What is ZIL anyway?

http://blog.zarfhome.com/2019/04/what-is-zil-anyway.html

by MrRadar • 7 years ago

204 69 7 years ago

blog.zarfhome.com

Maybe there can be a new golden age of interactive fiction using modern NLP and other AI related technologies?

jordanpg2• 7 years ago

I enjoy thinking about the future of IF. I see it as a gaming/storytelling medium with a simultaneously huge potential but still vast chasms to cross before it gets there.

NLP etc. has great potential to improve the parsing capabilities, to be sure, but it's not obvious to me that that will dramatically improve the experience. Is `pick up the shiny brass thing` really better than `GET LAMP`? IMO, it might make IF more accessible to non-techy, non-text parser gamers, but that's about it.

I think the next leap for IF games will involve non-linear stories that are better suited to the medium. The success of games like "Papers, Please" in particular has sparked my imagination in this area.

willvarfar2• 7 years ago

I recall a Ludum Dare entrant which was an android app that pretended to be a walkie-talkie to control some security operatives you were telling to search a building to investigate some noises. Very spooky. And a very good use of some basic voice parsing and NLP.

PaulHoule2• 7 years ago

Inform 7 lets you write games in a language that reads like English even if it doesn't write like English.

alchemism• 7 years ago

The most memorable IF I’ve played through was one[1] in which you were a captured spy being interrogated in semi-linear fashion. The game was a retelling of your capture, where every wrong choice amounted to ‘It could not have happened that way, go back and tell me what you really did.’

The only way to win was to slip one lie into the retelling, and then when the game catches up to the present moment...

[1]Spider and Web

Zanni• 7 years ago

Struggling with the limitations of the parser was one of the recurring frustrations with Infocom games. "GET LAMP" is fine, but anything more complex quickly becomes challenging. I remember getting stuck in "The Lurking Horror" because I couldn't figure out how to climb down off a ledge, despite having a ladder. Use ladder? Climb ladder? Climb down? Climb down with ladder? Climb ledge with ladder?

msla• 7 years ago

The problem to solve is game length and replayability.

If you don't make moon logic puzzles, your puzzles can be solved relatively quickly unless you're very, very good at making fair puzzles which are also legitimately difficult. Therefore, you're going to need more puzzles than an FPS or action game which incidentally has some puzzles in order to achieve the same total game length, and making content like that is legitimately hard.

Similarly, in order to make your games playable more than once, you need branching paths, such that you have different routes for the players to go down each time, and I really shouldn't need to explain the concept of "combinatorial explosion" in this forum. Add to that the expense of things like art assents and voice acting (unless you really think text adventures are going to come back in a big way) and it becomes prohibitively expensive to not railroad everyone down one main plot with, at most, a few minor side-treks which quickly rejoin the main line.

PaulHoule2• 7 years ago

I want a procedurally generated version of Danganronpa.

zantana• 7 years ago

I do think there is a Sturgeon's Law problem where it's hard enough to come up with linear stories which are compelling. Based on some of the experimentalism in modern day game development in places like ichi.io hopefully they can come with some compelling genres

I'm not going to go down some half finished path in Bandersnatch when Black Mirror has enough problem with single stories.

jordanpg2• 7 years ago

This is why I think the "Papers, Please" paradigm is so interesting.

The overall storyline is replaced with a repetitive but novel task with increasing difficultly, and various other loosely coupled storylines are grafted onto the task in different ways. The task also replaces the explicit puzzles.

It is as though rather than being told the story, you are viewing vignettes of the story through a very narrow window

bayesian_horse4• 7 years ago

The reason why IF Games are so focused on puzzles is that they essential can't do anything else. Even dialogs are either about solving a puzzle or choosing a branch in the story.

The point about AI in IF would be to break that mold. Dialogs can be more challenging, for example you could make a game about interrogating people. Or you could tell NPCs to do certain stuff for you. Or these NPCs could have some advanced planning or decision making.

To make IF more interesting, especially for those outside a very narrow niche, you would need them to be a lot easier to navigate, and offer more challenges that don't look like puzzles and don't hit you with a brain-twister at every turn.

sehugg• 7 years ago

Versu was one such attempt, with an engine designed by the Sims 3 AI developer: https://versu.com/about/

Though I'd guess that's more of a sophisticated GOFAI than RNNs.

blacksmith_tb• 7 years ago

Seems like smart speakers are a good fit, my eight year old spent a little while playing with an IF Alexa skill not long ago. Of course, backtracking leads to a lot of agonizing repetition, which isn't really different than seeing it in text, but at least there you can skim over descriptions you've already seen.

egypturnash2• 7 years ago

This is why "verbose/brief/superbrief" became a thing in IF.

Verbose: Always show the full description of a room when you enter it.

Brief: Show a short description when you re-enter a room.

Superbrief: Always show the short description.

(Regardless of which you chose, you could always 'look' to get a full description.)

imglorp• 7 years ago

I love this idea, but encourage people not to get hung up on the UI--ie how you interact with the game world--and instead think about what's going on in the game world.

What if the world included a number of fully independent agents with their own motivations, values, and policies and equipped with pattern matching, goal solving, optimization, and even a hint of common sense[1]. The game state would never be the same twice, and it would depend on how you interacted with the agents and they with each other.

TlDr: If you open a door, you might let the Grue out of its maze.

[1] https://github.com/AndrewSmart/opencyc

bayesian_horse4• 7 years ago

I would suggest dialogs as the main focus for AI.

I don't believe "complex and dynamic" NPCs would work well enough to make a fun game.

egypturnash2• 7 years ago

I suspect that making it not feel like you're pushing buttons to get the same lines of dialogue out of wandering bots is going to require solving some pretty hard AI problems.

You could probably make it work now with, say, a game world full of broken robots or something.

gmueckl• 7 years ago

Well, I would like to think that the ties of interactive fiction to text as the mode of interaction are likely the first to be broken.

Both speech recognition and speech synthesis have progressed by leaps and bounds. Computer graphics have progressed to a point where human characters can be made expressive and nuanced.

Imagine an engine that that takes games scripted as interactive fiction, procedurally generates the game world from the prose (far fetched, but first steps exist as research papers), populates it with the characters from the text and makes them act out the written dialog while honoring the implicit stage directions given in the written form ("he whispers in your ear" etc...). All of that presented in VR, which would make having a dialog with these virtual characters more natural.

Looking at the tech that exists today, I believe that we are about halfway there.

mruts2• 7 years ago

It seems like MDL uses both greater/less than and parens. Is this analogous to Scheme using both brackets and parens? That is, just for readability?

jerf3• 7 years ago

It was the early 80s. Readability wouldn't be invented for about another 20 years or so.

Actually kinda serious. You need to be able to spare the RAM to have the comments, whitespace, large variables, breaking things nicely into functions, etc. I've got modules I've documented where the documentation alone would fail to fit into, say, a Commodore 64's RAM, as plain text. Early 1980s "big iron" would still be considered on the high end of embedded programming today. (In the sense that a Raspberry Pi is embedded programming.)

blattimwind• 7 years ago

That's a really interesting perspective. If I think about it an application really doesn't need to be large or complex at all to accrue a couple hundred kB of source code, libraries not included.

abecedarius3• 7 years ago

They developed these on a https://en.wikipedia.org/wiki/PDP-10. It's not like they were editing/compiling on the C64. Of course targeting small systems does influence the coding style, but...

tyingq• 7 years ago

COBOL is pretty readable.

jmull• 7 years ago

I first learned programming in BASIC on an 8K PET. The 8K was shared by the source code, data, the executing program and display memory.

Spaces were for suckers. Each line had a few bytes overhead so you put as many statements on each line as you could. You stuck to 1 character for your common variables (only 2 characters were significant anyway). You used the abbreviated name for commands. (This was around 81, 82.)

Something like this:

10 FORT=1TO10:?"IHN!":NEXTT:FORT=1TO1000:NEXTT:GOTO10"

This is from memory, so probably not actually correct code but this is meant to print "IHN!" 10 times, wait about a second and then repeat. T is used as a FOR loop variable twice.

kps• 7 years ago

<> is a FORM, which is evaluated. () is a LIST, which is not.

    <+ 1 2>          ↦  3
    '<+ 1 2>         ↦  <+ 1 2> 
    (+ 1 2)          ↦  (+ 1 2)

    <TYPE <+ 1 2>>   ↦  FIX
    <TYPE '<+ 1 2>>  ↦  FORM
    <TYPE (+ 1 2)>   ↦  LIST

mruts2• 7 years ago

So parens are just a quoted list? Seems a little unnecessary but I wasn’t alive so I guess I don’t know.

kmill6• 7 years ago

Angle brackets set out FORMs, parens set out LISTs. In English: FORMs are to function calls as LISTs are to list literals.

A translation from MDL to Python would look like this:

   <a b c>   ->   a(b, c)
   (a b c)   ->   [a, b, c]

(This is in contrast to a sibling comment: in MDL

   (<+ 1 2> <+ 3 4>)

would evaluate to

   (3 7)

)

However, there is a bit of a caveat in the translation. In MDL, atoms are self-evaluating, and it takes a call to LVAL or GVAL to obtain their stored values (as a shorthand, dot and comma prefixes give LVAL and GVAL calls, respectively). I haven't been able to find out from exactly which scope the evaluator gets the definition of a function. Which of the following two corresponds to what the evaluator actually does?

   <a b c>   ->   <.a b c>
   <a b c>   ->   <,a b c>

taradinoc8• 7 years ago

Function definitions normally go in the GVAL, so

  <DEFINE FOO () ...>

is equivalent to

  <SETG FOO <FUNCTION () ...>>

But EVAL will fall back to the LVAL if the atom doesn't have a GVAL. In other words:

  <A B C>

is equivalent to

  <APPLY <COND (<GASSIGNED? A> ,A)
               (<ASSIGNED? A> .A)
               (T <ERROR>)>
         B
         C>

https://github.com/taradinoc/mdl-docs/blob/master/docs/03-bu...

taradinoc8• 7 years ago

Angle brackets and parentheses are just two of the many flavors of brackets you'll find in MDL. First, you got your regular, "boring" brackets:

  <SETG X 0>           ;FORM
  (NORD BERT)          ;LIST
  [1 MONEY 2 SHOW]     ;VECTOR
  "Bracket-ish"        ;STRING
  { what goes here? }  ;TEMPLATE

Then you got your exciting brackets, where the first ! is mandatory but the second is optional:

  !<REST .FOO!>        ;SEGMENT
  ![2 4 8 16 32 64!]   ;UVECTOR

Then you got your hashtag brackets, which pop up in the corner of the TV screen reminding you that there's another way to CHTYPE:

  #FORM (+ 1 2)        ;"same as <+ 1 2>"

And your hashtag voucher brackets, which are really just defective angle brackets you can exchange for a free hashtag:

  <>                   ;"evaluates to #FALSE ()"

Finally, you got your bogus brackets, things that look like they could be a new kind of exciting brackets, but when you look closely they're just prefix operators that tend to appear before brackets:

  '(OSCAR WILDE)       ;"same as <QUOTE (OSCAR WILDE)>"
  %<+ 1 2>             ;"evaluate immediately"
  %%<CRLF>             ;"evaluate and discard result"

lousken• 7 years ago

FYI the article is not about ZFS ZIL but Zork Implementation Language

chungy• 7 years ago

I too assumed this was going to be about the ZFS Intent Log.

zzo38computer2• 7 years ago

I have not programmed in ZIL or Inform, but have made implementations of Z-machine, as well as a "Tricky Document" describing various optimizations that can be used in a Z-machine code (such as Black-Johansen text packing, the SET->BCOM optimization, storing introduction text in the input buffer, etc).

taradinoc8• 7 years ago

Do you like this?

zzo38computer2• 7 years ago

I like Tricky Document; that is why I wrote it. (Or do you mean something else? If you mean ZIL, well, then I don't know a lot about that.)

linsomniac• 7 years ago

What does this have to do with the Zfs Intent Log?!? :-)

myself248• 7 years ago

That's what I thought I was coming here to read about. HN article titles are the worst.

kmill6• 7 years ago

Looking through a bunch of code in the archive, it was interesting to come across the macro for TELL, which is used everywhere in games as a sort of printf, and it has some special printf-like codes to print out names of objects and such. There are also some functions that have been declared with DEFINE rather than ROUTINE. Both the macros and the DEFINEd functions use the features of the greater MDL at will.

What is interesting in particular is how it possibly reveals their setup: they had a full MDL interpreter running on their workstations, and ZIL is a subset that could be cross-compiled to a stripped-down target: the Z-machine.

It looks like ROUTINE, OBJECT, ROOM and such are macros (in sources unknown) that would define standard MDL objects, possibly stored in some structure for findability. Then, the cross-compiler would be a program that takes these objects and writes out some Z-code. The workflow probably allowed (1) playing the games directly in the interpreter without compilation (2) making small changes to a running game and (3) recompiling the Z-code after having made those small changes, without needing to reload all the game's source code.

Except for the cross-compilation step, this is the sort of environment the original Zork would have been written in, and I find it hard to imagine they would have accepted anything less.

pinewurst• 7 years ago

Except their “workstations” were the likes of VT100 terminals connected to their DEC-20.

mratzloff• 7 years ago

Very cool! Being an IF collector, I acquired the source code to Restaurant at the End of the Universe years ago. It's generous to call it unfinished—it's barely a game. I wasn't sure whether or not I could publish it, but it's great that this and much more are now available to everyone.

I also have some internal Infocom emails and photos and things. Not sure if posting those would be welcome by the participants since most of them are still alive.

RandomBacon• 7 years ago

Not sure if posting those would be welcome by the participants since most of them are still alive.

Genuinely curious... is it okay to post private details of a person once they're dead? There are things that I want to keep private while I'm alive, and that I hope stay private when I'm dead. Does other people's curiousity trump my wishes just because I'm no longer around?

LanceH2• 7 years ago

Once you're dead, what someone says or writes about you doesn't even need to be true (in the US).

aidenn0• 7 years ago

IANAL, but...

In the US, it's reasonably well established that you have no privacy rights after death[1].

That doesn't necessarily make it safe to publish details about someone after they are dead though as property rights do exist through the estate of the deceased, so you may have to end up defending yourself from a defamation suit.

In terms of it being "okay" there are unwritten rules about allowing some not-well-defined period of mourning before publishing private details, particularly ones that reflect poorly on the deceased.

Publishing e.g. private correspondence of people from just 100 years ago seems completely non-controversial though.

1: https://en.wikipedia.org/wiki/Post-mortem_privacy

setpatchaddress• 7 years ago

Short answer: we've already been down this path.

https://waxy.org/2008/04/milliways_infocoms_unreleased_seque...

edit: if I had access to that data, I'd be interested in finding any hidden source files -- there's a copy of the source to a version of the ZAP assembler hiding in one of the minizork repos Jason Scott posted, for example. Might there be some others?

Tepix• 7 years ago

Wow, these are layers upon layers of archaic stuff. The article got me curiohs about Inform7 so i watched the introductory screencast on Vimeo https://vimeo.com/4221277 . I‘m blown away! It appears to be easy to learn, feature rich and also well documented. Oh and it‘s also free! If you are interested in IF, go check it out.

What is ZIL anyway?

Related Stories