The Lost Treasures of Infocom II CD (1992) - findings
Established against a physical pressing read on an Apple Silicon Mac with a Pioneer BD-RW BDR-XS07U. The general method is in docs/imaging.md; this file holds what is specific to this disc.
Identity
A single 12.4 MB ISO 9660 data track: one session, 6,075 blocks = 12,441,600 bytes of user data, volume LOST TREASURES II.
Apple ISO extensions carry the MAC and DOS trees side by side on the one filesystem, exactly the shape of volume 1; macOS mounts it but its cd9660 driver cannot open the fork-bearing Mac files, so extraction goes through scripts/appleiso.py (114 files, 18 resource forks, 0 failures).
The PVD was created 1992-11-06 14:12:33 (volume 1’s was mastered 1992-05-26, so the two CDs were mastered about five months apart).
114 files; see lost-treasures-2-1992.ls.txt.
The pressing is Redump-verified. The raw dump made 2026-09-11 hashes byte-identical to redump.org disc 105382, edition “CD-ROM Version (3 Bonus Titles)”, write offset 0; see checksums.txt.
Structure
MAC/holds all fourteen games: eleven as double-clickable applications with the z-code in the data fork, plusARTHUR FOLDER,JOURNEY FOLDERandSHOGUN FOLDERwithSTORY.DATA/PIC.DATA/CPIC.DATAbeside the application, andSOUND/with Sherlock’s seventeen sound files (thirteenSfiles plus four 32-byteMfiles, February 1988).DOS/holds one folder per game - story file, per-game interpreter,SETUP.EXE(1,584 bytes) andYES.COM- plusINFOCOM2.BAT, an F-key launcher menu, andREPLY.COMat root.- No shared
DATA/directory and no padding. Volume 1’s padded-to-round-sizes, caret-filled story files are absent here; every DOS story file is natural size plus a$1aterminator (AMFV.DAT262,018 for 262,016 bytes of z-code,BORDERZO.DAT178,373 for 178,372). The two 1992 CDs were mastered by different processes.
The full per-title release table for both platforms is in versions.md.
Findings
- This disc is Masterpieces’ other source. All thirteen titles shared with Masterpieces (1996) are byte-identical to Masterpieces’ builds on both platforms - and the DOS files match at full-file level,
$1aterminators and all, which identifies Masterpieces’ “natural size +1” provenance of PC story file as this disc’s file set. The Mac/PC build splits Masterpieces shows for Arthur (r54/r74), Journey (r26/r83), Seastalker (r15/r16), Sherlock (r26/r21) and Wishbringer (r68/r69) all originate here, completing the pattern begun on volume 1 (Planetfall, Hitchhiker’s, Zork Zero). - Shogun survives here and nowhere else in this catalog. The three bonus titles are Arthur, Journey and Shogun - and Shogun is the title Masterpieces later dropped along with Hitchhiker’s. The Mac build r292.890314 appears on no other cataloged medium (the leaked source repositories carry only r322); the PC build r322.890706 is byte-identical, full file, to
COMPILED/a5.zipin the leaked shogun repository. Both builds carry built-in hints. - Shogun ships three graphics libraries and that is the complete set:
SHOGUN.CG1,.EG1and.MG1, all dated June-July 1989, with no.EG2- but unlike Zork Zero’s genuinely missing files, nothing is absent: Shogun’s.EG1covers all 50 picture numbers by itself (see the.EG2note below), so there was never an.EG2to omit. - Arthur’s EGA files were made for this disc; Journey’s are the 1989 originals restamped. Doherty documents Arthur’s original MS-DOS release as shipping no EGA file at all - just
MG1,CG1, andFILECVT.EXE, a converter that generates the EGA and CGA renditions from the MCGA master on the customer’s machine. This disc’sARTHUR.EG1/.EG2, dated 1992-10-22 against 1989 dates on Arthur’s other files, are therefore pregenerated for the CD (consistent with running Infocom’s own converter during mastering, two weeks before the PVD date). Journey’s.EG1/.EG2carry 1992-10-27 stamps but their sizes (360,500 / 249,241) match Doherty’s original 1989-07-07 files exactly - originals recopied, not reconverted. The same files resurface inside Masterpieces’DATA.Z- includingJOURNEY.MG1, whose 1996-06-13 stamp on Masterpieces covers a file byte-identical to this disc’s 1992 copy. The timestamp stratigraphy recorded in the Masterpieces notes traces back to this mastering. - The
.CG1files use a different pixel format, andpix2gifcannot decode it. Arthur’s.CG1looks broken in tooling -pix2gifemits blank white images for its art - but the file is fine: 170 directory entries, 136 with real compressed data (median 1.7 KB per picture), 34 with no data at all (the invisible layout-metrics pictures), byte-identical to Masterpieces’ copy. The tell is a flag: every data-bearing image in every.CG1here (Arthur 136/136, Journey 134/134, Shogun 44/44) carries directory-flag bit 3 (0x0008), which no.EG1image has - it marks the CGA format, andpix2gif, documented for “MG1/EG1 files”, ignores the bit and mis-decodes the stream. The dimensions give away what that format is:.CG1art is the same 584x196-class geometry as the EGA art, impossible in CGA’s 4-color mode but native to its 640x200 two-color mode, and each picture’s compressed data runs one-half to one-third of its EGA counterpart - one bit per pixel instead of four.scripts/picdir.pylists all four libraries correctly; only pixel decoding is affected. .EG2is the EGA overflow file, and the overflow threshold is a 360K floppy. Arthur’s EGA art is split acrossEG1(125 images, 223,522 bytes) andEG2(101 images, 154,137 bytes), Journey’s across 360,500 and 249,241 bytes - and in each case the two halves together exceed a 360K disk’s capacity while each half fits, Journey’sEG1with under 2K to spare. The EGA rendition targeted the era’s lowest common medium; the CGA files fit a floppy whole, and MCGA (Journey’s is 465,606 bytes, over 360K by itself) implied a PS/2-class machine with 3.5” drives, so neither needed splitting. Period testimony corroborates this: the IF Archive’s MCGA index quotes Graeme Cree that Journey’s dual-media box put the VGA version only on the 3.5” diskettes, which is why used copies sold without them lack the.MG1. The split is play-aware, not a blind cut: 55 of Arthur’s pictures and 12 of Journey’s appear in both halves, duplicated so common elements need no disk swap. Shogun’s fifty EGA pictures fit one file (verified: itsEG1covers every picture number itsCG1has), so its lack of anEG2is not an omission - and per Doherty’s graphics-file table, Zork Zero’s EGA likewise fit one file and never had anEG2at all.- Interpreters match Doherty’s size table (11394=3L, 11402=3M2, 12004=3N, 12640=4A, 12682=4E, 12688=4D, 33946=5J, 47442=6.68, 47528=6.71) including Shogun’s
SHOGUN.EXEat 47,402 bytes - a size the other discs never showed, but present in Doherty’s full interpreter table as a build of 6.68. - Mac
SOUND/is byte-identical, all seventeen files, to Masterpieces’MAC/SOUND/.
Open items for this disc
None.
Resolved 2026-09-11: imaged, Redump-verified, extracted with forks, versions enumerated - same day the disc arrived.