The Xanadu Saga

[David Farber <farber () central cis upenn edu>]

Date: Mon, 6 Dec 1993 17:00:47 -0500
From: shap () viper cis upenn edu (Jonathan Shapiro)


Dave:


Recently you forwarded a review of Hypertext '93, which included some
comments about Xanadu Light.  I thought that a brief history might be
of interest to your readers.


Depending on who you ask, the Project Xanadu has been running since
1960 or 1961.  It was Ted Nelson's brainchild, and after some time he
got Roger Gregory, Mark Miller, and others to start in on an
implementation in Ann Arbor.  That effort was volunteer, and the
hardware then available wasn't nearly ready for the problem.
Eventually the coding team disbanded, and that iteration of the system
is long gone.


One of the principle issues that defeated that effort was a struggle
for control between Ted, who had a specific vision for Xanadu, and the
programmers, who had ideas about eventually getting paid for their
efforts.


Nearly a decade passed in which the project proceeded as a volunteer
effort.  In 1987 (or thereabouts), Roger Gregory showed a prototype
system to John Walker of Autodesk.  After some discussion, Autodesk
invested a sizable chunk of money in the Xanadu Operating Company.  In
return, Autodesk received operating control of the company.


In parallel with the Autodesk investment, Mark Miller managed to
arrange something called the Silver Agreement.  In this agreement, Ted
agreed to allow the development team to pursue commercialization of
the Xanadu system, subject to the restriction that anything involving
royalty collection was Ted's exclusive domain.  The Xanadu Operating
Company was formed to do this.  While Ted was the largest individual
shareholder in XOC, he did not control the company.  This agreement is
what enabled the Autodesk investment.


Walker's motivation was twofold: he was interested in whiz-bang
technology and he had realized that Autodesk's meteoric growth could
not be sustained unless they continued to generate BIG innovations.
He believed that long-term investment in innovation was the best way
to get large payoffs.  His book, _The_Autodesk_File_, is fascinating
reading for anyone interested in the rise of Autodesk.


Around 1990, Walker decided to relocate to Switzerland.  It has been
alleged that this was for tax reasons. He was also bitterly upset that
Autodesk had become a large (and in his view disfunctional) company.
Autodesk's growth was slowing, and after a year-long search and a
major reorganization, Autodesk hired Carol Bartz (previously at Sun)
as their new CEO.  She took over in 1992.


Carol sees Autodesk as a CAD company first, and took rapid action to
excise some of Autodesk's wilder investments.  In her eyes, the Xanadu
Operating Company was one of the excesses.  By this point Project
Xanadu had been going for 25 years without a product. The XOC effort
was essentially unrelated to the balance of the Autodesk product line,
both existing and planned.


The Xanadu group decided that if Autodesk was going to kill them
anyway, they had nothing to lose by asking to be spun out.  Since I
had previous startup experience, I was asked to take over the Xanadu
Operating Company as CEO to handle the spinout and build them a new
board.


After a lot of negotiation, Autodesk agreed to spin the Xanadu
Operating Company out, and friendly press announcements were published
by all.  The details of the agreement are confidential, but I can say
that they provided Xanadu with a generous (under the circumstances)
amount of bridge funding and rights to the source code.


It quickly became clear that Ted's involvement in the company was not
consistent with it's business success.  I stepped down as CEO, and Ted
received the rights to the source code and the trademark in exchange
for returning control of the company.


The Xanadu technology was eventually licensed exclusively to Memex
Information Systems.  They have not made their payment schedule, and
the technology is probably available for license to any bidder who wil
pay for it.  Most of the system is written in a custom Smalltalk
subset that generates (surprisingly readable) C++.


The engineering team joined Memex as an ancillary part of the deal,
but has since moved on to other things.  When last seen, Roger Gregory
was working with Ted on the Xanadu public information system
("zippy"), which may be what Ted is now touting as "Xanadu Light."


While I would dearly love to see Ted succeed in his current efforts,
my personal assessment is that the name lives on, but Xanadu is
essentially dead.  At this point, there is no money, no engineering
team, and no product.


The tragedy of it all is that within a month of the spinout the
implemenation was working reliably enough to demonstrate, and it
really did do most of what the XOC team claimed it would.  With the
team disbanded, we will likely never know if it would all have worked
in the end.  It certainly stands out as one of the most dramatic
examples of underestimated complexity in the history of the industry.






Jonathan S. Shapiro