
Apollo Computer
~ the goodbye ~
William Poduska had already built one computer company, Prime Computer, and left it to build another. Apollo was founded in 1980 in Chelmsford, Massachusetts to sell a machine that did not really exist yet as a category: a single-user computer powerful enough for engineering work, sitting on a desk instead of in a room. The DN100 shipped in 1981 and the category got a name shortly after.
For most of the decade Apollo was the company that made them best. By the end of 1986 it held the largest share of the worldwide engineering workstation market, at roughly double the share of the company in second place, which was Sun Microsystems. Under a year later it had fallen to third, behind Digital Equipment and Sun both. Nothing broke. The market simply decided it wanted something else.
On 12 April 1989 Hewlett-Packard announced it would buy Apollo for $476 million in cash, at $13.25 a share across roughly 36 million shares. Apollo had about 100,000 workstations installed at customers when the deal was announced, which is the number that shows this was not a failing product so much as a losing position. HP wound the operation down gradually between 1990 and 1997.
I have generated an enormous number of UUIDs over the years without once wondering where the format came from. It came from here, out of the networking work Apollo did in its last years, which is a strange thing to learn about a company whose actual products were $40,000 machines I would never have been in the same room as. The hardware was gone before I could read. The identifier scheme is in code I wrote this decade.
~ the chronicle ~
| Born | 1980, Chelmsford Massachusetts, founded by William Poduska and others |
|---|---|
| Peak | end of 1986, largest worldwide share of engineering workstations at twice Sun's |
| Killed | 12 April 1989, HP acquisition announced at $476 million |
| Lifespan | 9 years as an independent company, wound down through 1997 |
| Made by | William Poduska, who had already founded Prime Computer |
| Killed by | open systems, and Sun selling cheaper boxes running Unix |
~ the breakout ~
The technical case for Apollo was Domain, which was a network architecture pretending to be a computer. Machines were linked by a token ring and the whole cluster presented a single filesystem, so a file on somebody else's workstation had a path like a file on yours, with no mounting step and no copying. The operating system, Aegis and later Domain/OS, was written around that idea from the start rather than having it bolted on.
For the customers this was genuinely superior. Engineering departments at car and aerospace companies ran CAD across a floor of Apollos and got shared work without thinking about it, at a time when networking a set of Unix boxes was a project rather than a feature. The machines were fast and the software was integrated in a way competitors could not match for years.
~ the decline ~
Every one of those advantages depended on Apollo controlling the whole stack, and that was the position the industry turned against. Sun sold Unix workstations that were cheaper and less clever, running an operating system customers could get elsewhere and code they could move to another vendor's hardware. Buyers in the late 1980s started weighing lock-in as a cost, and once they did, the integrated system stopped reading as an advantage.
Apollo responded by making Domain/OS speak Unix, adding compatibility environments so its machines could run what Sun's machines ran. That is the correct engineering answer and the wrong commercial one, because it concedes the argument. If the reason to buy your machine is that it can pretend to be the other machine, the customer has already been told which one is the standard. The market share numbers between 1986 and 1988 are the whole story in two data points.
~ the next chapter ~
The piece of Apollo that outlived Apollo is the Network Computing System, its remote procedure call framework from the late 1980s. NCS went into the Open Software Foundation's Distributed Computing Environment, which inherited its identifier scheme. Microsoft then picked up DCE RPC for Windows networking. The 128-bit universally unique identifier that database schemas hand out by the million is a direct descendant of a naming problem Apollo had to solve for its distributed filesystem.
HP brought the Apollo name back in 2014 for a line of high performance computing servers, which is trademark reuse rather than continuity. The real continuity is invisible. Almost nobody generating a UUID knows they are using the last surviving output of a Massachusetts workstation company that was the biggest in its field for about eighteen months.
~ leave a rose ~
an anonymous tribute. one rose per visitor per day.
@}-->-- ... roses left here
~ leave a wreath ~
a rose fades by morning. a wreath stays on this grave for good, with your name on it.
from $1, paid once through stripe. it keeps the graveyard running: the hosting and the nightly digging. adopt it and the headstone reads "maintained in memory by" you. the groundskeeper plants new wreaths at the next tending, usually within a day.
» one email a week: new graves and the anniversaries of the dead. subscribe
other gadgets: Microsoft Band, Myo, Symbolics, CueCat
also died in the 1980s: MacBASIC, Memotech MTX, Camputers Lynx
~ leave a note ~
visitors before you have left these graveside notes. anonymous welcome.