Sunday, June 23, 2013

How Mazewar Escaped from a Lab at MIT in 1977

[6/28/2013 See comments at end from MIT Alumni that fill in some details here]

This is the story of how an early multiperson computer game, way ahead of its time, escaped from a lab at MIT and ended up on networked computers on the West coast and from there out to the rest of the world.

This is also an example of how difficult it is to understand events that happened in another time, another period of history, when the technology was different than we are used to. When this story took place, people used minicomputers or ran batch on mainframes, there was very little graphics, local area networks were research projects, and on and on.

I was taking a break from college and worked at the RAND Corporation and had been on the ARPANET since 1973 which is about as early as you can be on the ARPANET. I had made a lot of friends at MIT at the AI Lab and what we then called the Architecture Machine Group.

RAND sent me on a trip to Cambridge and I stayed a few extra days and slept on Lee Parks' couch at the Architecture Machine. It was on one of these tours that I saw the Spatial Data Management System at the Architecture Machine, or "Put that There". Seth Steinberg was working with Bob Frankston and working on their product after Visicalc, something called TK Solver! which was spectacular.

For some reason, Charles Frankston, who was at the AI Lab then, took the time late at night to show me a multiperson game called Mazewar. Mazewar ran on a PDP 10 computer that had a bunch of graphics computers attached, something called an Imlac. An Imlac was a 16 bit computer all its own that could do dozens of vectors a second, barely. I remember a room with a dozen or so Imlac's against a wall, so I am guessing that this was a graphics lab at MIT of some sort.

The basic game was this: you were in a maze. You can see whatever direction you were facing, down the maze, at a wall, whatever. If you saw another player, you could see them as well, represented as an eye and the eye had a direction so you could see which way the player was facing. If they were facing away from you or at right angles it was quite possible they had not seen you. Using the keyboard (there were no mice), you could navigate (forward, backwards, to the sides), or turn right or left, or stop, or fire straight ahead. If the bad guy was ahead of you, and you fired first, you won. The other player would be reincarnated somewhere else in the maze. In modern terms, it was an early 1st person shooter.

This was probably 1977.

I return to LA and go back to school to get my degree, and my friends leave RAND and move to Xerox where they are working on a secret project. I get a demo of some of their technology, called the Alto, and I am blown away. This is the future. It is the Alto that Steve Jobs was shown when he came up with the idea for the Macintosh, so they say.

The Alto was perfect for Mazewar. It had the screen, the user interface (keyboard/mouse), the network to communicate. It did not have a central computer like the PDP 10 so we came up with a distributed architecture for the game played over the network. Jim Guyton at Xerox did most of the programming. I described the game (Jim had never seen it) and figured out how to make the graphics efficient. Jim releases the game inside Xerox.

My friend Marc Cantor, founder of Macromind aka Macromedia, sees it and does a Mac implementation. Jim is asked to write an article about the Alto implementation for Byte magazine and he does in 1980. This turns out to be one of the first, if not the first, network distributed multi-person game with various points of view, in the public literature.

I am sure we were not the first. But apparently we were close to the first to talk about the ideas in print. Jim now becomes an expert witness to break weird patents on networked games. So do I.

Anyway, Mazewar has a loyal following, it even had a 30 year reunion that I did not know about.

The point is, that back then, people helped you, you shared ideas, it wasn't about making a fast buck, it was about showing these ideas would work when no one but us believed it would.

Now of course, things are different.

We got the following comments from MIT Alumni:

From Tom Knight on 6/24
That would have been in the Dynamic Modeling group Imlac installation, on the second floor of 545 Technology Square.The Imlacs were connected by serial lines to the Dynamic Modeling PDP-10, running ITS, one of three KA-10 ITS systems on the 9th floor of Tech Square at the time. JCR Licklider, who ran DM, didn't like those new fangled bitmaps. In my opinion (and that of many others) the Imlacs were a programming and support nightmare. The epitome (with the possible exception of the similar GT-11) of the catch phrase "There is a special name for a little bit of intelligence. It is stupidity." Cleverness in the console program led to unending complexity and failure in the mainframe.
From Ed Schwalenberg on 6/24
Here are a few things I remember:
The Imlacs were owned by the Dynamic Modeling group of MIT LCS, headed by Al Vezza.  Vezza was not fond of Maze, because randoms like you would come in at midnight, pound on the keyboards and break them.  So the installed version of Maze was typically neutered; you had to have a guide like CBF to know where to find a good copy.  Also, there was a screensaver for idle Imlacs; one of the images was a Maze playing position where user AV (Vezza) was directly in front of you, his eyes directly on you.
The cognoscenti also knew how to activate various cheat modes. A regular shot had a propagation delay to the target; control-mumble-cokebottle eliminated that delay.  Another patch activated keystrokes that would let you remove walls in your copy of the maze.  A third would show you the positions of others in the overview.
SAIL had some Imlacs, notably one at John McCarthy's home; I wonder if that was the first "home computer"?  I also wonder whether he ever played Maze on it.
Dynamic Modeling or Dynamod was located in what was then NE43 aka 545 Technology Square. I well remember (from midnight tours led by KLH) the room with a bunch of Imlacs, but I don't remember the room number.  That building has been engulfed (the east wall is now the west side of an enormous atrium) and renumbered 200 Tech Square; it's now Novartis.
Dynamod the research group, and DM the machine, played another role in gaming history, employing the hackers who wrote the original Zork.
Kris Karas adds
Imlacs not withstanding, DM was also home to MDL, a wonderfully cuspy
language, if anybody remembers it.  (I still have my MDL software
reference, forlornly gathering dust on a bookshelf.)  I probably owe
some personal success in the field of software to MDL, MACLISP, and DM.
I taught myself elements of good software structure and design from that.
Bill York adds
I remember that as an early MIT student, getting in to the 2nd floor of 545 TS to play Maze was one of the rarest of privileges, and as others have said I owed my access and my Maze training to Charles.
As Ed mentioned, the key gameplay difference between the standard Maze game and most FPS games was that you weren't so much firing a gun as dropping a time-delay grenade in the hallway. This made for very challenging game play, allowing you time to avoid getting killed if you could manage to duck into a side passage in time, or to doom an opponent by baiting him into chasing you into a corridor where you had left a nasty present waiting.
In addition to Zork (which I lost much more time to than Maze) the DM group (or at least individuals) also produced one of the first wide-area multi-player games, an ongoing trivia contest based on user-submitted content. I believe that there were players from all over the ARPAnet-connected world. I think that Peter David Lebling (part of the Zork creation team with Tim Anderson and Mark Blank) wrote and maintained it. He also perpetually occupied the top ranking slot with a commanding lead over the rest of us peons, though I held down 2nd place for a while. Anyone else remember this?


Mazewar page on Wikipedia

Xerox Alto on Wikipedia


  1. I played MazeWar a bunch on the Altos while I was at Carnegie Mellon between 1984 and 1988.

    1. Daryll played on the Altos because Xerox made a special deal to release this hardware to certain schools even though they were never released to the general public. I think that Stanford, Carnegie Mellon and MIT got some of these, and maybe others for all I know. Although we take things like Altos for granted today, at the time they were completely amazing (well, at least in 1977 they were).

  2. Lots of good memories playing that at PARC when I interned there, including going at it with Randy Pausch, who was also an intern :-(

    It's spelled Canter. He could've been a cantor if he wanted. Mac Mazewar came out in 87 or 88.

  3. Marc is one of my best friends, you would think I would know how to spell his name by now. But obviously not !