A docuverse of independent servers, connected by permanent, cryptographically verified, attribution-preserving references — the vision Ted Nelson had in 1960, working today.
In 1960, Ted Nelson envisioned a universe of documents — a docuverse — where content could be connected, reused, and traced across independent servers. Not a web of pages, but a network of connections at the character level. Every quotation would maintain its bond to the original. Every reuse would carry its full provenance.
The World Wide Web (1989) gave us one-way links between pages. Nelson's Xanadu was different: bidirectional, typed, permanent connections between passages — with content staying on its origin server, never copied, always verified.
Xudanu implements this vision.
Two independent Xudanu servers connected by a tumbler reference. Content stays on its origin server; references travel between them.
On July 10, 2026, we verified the full cross-server pipeline end-to-end with two independent Xudanu servers running on separate ports with separate data directories:
"127.0.0.1:8081".03ed.1.0.0No central registry was consulted. No shared database was synchronized. No federation protocol was used. Two sovereign servers discovered each other, exchanged a permanent reference, and shared verified content — exactly as Ted Nelson designed.
Every piece of content in the Xudanu network has a permanent, globally unique address called a tumbler. This is Nelson's innovation from the 1970s — a hierarchical, dotted notation that identifies content across servers.
A domain-based tumbler. The domain replaces Gold's numeric server IDs, enabling DNS self-routing.
The domain format ("alice.xudanu.com".03ed.1.0.0) is Xudanu's enhancement over Gold's original numeric format (1.03ed.1.0.0). DNS provides self-routing — no central registry needed.
Any Xudanu server can join the network by publishing its identity and content.
./xudanu-server run 0.0.0.0:8080 data \
--server-name "Alice's Literature Server" \
--server-description "Essays and annotations" \
--public-address alice.xudanu.com
Create a document, write text, and click Published. Published works become available via the public content API at /api/public/work/{id}.
curl https://alice.xudanu.com/.well-known/xudanu-server.json
Returns the server's identity, verifying key, name, description, and stats.
curl https://alice.xudanu.com/api/public/work/03ed
Returns the work text, BLAKE3 content hash, span provenance, and server identity.
Once two servers are running and have published content, users can create cross-server links:
Choose "Link to a remote server" in the Link Creator wizard.
Enter the tumbler from Server A (e.g., "alice.xudanu.com".03ed.1.0.0) and the BLAKE3 content hash from the public API.
The link is stored on Server B with the CrossServerRef payload (tumbler + hash + author info). Server B automatically notifies Server A of the incoming reference.
Clicking the link marker on Server B triggers cross-server resolution: fetch from Server A → verify BLAKE3 hash → cache locally → display content.
Cross-server content sharing uses three independent layers of verification:
| Layer | What it answers | How | Even over HTTP? |
|---|---|---|---|
| Tumbler | WHERE does content live? | Domain-based global address | Yes |
| BLAKE3 | WHAT is the content? | 256-bit hash verified on every fetch | Yes — mathematically impossible to forge |
| Ed25519 TOFU | WHO is the server? | Key pinned on first connection (like SSH) | Yes (after first pin) |
A man-in-the-middle cannot substitute different content because they would need to produce content with the exact same BLAKE3 hash — which would take 1022 years (longer than the age of the universe). The hash is stored at link creation time and verified on every fetch.
| Protection | Status |
|---|---|
| SSRF protection (reject loopback/private IPs) | Done |
| 5MB byte cap on fetches | Done |
| Trust gate (untrusted servers rejected) | Done |
| Ed25519 TOFU (pin on first use, warn on change) | Done |
| HTTPS pinning (refuse downgrade) | Done |
| Byte tracking (RoyaltyEntry per fetch) | Done |
| Rate limiting on public endpoints | Planned (#105) |
This implementation inherits directly from the Udanax-Gold hypertext model (open-sourced 1999), with modern adaptations:
| Gold concept | Xudanu implementation | Enhancement |
|---|---|---|
| Numeric tumblers | Domain-based tumblers | DNS replaces central registry |
| Enfilade content storage | O-tree CRDT + chunk store | Collaborative editing | Push-based sync | Pull-based resolution | Server stays sovereign |
| Custom binary protocol | HTTPS + JSON | Standard web protocols |
| Royalties (Rule 9) | RoyaltyEntry ledger | Byte tracking, no payments yet |
| Bidirectional links | Cross-server backlinks | Automatic notification |
The Xanadu project spans over six decades — one of the longest-running visions in computing history. Understanding this history helps explain why the cross-server network works the way it does, and why certain design decisions (tumblers, bidirectional links, content addressing) were worth the wait.
The Xanadu timeline: 39 years from vision to open source (1960–1999).
| Concept | Origin | Year | What it does |
|---|---|---|---|
| Transclusion | Nelson | 1965 | Content referenced, not copied — live links to the original |
| Enfilade | Miller, Gregory | 1978 | Data structure for efficient versioned content access |
| The Ent | Drexler | 1978 | Tree structure managing complex versioning with low overhead |
| Tumblers | Gregory, Miller | 1978 | Globally unique, hierarchical addresses for every passage |
| Bidirectional links | Nelson | 1965 | Both ends of a connection are always visible |
| PlaceHolder / FillDetector | Miller | 1992 | Lazy template slots that auto-fill when content appears |
| Royalties (Rule 9) | Nelson | 1960s | Micropayments to original authors when content is transcluded |
The Xanadu project faced many challenges over the decades:
But the ideas were sound. Ted Nelson's 1965 paper introduced concepts that the web still hasn't fully adopted 60 years later: bidirectional links, transclusion, and content addressing. The open-sourcing in 1999 allowed these ideas to survive and influence new implementations.
Xudanu is not a port of Udanax-Gold. It is a reimagining of Nelson's vision using modern technology, informed by the Gold codebase and its innovations:
Udanax-Gold vs. Xudanu: same vision, modern implementation.
The Xanadu project involved many brilliant minds over the decades: