Computing gets personal - The Computing Universe: A Journey through a Revolution - page 146

company in exchange for free use of the company's new PDP-10 minicomputer.
There they learned new programming skills from the company's expert pro-
grammers. One of them, Steve Russell, had worked with John McCarthy at MIT
and had developed Spacewar!, one of the first interactive computer games. In
those early days, computer companies placed most value on their hardware:
software came free as an inducement for customers to buy the machine. As
a result, C-Cubed had access to the source code of the TOPS-10 operating sys-
tem, developed by DEC for the PDP-10 mainframe computer, and C-Cubed was
working to debug and enhance the system. Russell noticed Allen's interest in
learning more about programming and introduced him to the PDP-10 assembly
language. As a project, he suggested that Allen try to improve and enhance the
BASIC compiler for the PDP-10. When C-Cubed closed down, Allen and Gates
continued their projects by relocating unofficially to the computer science lab
at the University of Washington. During the next few years, they also worked
as programmers on various commercial contracts, writing code for PDP-10
machines. In the summer of 1972, they formed a partnership called Traf-O-Data
to develop both the hardware and software to automate the measurement of
traffic flows using Intel's newest microprocessor, the 8008, to do the data analy-
sis. They persuaded Paul Gilbert, an engineering student at the University of
Washington, to design and build the hardware. To write the software, because
the hardware did not yet exist, they decided to simulate the 8008's instruction
set on a PDP-10 minicomputer. Although Traf-O-Data was not a commercial suc-
cess, Allen and Gates built an unrivaled set of development tools for the 8008
microprocessor. These tools included an assembler, to translate from assembly
language into machine code; a simulator, to model and study real-life situations
on the computer; and a debugger that allowed the programmer to stop the
program in mid-execution.
By December 1974, Bill Gates had gone to Harvard, and Paul Allen had
also moved to the Boston area working as a programmer for Honeywell. When
Allen came across the January issue of Popular Electronics , the two friends real-
ized that their experience had uniquely prepared them for the challenge of
writing BASIC for the Altair. Allen describes Gates calling Ed Roberts ( B.8.2 ) in
Albuquerque, pretending to be Paul Allen:

Fig. 8.6. The interpreter source tape
for Altair BASIC. Paul Allen finished the
software while flying to Albuquerque
to demonstrate the interpreter to Ed
Roberts and his engineers at MITS.
Microsoft later created interpreters for
many other languages and processors,
although BASIC remained its most valu-
able product into the early 1980s. The
text on the tape reads “BASIC 8K without
cassette July 2 1975.”

B.8.2. Ed Roberts (1941-2010)
founded the Micro Instrumentation
and Telemetry Systems (MITS) com-
pany in 1970, and initially produced
electronics kits for model rockets,
and later, for calculators. When
calculators became too cheap for the
MITS kits to be profitable, Roberts
designed a $397 “personal com-
puter,” do-it-yourself kit called the
Altair 8800. After the January 1975
issue of Popular Electronics , orders
began to pour in and the Altair
became the catalyst for the personal
computer revolution.

“This is Paul Allen in Boston,” Bill said. “We've got a BASIC for the Altair that's
just about finished, and we'd like to come out and show it to you.” I admired
Bill's bravado but worried that he'd gone too far, since we'd yet to write the
first line of code. 8

Roberts had received many calls from people making similar claims. He told
Gates that he would give a contract to the first person to demonstrate a BASIC
that actually worked on the Altair.
With this as encouragement, Allen and Gates bought an 8080 instruction
manual and set about extending their Traf-O-Data development tools for the
new microprocessor. Gates led the design of the BASIC interpreter ( Fig. 8.6 ). A
compiler converts the entire source code of a program into an assembly lan-
guage program in one operation: an interpreter translates and executes small
pieces of source code at a time and therefore takes up much less memory. To