you can kind of break the history of hypertext into three lineages, with different lenses:

  • the engineer (Doug Engelbart’s lineage)
  • the scholar (Ted Nelson’s lineage) and, starting in the 80s, and
  • the publisher (which ate the other two)

Ted & Doug were very concerned with hypertext as an evolving thing — with creating hypertext iteratively, and using the affordances of hypertext as a tool for thought. Hypertext as marginal notes, correspondence tables, new indexes, common-place books, revision history.

This is how the social dimension was expected to play in some of these systems. For Doug, NLS/Augment was this collaboration platform for tight-knit engineering teams. For Ted, Xanadu was a natural extension of the republic of letters, making visible the invisible threads of scholarly history. You take previously fallible processes like quoting, citation, and marginal notes, & remove manual effort to free up the imaginative use.

Thing is, collaborative hypertext systems weren’t super interesting on small home computers, which had poor networking support & not much storage. You saw them in the navy, or at universities, or other places where a home brewed local hypertext system was put on top of time share. (And this also doesn’t mean these systems weren’t hampered. Like, InterMedia crosses more off the is-it-hypertext checklist than the web but you watch a demo and it was clearly an enormous pain to use because they followed the mac UX manual — making it easy to learn by removing everything that could possibly make an expert faster or better at doing something than a novice!)

This is where we start to see the third lineage. While hypertext systems for home computers in the 80s almost always shipped with hypertext editors, they would generally let you export your generated hypertext and ship it with a “light” version of the software with no editing. This is pretty natural. If you can’t collaborate on hypertext because it’s nearly impossible to coordinate or merge changes (because you’re sending floppies through the mail or something) why not treat hypertext stacks as finished products — like books?

This is the stuff that the hypertext literature people (and the cybertext people, etc) ended up analyzing. And, if you’re forced by the software into being a consumer of some hypertext — your only choices being which links to click — then the value propositions are overblown (as claimed in the book Cybertext).

Once the idea that you could have some “finished” hypertext stack shrink-wrapped on a shelf showed up (and this was actually how a lot of games were produced, like Myst), it became natural to professionalize it. So, where basically all previous hypertext systems had internal databases manipulated by graphical editors (what we might be tempted to call WYSIWYG), we started to see systems like the web where folks were expected to write markup by hand. When you have a producer-consumer distinction for a medium, the producer side has incentives to professionalize (and to gatekeep), and no incentives to make things easier for non-producers or non-professionals. You can win by having good content or by being one of the few who have put in the time to learn the rules of how to format content, and the more arcane the rules the more job security you have!

We didn’t really have collaborative hypertext on the web again until the advent of the wiki.

Wikis, while they on paper support things like backlinking and transclusion (the former through the use of a database and the latter through mere duplication plus a forward link), swallowed the markup pill and the mutability pill and therefore can’t do what real hypertext does.

For instance, on a wiki, there’s only one newest version of whatever page. Revision history is a line, not a tree (or even a DAG). There’s on-paper transclusion but it’s not real because it can be falsified or rendered out of date. So you can’t have transpointing windows either. Because it’s built on web tech (specifically HTML, which has a tree-shaped tag hierarchy), “backlinks” and “forward links” are actually just special cases of forward links to anchors. You cannot link a section to another section: the rules of markup say that tags must nest.

In other words, the ramifications of getting rid of a focus on collaboration led to design decisions that made other core features of hypertext hard or even impossible to implement, even when collaboration was reintroduced. (We have web-based xanadu demos and prototypes, and they have elaborate workarounds for the limitations of the web. Some of these workarounds, such as those for CORS, can’t scale the way we need them to.)

I’ve talked a lot about the publishing model because it’s what I hate most, but let’s compare Ted & Doug’s models too.

Ted’s model is expected to be global — it’s for the global intellectual community, as a platform for discourse. So, things like “see everybody who ever linked to something” actually isn’t socially viable. “See everything that links here” isn’t either. It’s hard to make technically viable too. You’d need a centralized database, which a lot of people assumed Xanadu would have (and some prototypes would have had, sort of). Actually, federation was on the table from very early. But basically you’d want to constrain who you saw. Think of, like, video game modding. Even if a game has a database of all possible mods, you don’t install all of them at once. Some are even incompatible. You’d break the game. Same with hypertext: you want sets of links that you think will add to your experience. At the same time, though, because this is a global system, you don’t want to be in a situation where you can’t see documents or links unless you personally know the author. Links shouldn’t be constrained by the author of the document either. You should be able to find links. Likewise, there’s the expectation that you’ll be able to make your own links, fork off a document with your own changes (creating a distinct revision history and a trail that leads back to where you forked it off), etc. Share documents & link packs with your friends, or not. Use cases for this: “Hey, I annotated all the references to The Divine Comedy in The Waste Land, and it works great with this annotation of all the references to the Illiad in The Divine Comedy!” “Yr Honor, I found all the litigation history that seems to bear on this ruling”

Meanwhile, NLS is for small teams who know each other and are working together. Like jira. So, you can see *everything*, but also, if you let the team get too big you end up with coordination problems & it gets fucked up like your average jira. I want to kind of qualify my statements on NLS because it’s the only one of these lineages where I have very limited experience & knowledge. I know a bit of the history from Engelbart’s publications & from coverage in books like @manjusrii ‘s Memory Machines. I recently found… and plan to dive deep into it, since the only thing on this page I have seen is the MOAD. There are people who have seen the MOAD who still say things like “Doug Engelbart’s main contribution is the invention of the mouse” with a straight face.

Anyway, I find the scholarly lineage most interesting not just because of my own workflow but because it fairly naturally extends back in time. Transpointing windows is a very complete and natural extension of marginal notes — so natural that it took a lot of work to get right! Hypertext really is “text, but more so” along Ong’s orality-literacy spectrum. The things we associate with text in literate cultures — density and complexity, permanence, and (with the codex) psychotechnologies for making use of random access — are all extended here.

In print, citation is this weird oral lacuna inside a text. The inter-textual element is not made literate! You have this fallible recording or some book that might exist, and something it might say on some page, but you need to expend a lot of effort to check. A jump link gets rid of the difficulties involved in finding the document — maybe, assuming that the link isn’t dead and is actually a link to the content (as opposed to meta-information like a store page). It doesn’t actually point to the relevant content, though. The absolute best case, with a web-style jump link, is that the other document happens to have a relevant section with an anchor in the section header. The normal case is that there are no anchors at all, and the relevant information is buried somewhere in the document. “Web surfing” was coined by analogy to “channel surfing” specifically because of the disjointedness of jump links. Just as your short term memory often loses data when you walk through a door, the change of context when you click a link makes it hard to connect the two ends. In other words, this is an ENDNOTE EMULATOR, and a bad one. Transpointing windows, on the other hand, expand marginal notes kaleidoscopically, with every document a potential marginal note to every other.

Resident hypertext crank. Author of Big and Small Computing: Trajectories for the Future of Software.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store