typed in an improper kind of a sentence, such as asking a question or using a
non-imperative voice. It would try to notice if they did that two or three times in a row
and then just say, “The way to talk to the game is,” and then give a few examples.
And I think that the really critical thing about the parser interface has nothing to do
with typing, it is being able to use natural language for your inputs.
Did you ever feel limited by the Infocom development system?
The system was extremely powerful and flexible, and could grow to meet the need of a
particular game fairly easily. A minor exception was any change that required a change
to the “interpreter.” Every game sold consisted of the game component, which was
machine independent, and an interpreter, which was a machine-specific program which
allowed the game component to run on that particular microcomputer. Since there were
twenty or more interpreters (one for the Apple II, one for the Mac, one for the DEC
Rainbow, one for the NEC PC-800, et cetera) a change to the interpreter required not
changing just one program, but changing twenty-plus programs. So that could only be
done rarely or when it was extremely important, such as changing the status line in
Deadlineto display time instead of score and moves.
A more stringent limit was imposed by the desire to run on the widest possible
array of machines, so we were always limited by the capabilities of the smallest and
weakest of those machines. In the earliest days, the limiting machine was the TRS-80
Model 1, whose disk drive capacity limited the first games to an executable size of 78K.
As older machines “dropped off” the to-be-supported list, this limit slowly rose, but
even when I wroteHHGTTG, games were still limited to around 110K. Generally, this
limit would be reached midway through testing, and then every addition to the game, to
fix a bug or to handle a reasonable input by a tester, would require ever more painful
searches for some text — any text — to cut or condense. At times, this was a good dis-
cipline, to write lean, to-the-point text. But often it became horrible and made us feel
like we were butchering our own children. OK, that’s a slight exaggeration.
How did the development process work at Infocom? Were you fairly free to
choose what games you made?
In the early days, things were pretty informal, and decisions were made by fairly infor-
mal consensus. In the later days, particular after the acquisition by Activision, decisions
were much more mandated by upper management. Generally, the choice of a game was
left up to the individual author. Authors with more of a track record, like Dave Lebling
and myself, had more leeway than a greenhorn implementor. Of course, there were
marketing considerations as well, such as the strong desire to complete trilogies or the
opportunities to work with a licensed property such asHHGTTG.
One thing that was standard over the whole seven-plus years that I was at Infocom
was the “Implementors’ Lunches,” or, for short, “Imp Lunches.” These were weekly
lunches at which the game writers would get together to talk about the games in devel-
opment, share ideas, critique each other’s work, et cetera. It was probably the most fun
couple of hours of the week.
There wasn’t too much oversight during the first few months of a game’s life, while
the implementor was working pretty much alone, other than at the Imp Lunches, any
184 Chapter 10: Interview: Steve Meretzky