The Internet Encyclopedia (Volume 3)

P1: JDW

UNIX WL040/Bidgoli-Vol III-Ch-41 August 13, 2003 17:26 Char Count= 0

496 UNIXOPERATINGSYSTEM

Ritchie and Rudd Canaday had already begun sketching

out a design for a file system of some theoretical new op-

erating system that BTL was not going to allow them to

develop. This file system design was compositional and

had a root node from which descended either files or di-

rectories, which were just another type of file. Ken liked

the design and implemented the file system over a single

night on the PDP 7.

Having the PDP 7 provided a cost-free machine (some-

one else at BTL had paid the $72,000 for the computer) on

which to develop that new operating system they wanted

to work on. When Ken’s wife decided to take their new

baby boy out to California to visit Ken’s parents for a

month, Ken used that month to implement a new oper-

ating system kernel. In one month he produced a kernel,

a shell, an editor, and an assembler to natively compile

programs on the new kernel. In one incredible month, he

had produced a new multiuser multiprocessing operating

system. The moniker UNICS was given to the new operat-

ing system, reportedly by Peter Neumann (but some say

Ken Thompson), as a pun on the word Multix: UNICS was

punned the UNiplexed Information and Computing Ser-

vice, a not-so-subtle implication that it was some form of

“castrated” Multics. But it was hardly that.

There was a method to all this madness, however, be-

cause whereas Multics was a “serious” operating system,

Ken, Dennis, and Rudd wanted to keep their new system

simple.Simple meant manageable and comprehensible

and extensible. They believed that one of the things that

had led Multics astray was its complexity. The sheer small-

ness of the 18-bit PDP 7 machine forced an economy of

vision, and certainly an economy of bits, and this eco-

nomic vision produced the central philosophical tenet of

Unix:small is beautiful.

The Sting

Even though the initial version of Unix was written in PDP

7 assembly language, it was still painfully slow. They had

to have a new machine with the hardware resources ca-

pable of supporting the new multiuser and multiprocess-

ing operating system they had developed and were envi-

sioning. There was certainly a desire, again learned from

Multics, to have the operating system written in a high-

level language, for portability reasons. Current compilers

such as Fortran were ruled out immediately because they

were much too large. The GECOS machine (a GE 635)

had a compiler that had been ported to it called Basic

Combined Programming Language (BCPL). Even BCPL

was too large for the PDP 7’s fledgling new operating sys-

tem. So the team cut out all but the essentials, and turned

BCPL into an interpreter (which only generates interme-

diate code rather than machine code) and called it simply

“B” to indicate that it was a highly abbreviated form of

BCPL. It was a temporary solution, but not an ideal one,

and, as it turned out, not a permanent one. What they

really needed was a larger and more capable machine.

Knowing that the powers that were at BTL were go-

ing to be stricken with apoplexy on any proposal with the

words “operating system” in it, Joe Ossanna, forever in-

terested in text processing, reminded Ken Thompson that

the Patent Department was in need of a text processing

system, something akin to a program called “runoff” that

had been deployed on another system. Ossanna suggested

that the managers might actually accept a proposal that

provided an actual solution to an immediate need, which

was a text processor that could support specific terminals

and provide line numbered output.

The Coup

A proposal was made and eventually accepted and a new

computer, a Digital Equipment Corp. PDP 11/20, was pro-

vided for the development of the new text processing sys-

tem for BTL’s Patent Department. But an operating system

was needed on which to develop the text processing sys-

tem. What operating system did they choose to run on

this brand new computer? Unix, of course! (The UNICS

name soon morphed into the now familiar Unix form.) On

November 3, 1971, K. Thompson and D. M. Ritchie pub-

lished the firstUnix Programmer’s Manual,marking the

first official release of the Unix operating system running

on the PDP 11/20. The last sentence of the introduction

to the manual simply read: “This manual was prepared us-

ing the UNIX text editoredand the formatting program

roff.” The first released version of Unix supported three

typists from the Patent Department at Bell Telephone Lab-

oratories in Murray Hill, NJ. The roff text processing sys-

tem was written literally over a single night. The rest of

the time was spent on Unix.

The Patent Department liked the system and eventu-

ally bought the 11/20 machine from Computer Research

Group. The funds received for the 11/20 were used to-

ward the purchase of another faster and better machine:

the DEC 11/45, onto which Unix was immediately ported.

Infancy and Politics

With the publication of the third edition of Unix in early

1973, the number of Unix installations had grown to 16,

far beyond just the Patent Department. The fourth edition

appeared later that year and provided a complete rewrite

of the Unix operating system from assembly language into

C, derived from B, a highly portable systems program-

ming language that produced native machine code, de-

veloped primarily by Dennis Ritchie. This was a seminal

event in the history of Unix, because for the first time,

porting Unix onto a new hardware platform was signifi-

cantly easier compared with trying to port an assembly-

based operating system. Unix grew up within BTL like any

inventive intrigue, by oral tradition and reputation for de-

livering the goods. And news of its abilities and philosophy

spread. The fact that it actuallyworkeddidn’t hurt either.

A Slight Marketing Problem

In October of 1973, Ken Thompson and Dennis Ritchie

delivered a highly successful talk on the new Unix op-

erating system to an Association of Computing Machin-

ery conference on operating systems. The paper was titled

“The UNIX Time-Sharing System” (Ritchie & Thompson,

1978). It gave a technical summary of the key compo-

nents of the operating system. It also mentioned that the

system was from the beginning self-maintaining through

the fact that the source code was freely available to li-

censed users. This talk was met with enormous inter-

est, and requests began pouring in for the opportunity to