Multiplan

~ the obit ~

multiplan finished before i started. it shipped in august 1982, two years before i was born, and its last version came out around the time i was learning to read, so this is a grave i can only build out of documents. the thing that makes it worth building anyway is that i have used a piece of it. so has anyone who has ever opened a spreadsheet, without any way of knowing.

in 1982 microsoft did not make operating systems for a living so much as it made languages and applications for whatever hardware happened to exist. multiplan was its spreadsheet, called electronic paper internally, and it went out against visicalc, which had invented the category on the apple ii two years earlier. multiplan was in most respects the better program. it had named ranges and it could link one sheet to another, which visicalc could not do at all.

then lotus 1-2-3 arrived in january 1983 and did the one thing multiplan had deliberately not done, which was to be written for the ibm pc specifically and to talk to that hardware directly instead of politely through the operating system underneath. it was fast in a way people noticed from across a room. multiplan was portable and 1-2-3 was quick, and quick won so completely that lotus spent most of the eighties as the company microsoft got measured against.

gates has been direct about what went wrong and it was not the code. multiplan was built portable because in 1982 nobody knew which machine was going to win, so the effort went into ports rather than into making the ibm pc version fast. by the time it was obvious which machine had won, 1-2-3 already owned it. multiplan sold over a million copies and was never once the thing somebody bought a computer in order to run.

~ the rap sheet ~

~ how it ended ~

microsoft's answer was not to fix multiplan. it was to write a new spreadsheet from scratch, and to write it first for a machine lotus was not on. excel shipped for the macintosh in september 1985, it was very good, and it killed multiplan on the mac long before it did any damage to lotus. excel reached windows in 1987. multiplan never did.

what was left carried on in the corner of the catalog where dos and os/2 users lived. version 4 added os/2 character mode, with a single executable holding the code for both systems. the last release was 4.2, which sources date to 1989 or to 1990 depending on which disk label you trust, and nobody announced anything about any of it. it simply stopped having a next version.

~ what survived ~

here is the part still running right now. multiplan addressed cells as r1c1: row one, column one, counted numerically in both directions. visicalc had used a1, letters across and numbers down, and lotus copied visicalc because that was what everybody already knew. by the time microsoft came to write excel it could not ship r1c1 as the default, because asking a 1-2-3 user to relearn cell addressing in order to switch is the same as asking them not to switch.

so excel does something quietly strange and has done it since 1985. it stores formulas internally in r1c1 and translates them into a1 on the way to the screen. really though, the relative-reference behavior everybody leans on when they drag a formula down a column is r1c1 logic wearing a1 clothes, which is the reason the translation layer has to be there in the first place. and r1c1 is still sitting in the options as a checkbox, forty-four years after the product that introduced it, waiting for somebody who wants it back.

@}-->--

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

« previous  ·  random  ·  next »