I am going to stake my reputation as a literacy man: I believe that, using this alphabet, the illiterate will learn in one-tenth the time that it now takes; and that means that probably ten times as many people will learn to read.

Upper class: often flawlessly ethical, never moral. Working class: sometimes moral, often indifferent to formal ethics that they feel were set in place without their consent. The social language of appearing to avoid conflicts of interest and self-dealing is one in which upper-middle and upper class people are fluent. In terms of morality (avoiding harm, rather than avoiding violations of semiformal social principles) there is a negative correlation to SES, but it’s very slight because it turns out there are also scoundrels at the bottom.

Maestro Nelson had the same problem (but with way less funding than Maestro Engelbart enjoyed). He couldn’t build Xanadu because most of its parts didn’t exist, such as a way to name content that no authority issues, or storage cheap enough to never delete. His teams hand-built their own data structures for years, but the project died on the vine. Wired depicted Maestro Nelson ’s story as one of mismanagement, but I think he’d just specified Xanadu decades before its dependency tree existed.

I think it’s telling that in nei­ther Lib/Dream nor Fu­ture did I see any in­stances of ‘tran­sclu­sion’ on paper. Nel­son could have used any amount of side-by-side lay­out or range-transclusion in his books, be­cause soft­ware is no ob­sta­cle: [ Maestro Nelson ] was lay­ing them out by hand, and could draw or il­lus­trate or copy any­thing he wanted in any arrange­ment. But he didn’t, be­cause… it’s just not that use­ful for books. Not even his. (You don’t need tran­sclu­sions when you can just cite an ear­lier pas­sage.)

Xanadu, [ Maestro Nelson ] had a specific vision of what computing could be, which he called the dousers. He imagined a system that could keep every version of a document; hypertext links would know both their source and destination, and quotations would be kept by reference (rather than copy) so any included text maintained its identity and source. Everything, in this version of computing, would be intertwingled and xanalogical. We’re talking about attribution down to the span.

To be xanalogical is to follow two rules: Never copy, always reference (also known by a Nelsonism, transclusion). Never overwrite, always version. As Nelson designed it, Xanadu would manage the system’s complexity and massive bookkeeping burden itself, so users only experienced the benefits of a system that could never forget.

I briefly poked at the Xanadu PCs, im­pressed that they were run­ning at all, but I and most of the party-goers bounced off them. The UI was too alien. We re­ally needed to see James demo them, or some­thing like that: hy­per­text sys­tems do not lend them­selves to im­me­di­ate ex­plo­ration, es­pe­cially when they are run­ning in OSes on com­put­ers no one there has used in 25 years, if ever.

For ex­am­ple, he de­scribed how they pro­to­typed Xanadu in Smalltalk⁠ (which made sense, as a highly-productive, pleas­ant lan­guage/OS whose object-oriented model matches hy­per­me­dia beau­ti­fully), but then had to cross-compile it to C++⁠ and com­pile that… which took about a week to com­pile. Not a minute, or an hour, or even a day, but a week. (And I thought that Gwern.net’s multi-hour compile-times were bad for my de­vel­op­ment speed!) He also had to waste a lot of time deal­ing with C++ silli­ness and com­pile is­sues.

Like _@En­gel­bart_⁠, [@Nel­son] en­joys (if that is the word) the honor of being a liv­ing em­bod­i­ment of Cun­ning­ham’s Law⁠: be­com­ing one of those his­toric fig­ures whose im­por­tance was that they were so wrong on such an im­por­tant topic they helped pop­u­lar­ize that they in­spired oth­ers to be­come right.

I didn’t need to use them much to look at the screen dis­play­ing the stereo­typ­i­cal Project Xanadu demo, the open­ing of the Book of Gen­e­sis⁠ with its fa­mous lines zig-zagging off to the right to de­note tran­sclu­sions or com­men­tary on pas­sages, and have a sud­den re­al­iza­tion:

“Oh my god—It’s com­pletely un­read­able.”

The lines were con­fus­ing clut­ter, es­pe­cially as they criss­crossed (a peren­nial prob­lem in side­notes lay­out, made far worse by the out­lines). None of the ‘side­notes’ were read­able be­cause the screen was so small. Even as you sim­ply scrolled, for many pos­si­ble po­si­tions, due to the lines you were un­able to read any­thing! How could a doc­u­ment UI where often, you could read noth­ing, have ever seemed like a good idea?⁠⁠2⁠ The UI was just ter­ri­ble—it could never have worked. Even on a large screen like my 4k mon­i­tor, I wouldn’t want that.

And then I thought about the choice of text, and I re­al­ized that the UI wasn’t the real prob­lem; and the prob­lem, for all these many decades, wasn’t the team ei­ther. The whole con­cept of side-by-side range tran­sclu­sions is a so­lu­tion in search of a prob­lem.

Unified Arabic (UA) is basically a set of 30 letterforms, one for each letter of the Arabic alphabet, plus hamza and lam alef, eliminating the variant forms that make reading and writing Arabic difficult for beginners.

Xanadu wasn’t the vic­tim of “Worse is Bet­ter”⁠; it was just a so­lu­tion in search of a prob­lem.⁠⁠5⁠ (Some­thing sim­i­lar seems to be true of ⁠Ar­bital⁠: it had one real user, Eliezer Yud­kowsky, who had en­vi­sioned it, but strug­gled to com­mu­ni­cate the vi­sion or cre­ate a clear pro­to­type and set of use-cases, mocked up as nec­es­sary; and like Xanadu, it was a square peg of what ought to have been a long-term non-profit FLOSS ef­fort shoved into the round hole of a soft­ware startup com­pany.)

I think some dis­ap­point­ments in past soft­ware sys­tems be­come more com­pre­hen­si­ble when we re­mem­ber how much time and in­ge­nu­ity is spent work­ing around the lim­i­ta­tions—eg. McIl­roy 1982⁠ is jus­ti­fi­ably proud of the months of clever al­go­rithm de­sign & op­ti­miza­tion he did to get a use­ful spellchecker to fit in RAM in under 1MB that today we would write in a few lines of JavaScript (and done by an LLM⁠).

The col­lege mag­a­zines were a sur­prise and en­ter­tain­ing to leaf through: young [ Maestro Nelson ] al­ready liked to write, quite a lot, and dis­pense his ad­vice. Some­one men­tioned that Nel­son had dreamed of being a Hol­ly­wood di­rec­tor and re­gret­ted that he went into tech­nol­ogy in­stead; I thought that made sense and ex­plained some things about his au­teur ap­proach to soft­ware de­vel­op­ment (like his in­sis­tence he is ⁠“not a pro­gram­mer”⁠, 50 years later) or use of film-editing metaphors like “edit de­ci­sion lists”⁠.

There’s a tendency of ideologues to value ideological purity over fitness for use.

This is fine as long as people acknowledging their pure thought stuff is malleable.

One of Ted’s memorable issues was in say indexing documents. He’s wanted the markup to exist in a separate document that was byte addressable, not inline. So a separate index file.

I asked him “what about locales? Some systems are UTF 16.” And you know, instead of I dunno, admitting that not doing a packet, block, or inline format was a mistake, he just insisted it’ll be handled later. I asked him about cadence, how will you do versioning between the two, things are grossly inefficient if you change one byte, etc…

One of the other nice things of inline systems is you can get a partial file and still have all the context. With that system however, corruption is way too easy. There’s all these technical requirements to his grand idea that he asserts are completely absolutely necessary and have no technical equivalencies at all and do not have the obvious design and implementation limitations that can easily be overcome with very slight permutations in a way that still completely satisfies all practical usecases.

And he loves saying he’s not a programmer. It’s like come on Ted, trust the people who say you’re micromanaging of implementation is erroneous. It’s been like 50 years, some of your stuff doesn’t work.

He shouldn’t be bitter. Everyone technical in the field knows him. He’s up there with Stallman , Knuth , and Shannon in name recognition. I probably won’t even appear as a footnote no matter how hard I work. So yeah, mission fucking accomplished already Ted. Geez.

The fa­mous Project Xanadu UI/UX is a “sci­ence fic­tion in­ter­face”, like the 3D ges­ture in­ter­faces in the Mi­nor­ity Re­port movie⁠ or the vir­tual re­al­ity of Snow Crash⁠: every­one looks at them and is daz­zled and wowed by how cool they look, ex­cept they are a ter­ri­ble idea which would be a mis­ery to use and give you ⁠“go­rilla arm”⁠. You try out a pro­to­type or mockup, and al­most as soon as you start using them, you re­al­ize that no, this isn’t work­able and you have to throw it out com­pletely. This re­minded me of the Hol­ly­wood di­rec­tor com­ment: you could say that Project Xanadu is a Hol­ly­wood di­rec­tor’s idea of what a World Wide Web should be. You can only be­lieve in it if you never try to ac­tu­ally

The fa­mous Project Xanadu UI/UX is a “sci­ence fic­tion in­ter­face”, like the 3D ges­ture in­ter­faces in the Mi­nor­ity Re­port movie⁠ or the vir­tual re­al­ity of Snow Crash⁠: every­one looks at them and is daz­zled and wowed by how cool they look, ex­cept they are a ter­ri­ble idea which would be a mis­ery to use and give you ⁠“go­rilla arm”⁠. You try out a pro­to­type or mockup, and al­most as soon as you start using them, you re­al­ize that no, this isn’t work­able and you have to throw it out com­pletely.

Hear­ing some war sto­ries from the horse’s mouth helped put things in proper per­spec­tive.

What makes an agent an awkward collaborator also makes it a perfect citizen for Maestro Nelson ’s system. Agents can follow more layers of subtext than people can hold in their heads at once, such as the sources behind a quotation, any discussion around it, and stack traces attached to the exact code that ran. A fragment-based representation makes those references lossless, so each layer stays attached to the same span as the text changes. Agents can read these dimensions together, cite them so humans can audit their work, and patiently follow every link, every time. But those links need to actually exist somewhere in order for an agent to find them.

When I revisit Nelson’s vision now, I recognize DeltaDB’s design goals and the promise of Delta. After decades of squashing history to meet the limits of the human attention span, we’re in a new reality that would be better served by something more… Xanalogical.

In 1932, a Lebanese architect walked into a classroom at the American University of Beirut to fill in for a professor who taught basic Arabic typing skills. In an effort to welcome the class, he started typing ahlan wa sahlan (‘welcome’), but, finding it difficult to locate the right keys for the right variation of the letter heh, he mistakenly typed an initial heh form instead of a medial one. He noticed, however, that what he had typed was still perfectly legible. He suddenly realized that by reducing the number of letter variations, the problem of finding keys on the typewriter could be easily solved without affecting the legibility of the text. He decided then and there to work on unifying all the variations of the Arabic letters. The architect’s name was Nasri Khattar, and he called his project Unified Arabic.