In June of 1998 some people at the Xerox Palo Alto Research Center (Xerox PARC) resurrected and got operational a pair of Xerox Star minicomputers and held a talk called "The Last Demo of the Xerox Star." I had the good fortune to read about it on a newsgroup and take time off from work to go. I recently found my notes from that lecture and present them here.
|Notes from "The Last Demo of the Xerox Star"
Xerox Star "Last Demo"
Xerox - "Architect of Infromation"
1973 - Alto & Ethernet. Pad-sized computer prototype.
1975 Jan "You can't extrapolate backwards and get there."
There have been discontinuities.
Interactive - Dumb terminals; "Intelligent" terminals.
Word processing - from IBM Deutschland; electromechanical.
IBM - biggest player.
Xerox - biggest threat. Diablo printer.
Software from mainframe vendors or custom for customer.
Turnkey office systems.
No personal computers yet.
Xerox had bought Scientific Data Systems.
Sigma Scientific -> Business.
1975 Xerox sold SDS to Honeywell.
But Xerox had computer experts.
Xerox had antitrust lawsuit over copiers; owned Shugart (Seagate);
big fast laser printers.
10 year system architecture for electronic copying & printing
long term for paperless office.
1975 OIS Architecture (Shipped by 1981)
PSN -> ethernet
scalable CPU instruction set - implementation independent
OS Baroque to Rococo then. VM 512kB!
Do more than just memos. More than a typewriter.
Equation typesetting would sell laser printers.
Expand Xerox brand name to peoples' desks.
1978 - personal interactive software was unknown.
1988 - Mac caught up. OS 6
1993 - Windows caught up. W3.1
1972 Alto: email, FTP, office automation, CAD. } redo it all from
16 bit 600 x 800 pixel display [tall]
170ns - 6MHz CPU
32bit - 32Mbit/sec bandwidth
CPU did disk I/O, 16 byte buffer.
1024 x 800
54 Mbits/sec memory
4 2901 CPUs National Semiconductor
1 ns instructions
300 ns memory
clock cycles were allocated to hardware i/o devices & instructions
- like CPU bus in modern machine.
8085 for llow speed I/O: leybolard, mouse, disk, clock
needed cluster processor: dandelion code name
(dandelion is a wildflower but lion is a beast)
last bipolar CPU
1981 -> CMOS
1981 Mac development started
Goal: picture on screen by Christmas 1979
accomplished by hooking scanner to memory
Machine architecture is based on financial & cultural influences
Huge: PDP/10 36 bits 2Mips 1Mword memory
too big: PDP11/60 16bit 1/2Mips 1/4MB memory
Star - office appliance, fixed function
- No GP I/O bus -> tiny I/O hardware
expand via ethernet
1/3 of power, weight, cost.
memory - 5 rounds allocated to HW
a device could access memory
guaranteed not to lose data
"Correct by construction"
7 MHz clock speed bit-sliced 4 parts 1/4 MB 8MB VM
3 Mbit ethernet only 88 chips 12" x 8" board
20 Mbit did not fit.
10 Mbit chip from Fairchild made it work.
That is why Ethernet is 10 Mbit/sec.
Disk ran synchronized with CPU
didn't work in Japan - 50Hz
1000x memory in disks and RAM - Physical
"So the software had better be a thousand times better."
38fps 800x1024; 32 lines border pattern
Could not use black border pattern to not imply [that] people were
Video ran synchronously with CPU, too.
8085 ran 8" floppy drive
keyboard controller, com ports, etc.
then: 800kB floppy
now: 100MB Zip
email, networked design environment allowed 24-hour design —
Moore's Law since 1959: CPU, RAM 1000x since the Star;
I/O has only increased 10x
Dave Smith - Human Factors engineer.
8010; 6085 v2.
"It's hard to give a demo on Star. People anymore say "What's
the big deal?"
bring computer to professionals, not professionals to computer.
World was character driven; command line.
1st visual point & click
1st UI standards
1st ethernet laser printer
but it tried to do too much.
but it was closed <- biggest mistake.
Mac toolbox was a huge improvement.
2 button mouse -
1 was too few; 3 were unnecessary.
"Mac has 2 buttons. The second one is on the keyboard."
Biggest reason for success was keyboard-
[Delete], [Copy], [Move], [Show Properties]
most important functions were keys on the left side
Font effects across the top: bold, italic, underline, …
very few commands to do complicated things
Demo uses only a dozen commands.
Powerful ideas - Pervasiveness
Copying - easter than creating from sratch. Multiple choice rather
than fill in the blank.
Command line forces you to generate a command (menu lets
you choose a command)
Star [was] networked from the start.
White bar at top is area for messages.
Network directory icon-
Winow lists categories
basic document: copy, then deposit on desktop.
Icons looked like docs, folders, file cabinets, waste basket.
Not until Mosaic was there a similar interface.
New doc - copy a blank document and edit it.
Objects have state and behavior.
Show properties shows state.
WP commands: ? Close Layout
Properties window - ¶ character
[Show Properties] key.
Remap with context.
keyboard key - top of window shows meanings of keys.
Like "Keycaps" but shows more meanings of special-function
Within graphics frame, top row of keys become graphics keys. Electronic
transfer paper copy a thing to graphics frame.
Copy bargraph, drag to resize
show properties - includes data table.
Equation frame remaps keyboard to equation symbols
[Sigma] for instance, fills in the parameters one by one with the
Mathematically typeset correct equation only using keyboard and
smart equation object.
To print or email, drag to printer.
Data icons - folders, cabinets, files
Function icons - mailboxes, Printers, File cabinets - "Move"
17 years, 1000x later.
"Star's interface is still simpler, more consistent, more
useful than today."
Fewer commands, not not fewer functions.
Generic function keys.
Objects with properties.
Smart objects with inspectors.
Other technologies developed at Xerox:
1st release had more than ascii
Japanese, 6 months later. 1983.
... this became Unicode.
Remote procedure calls.
CUSP - customer programming (consultants programmed)
Boot servers --> app servers, Java
In August, 1998 I wrote
I was at the "last" demonstration of the Xerox Star at PARC
a few weeks ago. It is an interesting computer, and was fairly advanced
for its time. I liked the way it did some things, and I wish the Mac (and
thus Windows) had done them the same way. But the Mac is not a clone of
the Star, it is a significant development on top of the Star's basic concepts.
Windows is much more similar to the Mac than the Mac is to the Star.