Macsyma

» see macsyma.com as it lived, on the wayback machine

~ the obituary ~

Macsyma did algebra. Not arithmetic, which computers had been doing for twenty years by 1968, but the other thing: give it an integral and it gave you back an antiderivative, as an expression, the way a person with a pencil would. The name is a compression of Project MAC's Symbolic Manipulator. Work started in july 1968 under Carl Engelman and William Martin at MIT, and Joel Moses ran it for most of the decade after that. It was written in Maclisp and grew into one of the largest Lisp programs anyone had attempted, which is part of why it ended up running on Multics and on Lisp machines rather than anything ordinary.

The mathematics inside it was genuinely new work, not an implementation of known methods. Joel Moses's thesis on symbolic integration and Bill Gosper's method for hypergeometric summation were research results that happened to ship as menu commands, and Barry Trager's work on integrating algebraic functions went in the same way. For roughly fifteen years, if you wanted a machine to manipulate an expression rather than evaluate it, Macsyma was more or less the only serious option on earth.

I started engineering school in the early 2000s, by which point the symbolic math you were handed was Mathematica or Maple, and the free option people passed around was Maxima. Maxima is Macsyma with the commercial half amputated, and I did not know that at the time. The original name was never something I typed into anything. It had been in a warehouse for years before I had any reason to look for it.

~ the case file ~

~ the climb ~

MIT licensed the thing to Symbolics in 1982, after granting Arthur D. Little the right to broker a deal, and Macsyma became a product with a price. Symbolics made Lisp machines, expensive dedicated workstations for an AI industry that was briefly convinced of itself, and it initially held back the VAX and Sun versions of Macsyma on the reasonable theory that a customer who could run it on hardware they already owned would not buy a Lisp machine. Those ports came out in the mid-eighties once that logic stopped working.

Richard Petti took over the business in the second half of 1986. He cut headcount while expanding the sales side, and pushed the developers toward whatever customers were actually asking for. Revenue roughly doubled in 1987. That was the high point in every sense: a large majority of the market, a product with fifteen years of research inside it, and a parent company about to fall apart underneath it.

~ the slide ~

Two competitors were growing at once. Stephen Wolfram's Mathematica arrived in 1988 with a notebook interface and real graphics, and Maple came out of Waterloo cheaper and faster on small machines. Neither was mathematically deeper than Macsyma. Both were far better at being software, which by the late eighties was the part that decided purchases. From 70 percent of the market in 1987, Macsyma was at 1 percent by 1992.

Russell Noftsker and Richard Petti founded Macsyma Inc. in 1992 and bought the rights out of the collapsing Symbolics. They rebuilt a good deal of the product. What they could not do was buy attention. Mathsoft spent something like ten million dollars on direct mail at very low prices right through the window when Macsyma Inc. was still reconstructing, and it took the growth. The staff was too small to add the graphics and interface work that would have been needed to answer Mathematica, and every large account was already somebody else's.

~ what got buried with it ~

In 1999 Tenedos LLC, the holding company that had already absorbed Symbolics, bought Macsyma. It never released the product commercially again. Over the following year the employees left, and the assets, including the servers holding the source code, went into storage. There was no announcement of a shutdown because there was nothing left running to shut down.

The survival happened through a side door. The Department of Energy had been distributing the 1982 MIT version to academics for years, and William Schelter maintained it. That branch was released under the GPL in 1999, the same year the commercial line went into a warehouse, and it is still maintained now under the name Maxima. A student today can run code descended from Project MAC in 1968 for nothing. The company that owned the better version of it managed to lose all of that value in the time it took to sign the paperwork.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »