Monday 28 November 2016

Missed Classic: Dungeon - And Another Thing… (Plus Final Rating)

Written by Joe Pranevich

Last week, I completed the adventure game marathon that is the original Zork. Of all the games that I have played here, this is the one that I feel proudest for winning. I have wanted to play and beat this game for years; thank you for giving me the opportunity to share it with you. I will get to the final rating shortly, but if you have been reading the comments then you know I have one remaining piece of unfinished business: the alternate versions.

Throughout my coverage, I have been aware that Voltgloss and others have been playing a different version of the game than I am. I was attracted by the “Dungeon” version because it was the one that I discovered for Linux many years ago and, as far as I knew, it was the only one. In this, I have betrayed my age because an alternate version of Zork has been available since 2004. I have a better idea now how these variants fit together and just won the other major contender for the “definitive” version. None of this affects the score so if you want to get right to that part, just scroll down.

Map of Zork from January 1982. Notice this is the 585-point version

For the last several weeks, I have been blogging the 585-point version known as Dungeon. I discussed the history briefly in the introduction, but it started as a fork of the original Zork codebase, freed by a hacker and ported to work on a wider array of systems than originally intended. From the early 1980s through 2004, this was the definitive version of mainframe Zork and the only version that could be played outside a small community of MIT-flavored PDP-10 users. This game claims to be derived from the “final” version of Zork and includes all puzzles created through 1979, but I have since learned that several refinements were made later and missed being included.

Mainframe Zork continued to have some development until 1981 when a new “final” edition was produced. This edition grew to 616 points and featured several additional treasures and puzzles. I do not know how widely distributed this version was; I have found documentation from 1982 that suggests the 585-point version was in wider circulation. The source code for this version was leaked sometime before the early 2000s (possibly in 1996), but it was unplayable, a historical curiosity only. It was written in a dead programming language called MDL (also known as “Muddle”) and could not be built on modern systems. Two later advances would finally bring this version back to a playable state.

In 2004, Ethan Dicks and others took that MDL source code and ported it to Inform, an adventure game creation language. In spirit, this was much like the original anonymous hacker that ported the 585-point version to other systems. This would become a 646-point version, but as far as I can discern it only changes the point values from the 616-point version rather than adding anything new. This version used the Inform parser and so while all the game elements were the same, a few solutions were worded slightly differently. It’s the easiest way to play something very close to the original game.

The final chapter in this story was finished in 2009. In that year, Matthew T. Russotto released “Confusion”, a Muddle interpreter written in C and C++ that knew “just enough” syntax that it could play the original MDL sources (616 points) with only a few modifications. It has its own bugs, but this version is the only one that utilizes the original game parser. You can get Confusion prebuilt for Windows, but I had to jump through a few hoops to get it to compile for the Mac. As I closed out my look at Dungeon, I played this version to conclusion so that I could get as much of the original experience as possible.

All of the above does not answer the most basic question: which version of Zork is definitive? Is it the 1979 version that predates the founding of Infocom? Or does the 1981 version best represent the authors’ intent? Does it matter that the later version may have been played by fewer people? I can’t answer that for myself, let alone for you. Instead, let me summarize the handful of differences that I discovered and you can decide for yourself.

Where is Saruman when you need him?

Same Game, New Puzzles

I did not solve the 616-point version of the game without help. Playing both versions side-by-side is maddening because new exits and changed descriptions do not stand out in descriptions you’ve read dozens of times before. I may have a similar challenge playing Zork I later. I tried to play the new puzzles straight but did not always succeed without help. These are the differences that I found:

The Three Palantirs

The largest addition is the three-part “palantir” puzzle, colored crystal balls that revealed distant parts of the maze. Each ball, when found, hinted at the location of the next.
  • The White Sphere replaced the “crystal sphere” that we found in the Wonderland area. Looking into it revealed the location of the Red Sphere.
  • The Red Sphere puzzle was fun. Just west of the Torch Room was a new room with a locked door. The White Sphere revealed the other side and showed that someone had left the key in the keyhole. With that knowledge, I was able to slide the welcome mat (a new object found on the surface in front of the white house) under the door and push out the door key using one of the skeleton keys. I then pulled the mat to retrieve it and opened the door to collect the Sphere. Looking into it revealed the location of the Blue Sphere.
  • The Blue Sphere puzzle was the most difficult of the set. The Red Sphere described a “Sooty Room” filled with coal dust so I checked the mine first. I re-mapped the whole place without finding any differences. I had to look up the solution and I suspect I never would have worked it out: if you tie the rope to the timber then drop it at the top of the Slide Room, the rope will dangle down the slide. You can climb down to access a hidden room in the middle of the slide which contained the final sphere.

Close cover before striking.

Other Puzzles and Changes
  • The trapdoor to the basement now stays open after you discover any alternate exit from the maze. This simplifies the process of getting treasure back to the case considerably.
  • The entrance into the Land of the Dead now requires the longer exorcism puzzle that I remembered. Once I had the bell, book, and lit candles at the entrance, I did the sequence: I first rang the bell, but it turned red hot and forced us to drop both it and the candles. I then picked up and re-lit the candles then read the book. The spirits departed and we accessed the crypt as before.
  • I missed one new puzzle completely. The matchbook has new text that tells you that you can send away for a brochure. I thought nothing of it, but if you type “send for free brochure” (even when not near a mailbox), the brochure containing a “Don Woods commemorative stamp” will arrive at the white house in a few turns. This puzzle both honors one of the creators of Adventure while also poking fun at an obscure puzzle in that game. I like the idea behind this puzzle, but it fails to make sense on multiple levels.
  • Once you return all the treasures to the case the Wraith tracks you down shortly no matter where you are. It was much easier to miss him in the older version as he seems to only appear in the Crypt. 
This version feels more polished overall. The parser accepted complex commands like “drop everything except torch” and had the proper use of pronouns. A handful of text changes made puzzles like the clockwork canary and the sliding blocks much easier to understand. In contrast, additions like the “Sooty Room” and “free brochure” puzzles set the game backwards in solvability. In the end, it’s a wash. My feelings on the game are unchanged. I appreciate why some fans love and swear by each of the two versions.

Introduction to the MDL version under Confusion.

Final Rating

Puzzles and Solvability - Most of Dungeon boils down to “just” a treasure hunt, but experienced through a set of some of the most diverse puzzles in adventure gaming. We had not only traditional mazes and fetch quests, but also a sliding puzzle, a “Simon Says” puzzle, and many others. The endgame itself is like a masterclass in puzzle design and unlike anything I have played before. Although some of the puzzles were hard, only a few were “unfair’-- the glass bauble being the worst offender-- and I always felt like there were new things for me to try nearly until the end. Well done. My Score: 7.

Interface and Inventory - I had to play the MDL version to appreciate the original engine but what I found was fantastic. The text input was intelligent and dynamic; it would be excellent for 2010s let alone the early 1980s. The fact that it was implemented in a LISP-esque language certainly helped. Suspect received a 6 for nearly the same interface. That is a bit too high for me on a text-only parser, but this is an excellent text-only parser. My Score: 5.

Story and Setting - There is little story to speak of in Dungeon, just an unnamed protagonist delving into the ruins of an ancient civilization. Players are rewarded with hints here and there of a deeper backstory with the Twelve Flatheads and the decline of a brutally inefficient empire. It’s nonsensical and fun but not particularly deep. Although the setting as a whole makes nearly no sense, individual areas are exceptionally well done, especially the volcano and Flood Control Dam #3. My Score: 4.

State of the art ASCII imaging.

Sound and Graphics - You might expect a zero here because we are playing a text adventure, but the ASCII art for several of the objects as well as the basic representation of the sliding puzzle maze give it enough of a fraction of a point that I need to round up. My Score: 1.

Environment and Atmosphere - I had a lot of fun playing the game. The designers kept it whimsical and rarely did the game go into darker territory than it should have. My biggest complaint is that the layout of the dungeon makes absolutely no narrative sense. Why have a bank below the white house? Why have a cyclops in the center of the maze? Why have a well that doesn’t lead anywhere but more puzzles? The designers tried to take an Adventure-like maze and pretended that it made sense for something closer to archaeology. My Score: 3.

Dialog and Acting - The prose in this game is well-written and I was never bored or misunderstood what the game was trying to say. My Score: 4. 

Final Calculation

Totaling it all up we get (7+5+4+1+3+4)/.6 = 40 points. I’ll award a single bonus point for the tremendous influence this game had on the rest of interactive fiction. Final score is 41!

While not astronomical, that places Dungeon above the other text adventures that we have played and it scores well against early graphical adventures. That feels right. A lot of what made Infocom great can be found in this game. It rewarded the time that I spent with it, but I can see someone with less dedication having a worse experience.

You guys did an amazing job guessing the score on this one. Average guess was 40.3. Niklas and Ilmari will split the reward for guessing 40 and 42 respectively. Congratulations! CAP distribution will be included in the Star Trek final rating which will land either this week or next.

What’s Next

This was the first step in a very long marathon, but it just got longer: based on some prompting from Voltgloss, I will add Starcross to the playlist. He was a great help with hints and if he says that has some elements that tie into Zork, I’ll play it and see. I also learned of Mini-Zork, a stripped-down 1987 variant of Zork I for systems that could not play the full thing. As of right now that is not on the list, but I might be convinced if someone feels strongly about it.

The next marathon post will be Zork I, but I’ve already won and will cover it in one go. Essentially, when Infocom was founded they created a version of Dungeon with fewer puzzles (such that could fit on a contemporary home computer) and launched it as Zork I. Please comment here with your score guesses since there won’t be an opportunity later.


  1. Nice.

    Zork 1 doesn't really change much, except for removing some solutions (fortunately, many of the more obtuse ones were cut), maybe one or two solutions, and that only improves slightly. I doubt it will change much, so I guess 42 for the score.

    And the changes in that version are indeed what happens in later Zork games; Zork 1 also stops closing the trapdoor once you find an entrance (at least in the latest version I played).

  2. One last "alternate solutions" comment: several different small pointy objects can be used to poke the key out of the keyhole in order to get the Red Sphere. Both the pointy stick and the screwdriver work, for example. I actually didn't know the skeleton keys also work, but it makes perfect sense.

    I'll guess 40 for Zork I. A couple of items in Dungeon that were worth points in your evaluation (in particular, the endgame and the Ascii art) are not present in Zork I, so I expect the score will dip slightly as a result. Some of the puzzles are, I think, better clued and therefore fairer (like how to open the egg and how to cross the rainbow), which might bump the score back up by 1 point; but I don't think by 2.

    So far, my shortest "number of turns needed to finish the game" score for Dungeon is 625 turns; and my shortest such score for Zork I is 286 turns. I'm sure these can be improved upon, both with optimizing and with better luck on the random elements (especially dealing with the Thief).

  3. First time the dice rolled in my favour :)

    1. Don't forget to roll them for Zork I!

    2. Thanks for reminding me, and lets see... 94! Guess there won't be a Zork blessing for the rest of the marathon on my dice.

  4. This will be like comparing Fellowship of the Ring with the Lord of the Rings. I'll wager that a first part of a trilogy cannot hold up to a complete game (especially as the first Zork has almost nothing new to offer). Let's say 39.

    1. My goal will be to judge it as if Dungeon/Mainframe Zork didn't exist because that is how nearly all players came to it. So I will not be deducting points because I saw the same puzzles before.

  5. Do we happen to have any readers near Montrose, Scotland? I'm doing some research for an upcoming game and I may need a favor...

    1. 8 months late as I am in commenting here (I've been trying to catch up from the very beginning for the last 6 months), I'm now looking forward immensely to finding out why you asked this - and I'm sorry no readers seemed to be close enough last November.

    2. Wow! Glad you are catching up! Plenty of games to go, but feel free to comment because we still see many of the comments on old posts.

      I was looking for someone in that area to help with some research on Tartan Software as part of my post on "Crisis at Christmas".

  6. If you don't want to use Confusion, but want to experience Zork as close as possible to the original, there's an exact (or as close to exact as I could make it) port of the MDL version to C++17 at Downloads are available for Win32, Win64, and Linux at