Hypermedia and Hyperbole

Posted: emacs

My partner and I are slowly working our way through the Myst series and have finally started Myst 3: Exile. We were both surprised by how much of a departure this entry is to the previous Myst games. The biggest change isn't so much the game itself as it is the implementation of freelook: the ability to look around a scene like it's a three-dimensional room. If this doesn't sound like a big achievement believe me, it is. The first two games are glorified powerpoint slides in comparison.

Original Myst (not talking about the remakes) was effectively a deck of static images with some clickable areas on top. Clicking on one of these areas could advance you to the next scene (i.e. movement) or jiggle an object so as to trigger an animation and play some sound effects. While the core gameplay and presentation is incredibly simple, it doesn't detract too much from the compelling setting and (mostly) interesting puzzles.

I was fascinated to learn that the control mechanisms for Myst 1 and 2 root back to early Cyan, before Myst was a thing. Rand and his brother Robyn made games for the Apple Macintosh with a piece of software called HyperCard, an application for building GUIs based on the principles of hypermedia.

All programs built in HyperCard consist of just two components: cards that store data and links that navigate between those cards or execute scripts. It's a simple system that offers a lot of flexibility; even today I think it would make an awesome prototyping tool. It also seems no coincidence that the scripting language is named HyperTalk, as the whole environment reminds me of Smalltalk.

All of the concepts of HyperCard (and the awesome late-80s nerd aesthetic) are well-explained in this interview with the creator, Bill Atkinson, and an enthusiast, Danny Goodman:

If that video piqued your interest, The Complete HyperCard Handbook, written by Danny Goodman in the video, is available to read online.

It always amazes me what creative people can create using simple tools. HyperCard and the early days of Cyan are no exception; it seems like the duo were able to find a lot of success by excelling within the constraints of their software.

Hypermedia systems are super interesting to me, particularly with the renewed interest in tools like htmx that promise to bring the hypermedia ethos back to the web. That said, this post isn't going to talk about how htmx is fighting against the SPA-explosion that the web industry has experienced over the past decade. Instead, I'd like to take a look at a far more esoteric hypermedia system: GNU Hyperbole.

I stumbled on this Emacs package recently and I had this overwhelming sense of deja vu due to its similarities to HyperCard. It was originally created in the 90s and retains an identity true to that era through some of its naming conventions (e.g. HyRolo, the rolodex extension). But more than that it is a full hypermedia engine, just like HyperCard, that is scriptable with Emacs Lisp.

The fact that Emacs contains an entire hypermedia system as an external package is probably unsurprising given its reputation as an operating system, but do people actually use it? The answer is yes, and rather extensively given the amount of material about it on the internet. There were no less than three talks about it during last year's EmacsConf:

All there are interesting glimpses into some of the functionality of Hyperbole, but I found that Ramin Honary's Building a Zettelkasten did the best job of answering the question of "why would anyone use this?"

As the title suggests, Ramin's talk demonstrates a Zettelkasten built on Hyperbole, using its built-in rolodex to store entries with scripted links between them. That's right, I said rolodex. Hyperbole is from the 90s, remember?

That said, a rolodex is probably not the best way to think about Hyperbole's record-keeping implementation (named HyRolo). It's not so much about managing contacts as it is managing record-oriented information. From that perspective it feels like a natural fit for a Zettelkasten, which literally means "slip box" in German.

Linking to a zettel in Hyperbole might look something like this:

<hyrolo-fgrep "Immanuel Kant">

In this case, the link reveals its implementation. It's a call to an Emacs Lisp function that greps over all HyRolo records for the string matching "Immanuel Kant".

Of course, you don't actually need to have an ugly link like that in your zettel text. Instead, you can create what's called an "explicit button" that hides the actual linking function code in a separate file. With an explicit button your zettel looks something like this:

<(Immanuel Kant)>

This link behaves in an identical fashion to the previous one, even though it hides away the details of the Emacs Lisp function powering the search. What's interesting is where it places those details. If you check your Hyperbole user directory, you can see the source code of the new button in a special file: _hypb:

("immanuel_kant" nil nil hyrolo-fgrep ("Immanuel Kant" nil) "me@my-pc" "20231025:04:20:37" nil nil)

It's just s-expressions all the way down!

Hyperbole is difficult to explain because it's such an abstract concept. Sure we can link between things, but what's the use? I think looking at the _hypb file and uncovering the Emacs Lisp engine underneath helps illustrate its flexibility; it's more than a linking engine because you can make your links any valid Emacs Lisp expression. Pretty cool.

Now, I don't think anyone is likely to create an entire game with Hyperbole, like Cyan did with HyperCard. Not because they couldn't per se, but asking players to learn Emacs as a prerequisite feels a tad much. But if there's anything to learn from the success of HyperCard, it's that simple interfaces empower creative people to build interesting things and I think Hyperbole exemplifies that same characteristic.


Thanks for reading! Send your comments to [email protected].