I've recently read two separate accounts of the history of hypertext (here and here). Neither is particularly accurate, original, or insightful, but they share one odd characteristic: they leave Douglas Engelbart out of the story entirely. The genealogy of the Web, in these narratives, runs from Bush to Nelson to Berners-Lee.
Warning: in my opinion, both these articles misunderstand Bush's Memex, which each describes poorly but at some length. Bush's article is readily available and, like many popular science articles, is easy to read. There's a good book about Memex by Kahn and Nyce, which neither author appears to have consulted, and MIT hosted a symposium on Memex a few years ago whose proceedings would also be useful. I have long been skeptical of the true extent of the influence of Bush's essay, and this question awaits the work of a dedicated scholar capable of analyzing sources. I should also mention that Author #1 apparently misunderstands the genesis of Storyspace, while Author #2 misreads Landow to hilarious effect. Proceed at your own risk.
To omit Engelbart is strange, because not long ago it seemed likely that the forgotten progenitor would be Nelson, not Engelbart. It used to be Nelson who was left out of abbreviated accounts or dismissed as a visionary.
At the first hypertext conference, it was clear that the hypertext community was divided in two groups, the Nelsonites and the Engelbarters. (I was a Nelsonian; Engelbart's work was new to me but very exciting) Over the dinner when I first met (if memory serves) Randy Trigg, Cathy Marshall, and Polle Zellweger, I actually asked the PARC crew "are you Nelson or Engelbart?", and the question seemed perfectly natural. One reason the Hypertext Conference was important was that it provided a meeting ground; lots of people coming out of these two traditions had never met. Reading the early literature, you can see traces of this divide, chiefly in attitude toward hierarchy. Engelbart's work embraced hierarchy and outlining, Nelson's battle-cry was "Everything is intertwingled!"
In the early 1990's, outliners and stretchtext systems like Guide seemed the wave of the future, and increasing formality appeared to be the natural direction for hypertext. At the same time, Nelson's utopian insistence on the viability of Xanadu seemed an embarrassment and his liberatory rhetoric was a hard sell to corporate buyers. Engelbart (who had aerospace connections) and Bush (with an impeccable defense department heritage) seemed to overshadow Nelson entirely, and Nelson's followers felt the need to remind audiences of his role at every opportunity.
As it turned out, Xanadu wasn't utopian; one year we woke up and, there (more or less) it was. Over the course of the 90's, moreover, the pendulum swung against formality. Outliners disappeared into presentation systems, and then Microsoft killed the software category by giving away PowerPoint and outliners were nearly forgotten. The Engelbarter/Nelsonian split, which was social and geographic, was soon resolved. If it was remembered at all, it was seen (incorrectly) as a prelude to the battle of the engineers and the literati.
Just at the moment when Engelbart falls out of these histories, as it happens, the pendulum is swinging back toward Engelbart-style systems. Tinderbox and Radio Userland, for example, are each strongly influenced by Engelbart. Tinderbox, with its semi-formal attribute structure and template transclusion, echoes Aquanet and SEPIA, and its rhetoric sometimes recalls NLS/Augment. Radio, like Augment, is all about outlines -- and OPML may the first outliner since Augment to support outlines that cross machines and file systems. (Gopher did this too; it might be interesting to look at Gopher again....)
Engelbart also gave what is generally considered to be the best demo in the history of computing.