Udanax-Gold vs. OpenXanadu vs. Xudanu

A detailed comparison of three Xanadu-lineage systems: the 1998 open-sourced codebase, the 2014 web demo, and the 2026 independent implementation.

1. Timeline: 60+ Years of Xanadu

2. At a Glance

3. Udanax-Gold 1998

In 1998, facing the irrelevance brought by the Web's dominance, the Xanadu team released their source code as Project Udanax — two codebases:

• Udanax Green (formerly Xanadu 88.1) — the C version, developed 1979–1988 by the Swarthmore team. Focused on "global transclusive address space" — tumblers, basic transclusion, and the FeBe (Frontend-Backend) protocol. The earlier, simpler system.
• Udanax Gold (formerly Xanadu 92.1) — the Smalltalk + C++ version, developed 1988–1995. Developed in Smalltalk and cross-compiled to C++. Focused on "full transclusive versioning" — General Enfilade Theory, The Ent, deep versioning, and the royalty system. The more complete and ambitious codebase.

This document focuses on Gold, which contains the advanced algorithms (enfilades, intercomparison, royalties) that influenced Xudanu's design. The release was under the Xanadu X11 license, which is source-visible but not open source by modern definitions (OSI-incompatible due to restrictions on derivative works).

The release had a dual purpose: to establish prior art against software patents, and to let the world see what Xanadu actually was after 38 years of vaporware accusations.

Architecture

Strengths

• ✓ Complete algorithm suite — the only implementation with the full set of Xanadu algorithms (enfilades, tumblers, transclusion, intercomparison, royalties)
• ✓ Battle-designed data structure — the enfilade tree was refined over 15+ years by Roger Gregory and Mark Miller
• ✓ Royalty system — built-in micropayment tracking, still ahead of its time
• ✓ Deep versioning — every revision preserved with full provenance
• ✓ Historical significance — the canonical reference implementation of Ted Nelson's vision

Limitations

• ✗ Difficult to build today — requires significant effort to set up the period C++ toolchain, resolve dependencies, and work through minimal build documentation. The code is complex and cross-compiled from Smalltalk, adding to the challenge.
• ✗ License incompatibility — Xanadu X11 is not OSI-approved; derivatives are restricted
• ✗ No concurrency — lock-based work grabbing prevents real-time collaboration
• ✗ No web integration — custom protocol, no HTTP/REST/WebSocket support
• ✗ No content-level crypto — had session auth and access controls, but no per-author attribution signatures, no content hash verification, and no transport encryption
• ✗ Dormant publicly — no public commits since ~1999. Continued internal development.
• ✗ Steep learning curve — enfilade concepts are complex and poorly documented

4. OpenXanadu 2014

In 2014, Ted Nelson released OpenXanadu as a web page at xanadu.com. Despite the name, it was explicitly not open source — the site stated it was "open" in the sense that "you can see all the parts." It was described as "a working deliverable," the first in 54 years of Project Xanadu's history.

The demo showed parallel documents — multiple source documents displayed side-by-side with visible transclusion connections between them. A user could see how a passage in one document was quoted or derived from another, with the connections drawn as visible lines.

What It Is

Strengths

• ✓ Visual proof of concept — demonstrates that the Xanadu experience is comprehensible and compelling
• ✓ Ted Nelson's seal of approval — the only "official" Xanadu deliverable
• ✓ Accessible — runs in any browser, no installation required
• ✓ Shows parallel documents — the core Xanadu visualization that 54 years of promises hadn't delivered

Limitations

• ✗ Not open source — despite the "Open" in the name, no source license granted
• ✗ Read-only — cannot create, edit, or link documents
• ✗ No backend — no server, no storage, no networking
• ✗ No collaboration — single-user, view-only
• ✗ No versioning — shows static snapshots, no revision history
• ✗ Slow to load — "takes a while to open because it's downloading a lot"
• ✗ No API — cannot be extended or integrated
• ✗ Stale — no updates since 2014
• ✗ Anti-Web rhetoric — site claims Tim Berners-Lee "stole" the idea and that the Web has a "terrible programming language"

5. Xudanu 2026

Xudanu began as a direct reading of the Udanax-Gold C++ source, migrating concepts into Rust. Early development followed Gold's design closely. The project then diverged significantly: the adoption of BLAKE3 for content hashing opened the door to verifiable cross-server content (impossible in Gold's trust model); React and WebSocket replaced the traditional client architecture; and most fundamentally, the O-tree CRDT with inline transclusion elements replaced the enfilade as the core data structure. Many components from Gold were dropped along the way — the FeBe protocol, the enfilade tree, the Smalltalk cross-compilation pipeline. What remains is the conceptual DNA: tumblers, transclusion, typed links, bidirectional connections, provenance tracing, and the principle that links are unbreakable.

Architecture

Strengths

• ✓ Buildable and runnable — cargo build + npm install, no exotic toolchain
• ✓ Apache 2.0 licensed — true open source, commercially usable, derivative-friendly
• ✓ Real-time collaboration — O-tree CRDT enables concurrent multi-user editing without locks
• ✓ Modern cryptography — Ed25519 attribution signatures, BLAKE3 content verification, TLS, tamper-evident audit trail
• ✓ Cross-server networking — domain tumblers make content globally addressable; BLAKE3 hash prevents substitution
• ✓ Visual link system — coloured typed links with margin description boxes, connecting lines, resolve/edit lifecycle
• ✓ Active development — 2,444 backend tests, 412 frontend tests, regular commits
• ✓ Rich text editing — headings, lists, blockquotes, code blocks, bold, italic
• ✓ Transcopyright licensing — first implementation of Ted Nelson's TCo license, with compliance badges
• ✓ Reactive UI — Zustand store propagates changes instantly across panels
• ✓ Image support — upload, crop, resize, drag-and-drop, caption persistence
• ✓ Web-native — HTTP API, WebSocket sync, runs in any modern browser
• ✓ Compound documents — inline transclusion in the CRDT itself (no side-table drift)
• ✓ Span migration — links and annotations survive edits to underlying text

Limitations

• ✗ No enfilade — the O-tree CRDT is conceptually different; some Gold algorithms (cranking, umbing) have no equivalent
• ✓ Intercomparison implemented — side-by-side document comparison with bezier connections, shared passage detection, coverage metrics, and AI-powered diff narration
• ✗ No royalty system — byte tracking exists (record_royalty) but no payment mechanism; Transcopyright licensing (FR-24) provides the legal framework
• ✗ Stable release Not production-hardened — v0.9.x, still in active developmentmdash; v1.0.1, production-ready for self-hosted use
• ✗ Single-server focus — cluster federation (FR-3) exists but is optional; cross-server (FR-7) is the primary model
• ✗ No mobile client — desktop browser only (though responsive design works on tablets)
• ✗ Wasm incomplete — browser-side CRDT compilation is stubbed but not functional

6. Feature-by-Feature Comparison

7. Data Structures: Enfilades vs. O-Tree

The most fundamental architectural difference is the underlying data structure for storing document content.

8. Networking: Xanadu Protocol vs. HTTP + Tumblers

9. What Xudanu Inherits from Gold

Despite sharing no code, Xudanu draws heavily on the conceptual framework established by Udanax-Gold:

10. What Xudanu Does Differently

11. The 17 Rules: Compliance Scorecard

Ted Nelson defined 17 requirements for a true Xanadu system. Here's how each implementation measures up:

12. Verdict