Does anyone know if there is any information available on how Level 9 Computing created their text adventures? I’m especially interested in the coding methods they used to fit large games into machines with only 32 Kb of memory and cassette tape loading.
As it happens, I got curious about Level 9 last week and went on a hunt for information myself. I didn’t find much in-depth technical information, but Jimmy Maher’s Digital Antiquaria has some lovely articles on the history of the company:
The main resource for Level 9 is the Level 9 Memorial. It’s a rather dated-looking website, but it does have contain a factsheet that you might find useful, and links to all of the games:
Thanks. I had seen the Level 9 Memorial before, but not the Digital Antiquaria posts.
It’s a shame that information on the development process is so hard to find.
Whereas some of the minds behind Magnetic Scrolls, Adventure International and Infocom are still contactable and in some cases still involved with IF, the Austin brothers have long since disappeared into the digital ether. You might try emailing the author of the Level 9 Memorial website.
We don’t really know a lot beyond what’s been learnt from reverse engineering their own interpreters for the Level9 C interpreter. They certainly had their own virtual machine format, but what development tools they used to generate that format seems to have been lost in the mists of time.
It’s also worth noting that their compression, while good, is not really much better than, say, Infocom’s. It’s just that Infocom’s games were distinctly larger.
I managed to find a couple of magazine interviews from links at Level 9 Computing at Wikipedia .
It seems that A-Code was created from scratch by the Austin brothers and is “kind of like machine code” with a reduced instruction set. It “makes programs about six times smaller than they would be in any other language including machine code and the text is compressed”.
Their compression methods could reduce text to “anything between 30% of the original and 50%” and at one point they could also “store a picture in about 30 bytes”.
It would be interesting to know how much memory the original Colossal Cave Adventure would use if rewritten in Z-code, as Level 9 managed to fit their version into 32 Kb machines with another 70 locations added.
Infocom games were certainly more verbose, but I’m not sure they were that much bigger on average compared with Level 9 games. Snowball had over 7000 locations and although the descriptions were shared for many of them (it was set on a huge spaceship), you could drop an object in any one of them and come back later to find it in the same place.
Well, yes, but 6,900 or so of those locations in Snowball were identical. And even if you can drop objects in them, you wouldn’t need that much memory to track objects - the obvious way would be to associate a number indicating the room with each object, so that would only cost you 2 bytes per drop-able object. It’s the text that is most likely to take up lots of room.
What I meant was, you can prove they are separate locations by dropping objects. Some games leave you at the same location when you type a direction instead of actually moving you.
Even ignoring Snowball, most Level 9 games had 200 locations or more and fit in 32Kb machines. I don’t think Zork I even had 100 locations and the Z-code files I’ve seen for that are 60 Kb or more, without the interpreter.