The Answer to the DRM questions for Where in the world is Carmen Sandiego? Enhanced (DOS, 1990) are, in no particular order:

23
Kent
dragon
calcium
1796
Warren
revenue
1792
Willard
1937
Crater
Tanzania
Hartford
Duluth
London
Gem
Silent
squeaker

if ((0x80 >> ((byte)local_4 & 7) &
(int)(char)*(byte *)((int)((int *)param_1 + 1) + (local_4 >> 3))) != 0) {

COULD YOU USE SOME MORE CASTS MAYBE?

oh it's because ghidra's near/far pointer support is shit.

I had param2 defined as a byte*32 and it was casting it to a byte* before using it

if I define it as byte* and let the calling convention implicitly define it as 32bit, it doesn't do the cast

well I found the decompression method.

as always, I hate it. decompression routines are probably my least favorite thing to reverse engineer

I think this compression is specifically designed for ASCII text, which is annoying because they've also got compressed images... which probably use a DIFFERENT COMPRESSION!

it looks like this chunk has length 256, which means 253 usable bytes, and it expands to 374 bytes.

Not the greatest compression. a little better than just doing 6-bit ASCII.

it's some kind of shifting bit mask but it starts at encoding values in 4 bits, then it can increase (or decrease, I guess) based on the input stream.

then it has an output filter, where if the number specified wasn't 8 bits, it's actually an index into a predefined text table

the predefined table starts with NUL, space, then:
aetonisrdlhugfcwypbmk,vSA.T'PMxBCIRGDWHqE-zNFKL0j:51YJ8\U?73Q;2!469
\r\nOVXZ()*+"#$%&<=>/@[]^_`

given that the most comment symbols are near the beginning, this is presumably a sort of lazy huffman coding

but I've got the predefined table, an input file, an output file, and now I need to write some python code to replicate this, hopefully without crying

"vs ses oa is isgit's tc eital and largest t u anhtA ttggh os nnotosnhrdsmarosogdn ss drte tishoth's isdhsceohtsnthminder of isgit's t nuorhdhtpast\x00 geru is slightltsn oaller than ndhd na and is o nnsgtgstbtst oa dotlalssaaolootbiaoht Sal gh, sonuhvia and sl ghh\x00isgit, ontvdn ss nhsiaalgarsnadlfnaatawlarst oadrlhrs i is a rugged land dooousr'casrbhe nrdsgs fountainsnht iah"

I mean, it's not 100% wrong, but it's not right either

that's supposed to read:
"\x03Lima is Peru's capital and largest city. A well-known landmark is the Archbishop's Palace, a reminder of Peru's colonial past\x00Peru is slightly smaller than Alaska and is bordered by Ecuador, Colombia, Brazil, Bolivia and Chile\x00Peru, once the center of the mighty Incan Empire, is a rugged land dominated by the Andes Mountains. Forests and jungles cover half its land area\x00"

I somehow confused the dosbox-x debugger into not accepting letters anymore

it was a trivial off-by-one error.
I was doing saved_byte=input[3]

but while I needed the 3rd byte, that's at input[2]

yess!

C:\DOSBox-X\drive_c\carmen\py>python datfile.py cities.dat --dump=12803 --decompress
"\x03Sydney, with a population of more than 3.3 million people, is Australia's largest city. A well-known sight is Sydney's distinctively designed Opera House\x00An island continent, Australia is nearly as large as the United States but has only one-fifteenth the population\x00The capital of Australia is Canberra, located in the southeast corner of the country between Sydney and Melbourne\x00"

It starts with \x03 to indicate there's three strings: then it describes the city three times. at runtime it uses select_string function with a random input to select one of the three strings

okay now that I can decode the chunks (well, most of them) I can identify a lot more of them:

00 Name and (some other info)
01 ???
02 Image
03 City descriptions
04 Items to steal
10 ???
11&up: Hints leading here

So like, the 12 chunk for Tokyo says:

b'\x05asked about the exchange rate for yen\x00was practicing Japanese characters\x00said\x81planned to take photographs of Mount Fuji\x00asked about tours of the Imperial Palace\x00was interested in visiting Shinto shrines\x00'

So it picks from one of those 5 options