Hacker News Reader

Comment by specialist | Hacker News Reader

"...he had great visionary scale ideas, but couldn't describe a concrete implementation."

I revisit Xanadu every couple of years. For its time, its "closed" architecture seemed reasonable. How else would one implement persistent two-way links? Once you kinda grasp Xanadu, kludges like pingbacks and link liveness checkers seem ridiculous.

Alas, Ted Nelson, like myself, didn't anticipate, or was slow to fully appreciate, how DNS, HTTP, URL/URI would up-end the world. Whatever else was going on with Nelson, Xanadu, Autodesk, whatever, "the web" tsunami obliterated competing notions.

Animats' comment (below) about git (eg github) being today's closest realization of the Xanadu ideal is insightful.

View on HN

"its "closed" architecture seemed reasonable. How else would one implement persistent two-way links"

That's a great example of a concrete implementation detail - unbreakable links means a central link database - that has significant implications in the shape of the vision. For example the only way I can think of to do unbreakable links but also distributed is an immutable public ledger perhaps something like a blockchain. And now we can talk about specifics like what other properties does a blockchain implementation shape? What does a database app with HTML forms and such look like? But did Nelson understand this? Can anyone cite an implementation source that goes into Nelson's analysis of this stuff?

Nelson was not a programmer. However, Mark S. Miller told me that he came up with http://e-drexler.com/d/09/00/AgoricsPapers/agoricpapers.html out of thinking about how to architect the full distributed Xanadu system. I believe Nick Szabo was heavily influenced by Miller's ideas in his first paper on smart contracts: it references Miller and mentions that Szabo did work for the company started up to commercialize those ideas (Agorics). It's not quite a coincidence that you think of blockchains.

I don't know how Xanadu was supposed to work on the inside. I do have a dim vague memory of seeing a mailing-list post many years later in response to someone seeking the earliest prior art on using content hashes for a distributed content-addressable store -- from someone in Xanadu. That's all I got in my memory. I want to say they also mentioned Merkle trees, but I may be confabulating. (I'd be surprised if Miller didn't know Merkle personally -- they both worked with Eric Drexler, and Drexler invented his ent just a few years after Merkle patented his tree.)

Robin Hanson published a paper back then with a non-Xanadu design sketch with similar goals: http://dl.acm.org/citation.cfm?id=43938 but it's paywalled. (Hanson also got involved with Xanadu for a little while.)

IPFS solves the addressability, availability and decentralized storage for this kind of thing.

> For its time, its "closed" architecture seemed reasonable.

The currently dominating platform gives the impression that "closed" has won nowadays.

Reply on news.ycombinator.com