Roads to Xanadu

I worked with him for a while. The most frustrating part is that he wants Xanadu to be a closed-source, for-profit, 1988 style piece of capitalist software with potentially, maybe, open standards.

The content he wants to be DRM'd through micropayments and all for-profit with fine-tuned rights and fees for attributions etc.

Also things like multi-phasic design where you do 80% and then the other 20% is a non-starter. He wants a waterfall-model cathedral or nothing at all.

So I really wanted to work more with him but he couldn't get that no matter how awesome (it involves a file that maintains binary offset pointers to another file WITHOUT a source file that generates both so yes, you append to byte 1 and then you need to ROLL YOUR OWN tool to change the offsets in the offsets file - oh yeah, they're also in a binary format), because of it's systemic structure, he was effectually just proposing another proprietary locked-down format that won't go anywhere and was imposing a structure that will ONLY lead to one-off proof-of-concepts and prototypes.

Which is exactly what we've seen. For 50 years. I have a VM with demos in Java, TCL, and many other bygone languages written by people like myself. I remember saying:

"There's a deeper structural reason why the web uses HTML and not Microsoft DOC as its language."

Everything is pay per view in Xanadu. That's the problem.

It's also why it attracts economists and developers of the libertarian persuasion.

I met the Xanadu crowd during the period Autodesk was funding them. They were building a cross between Wikipedia and GitHub, with micropayments. Any subscriber could create a new document based on old ones. Readers of the new document paid to view it, and payments were apportioned between the contributors.

They had no idea how to distribute this beyond one machine. The rigid bidirectional linking system in Xanadu required everything to be kept in sync, and it all had to be under central control so the micropayments would be paid.

They also had to deal with the dead weight of Ted Nelson's database design, with "tumblers" and "zipper lists". It wasn't a good way to architect a database. Everything was explicitly cross-linked. Today, we know to separate the indices from the data, which all SQL databases do. You can create and rebuild indices as needed in an SQL database. If you need backlinks, you can create an index to do that. This separation of content from index makes the problem much simpler.

If you wanted to implement Xanadu today, it wouldn't be that hard to do it on top of an SQL database. (That's how Wikipedia works.) What Xanadu was supposed to do isn't all that complicated by modern standards.

Xanadu didn't do images. Nelson was into text. Xanadu had no image capability at all. (Video and audio storage were too expensive to even consider at the time.)

As computing became cheaper, Xanadu became unnecessary.

For those unfamiliar with Xanadu or Ted Nelson, http://www.wired.com/1995/06/xanadu/ is a great read. People who criticize it for being the "largest piece of vaporware in history" are technically correct, but rarely understand its potential or scope because we're so used to the current paradigm that it's hard to picture anything being better than "the way it is".

I've been meaning to blog about Xanadu myself. Nice to see it being talked about.