[ih] Internet analyses (Was Re: IPv8...)

[John Gilmore]

John Levine via Internet-history <internet-history at elists.isoc.org> wrote:
> Ted Nelson had been faffing about trying to implement his Xanadu
> hypertext idea for decades, but never got very far because it was much
> too ambitious. Tim built what's pretty much the implementable subset
> (i.e., one-way rather than two-way links) and that's the Web. Unlike
> its predecessors it worked, he had users that liked it, and here we
> are.

Contrasting points about Xanadu and WWW:

Xanadu was proprietary and thus discouraged people from working on it.
Its high level structure was documented (in 2 books), but none of the
implementation details.  WWW was well documented in specifications,
explanations, and in free/libre software.

Technically Xanadu was not federated -- it was designed to be centrally
run.  The plan was to have locally run "franchises" in towns, where you
could go to access it.  In theory people could run their own servers,
but every server, every user, and every document had to have a centrally
assigned and validated identity, and a financial account tied to that.
This was much more complicated both organizationally and technically.
WWW was federated by using the DNS to find each resource, and no central
coordination was required beyond the centrally managed DNS system.

Xanadu had an economic model that charged readers and funded publishers
and authors.  But that was never implemented.  WWW had no economic
model, was free to access and free to publish (plus the costs of
installing and learning software, and running servers).  I believe this
was WWW's biggest failing, since it led to the free-TV
advertising-funded business model.  This was rapidly followed by
centralized search engines that were paid by third party advertisers,
technical and social tricks for deep surveillance on readers, and
technical and social tricks for coerced attention extraction.  Xanadu
would probably have had attention coercion too, since publishers were
planned to be paid based on how much reading users did.

WWW was competently implemented many times.  Xanadu was incompletely
implemented several times, but the small team kept throwing away the
early implementations (rather than releasing them to gain experience and
a community) and rewriting it from scratch in niche languages like
Smalltalk.  Ultimately it failed because it never shipped a working
version -- nor any version at all.  By 1992 it had burned out its only
serious financial backer, and most of its volunteers, and it never
recovered thereafter.

By 1999, a dedicated volunteer (Roger Gregory) resurrected the last two
dead Xanadu code bases and gathered the permission to release them as
MIT-licensed software.  See:

  http://udanax.xanadu.com/

By then, nobody cared; WWW had grown exponentially and couldn't be
caught by competitors.  And the latest half-done code (Udanax Gold)
had bit-rotted due to its arcane implementation language, so it's quite
hard to even extract, let alone run or modify.  The earlier
implementation (Udanax Green) is probably buildable and runnable, but
does not implement most of the design (such as the economic model).  The
last updates I found were from 2015:

  https://xanadu.com.au/mail/xanadu/msg00485.html

An interesting bit for me is that the difficulty of implementing "2-way
links" seemed to play no part in Xanadu's failure.  Presumably, someone
could build them as an extension to WWW if the world really valued them.
The world of 2026 has much better tools for widely distributed
replicated databases: not just commercially rentable server farms and
Terabit Ethernets, but distributed hash tables, blockchains, BitTorrent,
and conventional transactional databases that can work at scale.  When
Xanadu was being implemented, there were no scalable networked
databases, because there were only a few thousand machines on real
networks in the whole world.  Airplane reservations, for example,
were still handled on a single mainframe running very custom software.

	John