCSI Command Descriptor Blocks¶

[VERIFIED negative] Searching for the CDB opcode constants in the packed high-byte form the SINTRAN kernel uses: * 25 00 (READ CAPACITY(10), 0o022400) — 11 byte matches, all at odd half-word offsets (ram:7f6d.1, 8175.1, 9b2b.1, 9c27.1, be4b.1, be85.1, bf47.1, c70a.1) except three, none of which sit in a CDB-building context. * 12 00 (INQUIRY) — 18 byte matches, again overwhelmingly at .1 half-word offsets, i.e. they are not aligned 16-bit constants at all.

No aligned INQUIRY / READ CAPACITY / READ(6) / READ(10) / WRITE(6) / WRITE(10) / TEST UNIT READY / REQUEST SENSE / MODE SENSE constant was found in a CDB-construction context. NOT FOUND [evidence noted 2026-08-02: the negative turns on three ALIGNED 25 00 matches that are dismissed without addresses, disassembly, or a stated criterion for "CDB-building context" - and the passage says "all at odd offsets ... except three", which is self-contradictory. The conclusion is independently supported by section 7.1 (IOXT opcode: 0 matches); this bullet's own evidence does not carry it. To settle: list the three aligned addresses and show what reads or stores them.]

7.4 No disk name, vendor ID or product ID¶

[VERIFIED] Every device name MACM ever prints comes from its own static string table at ram:95ab–ram:96fc, selected by the MSTYP the operator typed (ram:9570–9575: LDA I *0x9589 ; ADD *0x9591 ; COPY SA,DX ; LDX 0,X ; AAX 1 ; JPL I *0x958f). There is no code that reads an ASCII vendor/product field off a device, and no comparison of any such string.

[VERIFIED negative] The complete operator-facing string inventory (§2.1) contains no WRONG DISC, ILLEGAL DISC, DISC NOT READY, NOT SCSI, or any vendor name. The only I/O diagnostics are the generic WRITE / READ / ERROR / STATUS= group at ram:9fb7–ram:9fc2, \r\nREAD ERROR at ram:b6c7, \r\nFEED ERROR at ram:b6ce, and CHECKSUM ERROR at ram:b4ac.

7.5 No validation of the disc against MSTYP — the heart of the question¶

[VERIFIED] The MSTYP dialogue (ram:94ad–ram:9587, disassembled in full above) contains not one I/O instruction. It is: print, read a number, range-check the number against a constant, index two tables, copy words into a parameter block, install a symbol, print the name. Nothing is read from any drive and nothing is compared against anything the hardware says.

MACM does not verify the operator's answer against the actual disc. It cannot. If the operator types the wrong MSTYP, MACM will happily generate a SINTRAN for the wrong disc.

7.6 No sense data, no SCSI status handling¶

[VERIFIED] The error reporting path prints a raw hardware status word — READ/WRITE + ERROR + STATUS= at ram:9fb7–ram:9fc2. That is an SMD controller status register, printed verbatim. There is no REQUEST SENSE, no sense key decode, no additional-sense-code table.

7.7 Conclusion, and where SCSI awareness actually lives¶

[VERIFIED] MACM handles device 0o144300 purely as a number: it is a constant in the MSTYP record for MSTYP 0o23, copied into ram:8343 and from there into a symbol value that the SINTRAN source assembles into the kernel. MACM never issues a single I/O to it.

[INFERRED] SCSI support therefore arrives only with the generated SINTRAN kernel itself — which, per the cross-referenced carving work, does issue INQUIRY, READ CAPACITY(10), reads the vendor control record from the last block, validates block size, and has 2210 / 144300B in MDISCS. MACM's job is to bake the right device number and the right library mark into that kernel; the protocol-aware driver is the kernel's, not MACM's. Reasoning: MACM contains neither IOXT nor any CDB construction, so no SCSI transaction can originate in it, yet the constant 0o144300 and the mark SCASI are both present as assembly-time data destined for the kernel.

8. Open questions¶

10,0$ semantics. Not resolvable from MACM — the binary contains the string and no parser. Needs the ND-100 MOPC microcode or the console handler. Same for the ! executor in 22!.
The eleven 3-word slots at ram:872d–ram:874d. Their value words are verified as the )9BYTT parameter destinations, but the name words do not decode as 6-bit symbols. Layout unknown. Raw words in §1.
Record words 2–9 — the disc geometry parameters. Only three distinct parameter sets exist (§6.3); which word is sectors/track vs words/sector vs cylinders is undetermined. Comparing against a known SINTRAN )9BYTT command line would close this quickly.
ram:833b — the second value from the DISK TYPE table. Purpose unknown beyond the 0o20 → 7 FIXED remap.
The full command-table extent and entry list. Format proven, five entries decoded; the table was not dumped end to end. nd_sixbit_decode.py makes this mechanical whenever it is wanted.
The 0o144300 bit pattern. No decomposition evidence in MACM.
The floppy build. Everything here is the standalone D:\ND\BPUN\MACM-1718L.BPUN. The L-floppy copy (0o76203 / 19,738 words) is 465 words larger and was not examined; the MSTYP tables may have moved or grown. Extract read-only with ndtool -x -o <dir> D:\ND\S\VSXL1.IMG (no -p — it strips bit 7 and corrupts binaries) if that comparison is wanted.

Ghidra annotations added by this analysis¶

In MACM-1718L.BPUN:

Functions created — disktype_menu_prompt_and_parse (ram:94ad), mstyp_menu_prompt_and_parse (ram:94cc), mstyp_install_libmark_and_report (ram:9546).

Data labelled — sintran_disktype_to_mstyp_table (ram:9483), mstyp_name_ptr_table (ram:9597), mstyp_record_ptr_table (ram:9715), var_mstyp_record_addr (ram:98b7), var_mstyp_fixed_index (ram:98b8), var_mstyp_return_link (ram:98b9), mark_DRUM (ram:9807), mark_REMOV (ram:980b), mark_FIXED (ram:980d), mark_BD288 (ram:980f), mark_BDFIX (ram:9811), mark_W8INC (ram:9813), mark_SCASI (ram:9815).

Comments — plate comments on ram:9483 and ram:9715 documenting both table layouts; a pre-comment on ram:951c documenting the FIXED bit-15 path.

Nothing was renamed that is not positively identified above. Nothing under D:\ND\ was modified.