[Axiom-developer] Why did Axiom fail in the 1990s?

[daly]

Scratchpad was a research project, not a commercial project.
The only reason that it became commercial was that Lou Gerstner,
hired from Nabisco to run IBM, set out to commercialize everything.

I'd like to blame Lou for everything but the collapse started
much earlier when the Federal Labs were closed putting an end
to much federally funded research. The trend eventually killed
off such great places as Xerox Labs, Bell Labs, and IBM Research.
(IBM Research exists but "it ain't the same", ask any old-timer)

I worked at IBM Research for years. Once Lou was hired there was
a push to "make research pay for itself". Scratchpad was going to
be sold in the market. We originally tried to sell it thru IBM.
However IBM had a requirement that programs must have extensive
documentation. And the documentation must be readable "at the
eighth grade level". There were programs which read the text and
flagged long words, long phrases, out-of-dictionary words, etc.
Since there was no way to document DistributedMultivariatePolynomials
under this stricture we could not market thru IBM.

We tried several paths. NAG was a logical path as they were well known
for their excellent mathematical tool (the NAG library) and their
ability to support mathematical software. Plus NAG had quality people
in the management team. Thus we managed to sell Scratchpad (as Axiom)
to NAG.

There was a huge cost however. None of the original IBM developers
(including me) were allowed to continue to work on Axiom. So Lou
managed to get money for the software but stripped it of the people.
We all were told to circle our chairs until we found some non-axiom
research that would pay for itself in real customer dollars. So
front-line computational mathematicians were stripped of the one
place where their skill sets could make a difference. (I ended
up working on a voice-recognition system for a bank).

Thus stripped of the people who made Axiom live, NAG had a huge task
on their hands. They could not afford to hire a 10-person support team
and open source did not exist at that time. (Well, it did since I
regularly shipped sources for scratchpad, but there was no "open
source developer network"). Without the people Axiom cannot be a
viable, growing product.

Thus I believe that NAG's only possible path was to use Axiom as a
market driver for their excellent NAG library. Given such a role Axiom
cannot and did not compete. I believe that once NAG connected to Maple
as an external library of choice it was clear that Axiom had to go.

Scratchpad was well ahead of the curve compared to Maple, MMA, etc.
Given the right support it could have been much better than it is
today. However, the IBM Research which Lou Gerstner shaped lacked any
sense of long term vision. What could "the 30 year horizon" mean to a
man who markets cookies (Lou's previous company)?

Scratchpad didn't die a natural death. It was killed by short term,
make it pay, thinking. Notre Dame took 300 years to build. Given the
short term mindset Lou would rather have built it overnight as a
prefab wooden house in a suburb using conventional tools and selling
for a nice markup.

Tim