I've always wanted to meet Ted, but I'm affriad he'd show up at the museum or somesuch and I'd spend the entire time asking him Celeste Holm questions Chris
When I went to Swarthmore in 1960, Ted Nelson (who had graduated a few years earlier) kept coming back because he's been a superstar there and couldn't figure out what to do next. Glad he thought of something. He is a bullshit artist with the emphasis on both words.
He's an interesting and very bright guy, but I understand what you mean. The conceptual work he did on hypertext was - and is - groundbreaking. Literary Machines raises many issues which modern hypertext systems (chiefly the Web, but also including the also-rans) don't and can't answer. He's made some very astute observations and our relationship with computers.
However, he's a dreamer, not a completer. Had Xanadu been implemented in a timely manner after it was conceived, it could well have been where the Web is today. The failure of Xanadu, up to and including the Autodesk years, is a matter of record. Ted views the Gary Wolfe article as a hatchet job; I've heard some of Ted's version of events from him. I suspect the truth lies somewhere between the two.
In many ways, Ted is a Charles Babbage for our time.
I've had a fascination with both Charles Babbage and Ted Nelson since the age of eighteen (I can blame Hyperland and The Difference Engine respectively), and have come to know Ted fairly well in person.
The parallels are striking. They're both creative polymaths who struck upon an idea well before its time (automatic computation by mechanical or other means, and a global hypertextual document system). They both received considerable support towards realising their ideas, and both failed due to a combination of irascibility and an inability to settle on a version of their idea that was 'good enough'.
In Babbage's case, this went as far as burning through £17000 of public money with little to show apart from some finely machined parts and some precision engineering equipment, which he promptly lost in a dispute with his engineer, Clements. Babbage then asked for more money on the grounds that he'd thought of a much jollier wheeze than the tired old Difference Engine (No.2 - he'd already had one major ground-up redesign), and only needed a little more money to build his whizzy new Analytical Engine.
As far as I'm aware, however, Ted has never shown an aversion to hurdy-gurdies or to street musicians in general, although he does claim to have invented to rock opera.
The devil's in the details with these things. The outline of Xanadu in Literary Machines (which corresponds to the 1987 version that Autodesk paid to be developed) is very detailed, but what is missing is very telling.
There's a rather clever scalable addressing system, but the algorithm for searching through the resulting data structure (the enfilade algorithm) has never effectively been published. There was source published about five years ago in the Xanadu Green and Gold releases, but that was a bunch of uncommented Smalltalk and C code. Ted's claims for that algorithm are almost certainly an exaggeration of its actual capabilities.
Similarly, there's a fairly detailed sketch of a client-server API (the FEBE, or Front-End-Back-End, protocol). Literary Machines includes the throwaway line that the corresponding BEBE (Back-End-Back-End) protocol was under development. This is *the* crucial part of the system, to my mind; it determines whether Xanadu could be a viable *global* hypertext system. My PhD research touched on some closely related topics, and I studied all I could about Xanadu. I finally got to ask Ted about it while he was living in Southampton, and he admitted that the BEBE had never been written.
I wouldn't go so far as to say that Xanadu couldn't be implemented, but getting it to scale while retaining global link consistency would be a significant problem.
That was the impression I got. For a single data repository it would all work - as soon as you made it a collection of different data repositories you'd be back to the current situation where links would break and data would get out of synch.
It strikes me that trying to maintain perfect hypertext is the Aneristic Illusion. Both UNIX and the Web demonstrate that it's actually mostly fine if things break, as long as they're useful when they're working.
Oh, it's infuriating then UNIX's error semantics say "no, I'm going to hard fail, rather than retry, on this I/O error" 99% of the way through processing 84.6GB of data (or, er, when a link goes 404). But I'm posting this to the Web using a UNIX clone, not to Xanadu using a MULTICS clone. If we can do better, it's not evident that it can be done from nil.
In many ways, Ted is a Charles Babbage for our time.
Does that mean that in 120 years we'll have working versions of Xanadu built by the London Science Museum that my grandkids we'll be taught to use for weekly demonstrations?
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
![[identity profile]](https://www.dreamwidth.org/img/silk/identity/openid.png){kind=small}
no subject
Date: 2007-05-17 04:22 pm (UTC)He has good taste, does our
no subject
Date: 2007-05-17 04:35 pm (UTC)Chris
no subject
Date: 2007-05-17 04:55 pm (UTC)no subject
Date: 2007-05-17 09:29 pm (UTC)no subject
Date: 2007-05-17 07:16 pm (UTC)no subject
Date: 2007-05-17 07:36 pm (UTC)However, he's a dreamer, not a completer. Had Xanadu been implemented in a timely manner after it was conceived, it could well have been where the Web is today. The failure of Xanadu, up to and including the Autodesk years, is a matter of record. Ted views the Gary Wolfe article as a hatchet job; I've heard some of Ted's version of events from him. I suspect the truth lies somewhere between the two.
In many ways, Ted is a Charles Babbage for our time.
no subject
Date: 2007-05-17 09:27 pm (UTC)no subject
Date: 2007-05-17 10:47 pm (UTC)The parallels are striking. They're both creative polymaths who struck upon an idea well before its time (automatic computation by mechanical or other means, and a global hypertextual document system). They both received considerable support towards realising their ideas, and both failed due to a combination of irascibility and an inability to settle on a version of their idea that was 'good enough'.
In Babbage's case, this went as far as burning through £17000 of public money with little to show apart from some finely machined parts and some precision engineering equipment, which he promptly lost in a dispute with his engineer, Clements. Babbage then asked for more money on the grounds that he'd thought of a much jollier wheeze than the tired old Difference Engine (No.2 - he'd already had one major ground-up redesign), and only needed a little more money to build his whizzy new Analytical Engine.
As far as I'm aware, however, Ted has never shown an aversion to hurdy-gurdies or to street musicians in general, although he does claim to have invented to rock opera.
no subject
Date: 2007-05-17 09:37 pm (UTC)Bit like Communism.
no subject
Date: 2007-05-17 10:28 pm (UTC)There's a rather clever scalable addressing system, but the algorithm for searching through the resulting data structure (the enfilade algorithm) has never effectively been published. There was source published about five years ago in the Xanadu Green and Gold releases, but that was a bunch of uncommented Smalltalk and C code. Ted's claims for that algorithm are almost certainly an exaggeration of its actual capabilities.
Similarly, there's a fairly detailed sketch of a client-server API (the FEBE, or Front-End-Back-End, protocol). Literary Machines includes the throwaway line that the corresponding BEBE (Back-End-Back-End) protocol was under development. This is *the* crucial part of the system, to my mind; it determines whether Xanadu could be a viable *global* hypertext system. My PhD research touched on some closely related topics, and I studied all I could about Xanadu. I finally got to ask Ted about it while he was living in Southampton, and he admitted that the BEBE had never been written.
I wouldn't go so far as to say that Xanadu couldn't be implemented, but getting it to scale while retaining global link consistency would be a significant problem.
no subject
Date: 2007-05-17 10:32 pm (UTC)no subject
Date: 2007-05-18 06:17 am (UTC)We did see a demo!
no subject
Date: 2007-05-18 07:31 am (UTC)What a fascinating speaker he is!
no subject
Date: 2007-05-21 09:29 pm (UTC)Oh, it's infuriating then UNIX's error semantics say "no, I'm going to hard fail, rather than retry, on this I/O error" 99% of the way through processing 84.6GB of data (or, er, when a link goes 404). But I'm posting this to the Web using a UNIX clone, not to Xanadu using a MULTICS clone. If we can do better, it's not evident that it can be done from nil.
</gross-oversimpliciations>no subject
Date: 2007-05-17 10:13 pm (UTC)Does that mean that in 120 years we'll have working versions of Xanadu built by the London Science Museum that my grandkids we'll be taught to use for weekly demonstrations?
Chris
no subject
Date: 2007-05-17 10:17 pm (UTC)no subject
Date: 2007-05-21 09:15 pm (UTC)no subject
Date: 2007-05-21 09:25 pm (UTC)