This site runs best with JavaScript enabled.

The Pattern Language of Project Xanadu

Planted Jul 10, 2020
Last tended Apr 05, 2021
The Web
Digital Gardening

Ubiquitous Computing, also known as "

," was a technological dream of the 1990's. A group of researchers at Xerox PARC predicted that computing interfaces would disappear, dissolving into the objects and architecture of our homes, offices, and city streets.Mark Weiser's article
The Computer for the 21st Century
is the canonical source of the Ubicomp dream. Published in Scientific American in 1991.
It would become ubiquitous – always present and available without us needing to sit in front of large, unwieldy work stations to access it.

Us moderns will recognise this dream of computational power being available to us anywhere at any time. Smartphones made it happen. The internet of things contributed a little, but continues to be disappointing compared to the flexibility and power of our pocket computers. I can be walking around my kitchen, down Oxford St, or through Angkor Wat while firing off a series of computational commands. It requires no heavy, cumbersome equipment and very little cognitive attention.

The funny thing is, there is a large research community who believes we have not achieved the Ubicomp dream yet. They hold annual

and publish ambitious research papers. They're chasing after a particular manifestation of the dream. One that looks more like the seamless UX experiences depicted in the film
, instead of the messy reality we have cobbled together with underwhelming smart fridges, API services, and smartphone apps.

When futures arrive that do not exactly match our predictions and promises, we fail to recognise them. They are never as slick and shiny as we want them to be. They don't arrive in a single keynote address, ready to be shipped out in neat packages that are fully backwards compatible with our existing systems, avoiding the messiness of driver installations, legacy software patches, and incompatible hardware. The future arrives iteratively, piece by piece, in slow, unceremonious moments that fly under the radar.

But this is not an essay about Ubicomp. It's about another futuristic promise we've arrived at without realising it.

When it comes to visionary, before-its-time software, it's hard to find a better archetype than Ted Nelson's


It's infamous in the technology community as a sixty-year project that

. An infinite tale of unfulfilled promises and dashed dreams. There is however a working demo with a few of its features
(takes a minute to load).

Xanadu's failure to mature into a widespread, useable commercial product is great fodder a dramatic cover story. It's also the least interesting thing about it. The most interesting part of Xanadu is its Pattern Language.

are sets of flexible, reusable solutions to common design problems. Identifying and recognising these patterns gives us a framework to craft solutions with.

As a collection of design solutions, Xanadu's patterns suggest how we might construct relationships between pieces of data on the web that lead to whole new ways of thinking and seeing.

They're deeply relevant to problems we're wrestling with today.

  • How do we structure information and build relationships between pieces of data that help us see them across contexts, and clarify understanding?
  • How do we build systems that allow people to collaborate on shared documents without losing authorship? How do we credit and compensate authors based on their contributions?
  • How do we bring ideas and data from other sources into conversation with our own, while leaving a clear trail back to the origin?

Curiously enough, many of these patterns are showing up in modern manifestations.

Blog themes built with

, and
are beginning to look an awful lot like those original Xanadu demos and prototypes.

Indie projects like

, and
make some of its core features possible for the first time.

We're finding ways to backwards engineer the original spirit of Xanadu into the existing structures of the World Wide Web.

The Patterns

1. Visible Links

Ted Nelson, the inventor of Xanadu, has many gripes. One in particular is what he calls "jump links" - the very kind we all use around the web today.

Clicking a hyperlink is a bit of a gamble. It jumps you somewhere unknown. You find out when you get there.

Nelson wasn't keen on this surprise version of linking. Visibility of where the link connected to was essential in Xanadu.

One of the original UI mockups for a "Xanadoc" with visible links between text blocks

Link previews that contain excerpts of the destination page hint at this solution. You get to see where you're headed to before you decide to jump.

They're fairly common at this point. The

helped fuel this pattern by making it easy to generate preview cards in social media feeds.

Wikipedia added hover previews in 2018

Among the

community, creators like
Andy Matuschak
have popularised the pattern on personal websites and blogs.

Andy's notes system featuring link previews

Previews aren't quite the same as physical links between parts of a page though. You lose the fine-grained associations between specific paragraphs and lines.

A few members of

have been leaning into this association with Xanadu, and taking it to the next level.
Azlen Elza
recently managed to craft up this crazy demo inside of their Roam Research database:


Coming Soon

Feel free to bug me on twitter to finish writing this.

2. Parallel Documents

3. Transclusion

4. Bi-directonal Links

5. Version Control

6. Annotation

7. Copyright & Micropayments

The original Xanadu vision was ambitious. Unreasonably ambitious. The above feature list is a small selection.

Perhaps you sense the problem.
Solving all the things all at once gets tricky.

Instead, what we're now seeing is the ad-hoc, decentralised manifestation of Xanadu in bits and pieces.

People are building Xanadu without knowing what Xanadu is.
Which is the essence of a good pattern language; true patterns evolve naturally within systems, and are found rather than crafted. Christopher Alexander's

goes deep on this principle

In the grand scheme of things, 60 years isn't an extraordinary amount of time to spend solving a problem. It's about 0.09% of history since humans first started creating symbolic cultural artefacts Using

in Spain as a relative marker - 65,000 y.a.

Who knows, maybe Xanadu can still happen (if you clap for it) ✨

Want to share?

Join the newsletter

For weekly notes on visual thinking, ethical technology, and cultural anthropology.

Maggie Appleton © 2021