| Filename | Latest commit message | Latest commit date |
|---|---|---|
| .. | ||
| manifest.xml | ||
| MSDOS320-DISK1.json | ||
| MSDOS320-DISK2.json | ||
| MSDOS320-DISK3-PROGREF1-thumb.jpg | ||
| MSDOS320-DISK3-PROGREF1.jpg | ||
| MSDOS320-DISK3-PROGREF1.json | ||
| MSDOS320-DISK4-PROGREF2.json | ||
| README.md | ||
MS-DOS 3.20
The MS-DOS 3.20 disks in the PCjs collection are the only known non-OEM copies of that release available online. All other known MS-DOS 3.20 disk images are OEM releases (eg, HP, Zenith, Data General, Seiko Epson, etc.) and can be found on sites like WinWorld.
Disks 3 and 4 contained a printed date of March 21, 1986 (see the photo below), which is consistent with
the file timestamps (2-14-86 and 3-14-86) on those disks. However, all the files on Disks 1 and 2 are
timestamped 7-07-86. Apparently, the Programmer's Reference files were finalized several months in advance
of the final MS-DOS 3.20 binaries.
Based on the dates of the binaries, July 7, 1986 is considered the release date of MS-DOS 3.20.
NOTE: Disks 1 and 2 are restored versions of the original disks, since the original disks had been slightly modified for personal use. The disks should be complete, and all files should be genuine and in the original order. As proof, the following data from Disk 1:
00013db0 0d 0a 0d 0a 4d 69 63 72 6f 73 6f 66 74 28 52 29 |....Microsoft(R)|
00013dc0 20 4d 53 2d 44 4f 53 28 52 29 20 20 56 65 72 73 | MS-DOS(R) Vers|
00013dd0 69 6f 6e 20 33 2e 32 30 0d 0a 20 20 20 20 20 20 |ion 3.20.. |
00013de0 20 20 20 20 20 20 20 28 43 29 43 6f 70 79 72 69 | (C)Copyri|
00013df0 67 68 74 20 4d 69 63 72 6f 73 6f 66 74 20 43 6f |ght Microsoft Co|
00013e00 72 70 20 31 39 38 31 2d 31 39 38 36 0d 0a 0d 0a |rp 1981-1986....|
matches the data listed at 16bitos.com, as do the MD5 checksums of IO.SYS, MSDODS.SYS and COMMAND.COM.
Surprisingly Bad Bug in MS-DOS 3.20
If you attempt to boot MS-DOS 3.20 on an IBM PC Model 5150 (with either an MDA or CGA video card), the machine will appear to crash after printing the first character.
When MS-DOS 3.20 prints characters to the screen, it uses this code:
0287:1FA8 CD29 INT 29 ;history=7
0070:017A 50 PUSH AX ;history=6
0070:017B 56 PUSH SI ;history=5
0070:017C 57 PUSH DI ;history=4
0070:017D 55 PUSH BP ;history=3
0070:017E B40E MOV AH,0E ;history=2
0070:0180 B307 MOV BL,07 ;history=1
0070:0182 CD10 INT 10
AX=0E43 BX=0B07 CX=0010 DX=0000 SP=0862 BP=36FE SI=00D3 DI=000B
SS=0287 DS=0287 ES=0287 PS=F012 V0 D0 I0 T0 S0 Z0 A1 P0 C0
whereas MS-DOS 3.30 uses this code:
022F:222E CD29 INT 29 ;history=10
0070:069E 50 PUSH AX ;history=9
0070:069F 56 PUSH SI ;history=8
0070:06A0 57 PUSH DI ;history=7
0070:06A1 55 PUSH BP ;history=6
0070:06A2 53 PUSH BX ;history=5
0070:06A3 B40E MOV AH,0E ;history=4
0070:06A5 B700 MOV BH,00 ;history=3
0070:06A7 B307 MOV BL,07 ;history=2
0070:06A9 CD10 INT 10 ;history=1
MS-DOS 3.20 doesn't clear the BH register before calling INT 0x10 Video BIOS function 0x0E, and the Model 5150 Video BIOS uses the BH register to select the video page, so when BH contains garbage (0x0B in the above example), it prints characters off-screen, where they can’t be seen.
As a result, the machine hasn't technically "crashed", but it's printing characters where they can't be seen, making the machine more or less unusable.
Starting with the IBM PC Model 5160, Video BIOS function 0x0E ignores BH and always use the current page.
Microsoft must not have tested MS-DOS 3.20 on any 5150 machines, or if they did, those machines must have contained
upgraded video cards, such as an EGA, which provided their own Video BIOS. This bug was fixed in MS-DOS 3.21,
but that release apparently wasn't made available until almost a year later (MS-DOS 3.21 files are timestamped
5-01-87).
MS-DOS 3.20 Photos
Here's what one of the MS-DOS 3.20 diskettes looked like.
And the box, courtesy of www.oldcomputermuseum.com.
