/LMAR=0/XLINE=3/FONT#0=BAXL30/FONT#1=BAXM30/FONT#2=BASB30/FONT#3=SUB/FONT#4=SUP/FONT#5=BASL35/FONT#6=NGR25/FONT#7=MATH30/FONT#8=FIX25/FONT#9=GRKB30
␈↓ α∧␈↓␈↓ u1


␈↓ α∧␈↓α␈↓ βLCRITERIA FOR USEFULNESS OF COMPUTERS IN OFFICES

␈↓ α∧␈↓α␈↓ εεby John McCarthy

␈↓ α∧␈↓␈↓ αTThe␈α∪thesis␈α∩of␈α∪this␈α∩lecture␈α∪will␈α∩be␈α∪that␈α∩there␈α∪is␈α∩no␈α∪difficulty␈α∩in␈α∪getting␈α∩people␈α∪to␈α∩use
␈↓ α∧␈↓computers␈α
in␈α
offices␈α
provided␈α
the␈α
computer␈α
and␈α
its␈α
applications␈α
are␈α
genuinely␈α
useful.␈α
 However,
␈↓ α∧␈↓the␈α∞criteria␈α∞for␈α∞usefulness␈α∞are␈α∞often␈α∞not␈α∂what␈α∞one␈α∞would␈α∞imagine,␈α∞and␈α∞some␈α∞further␈α∂research␈α∞is
␈↓ α∧␈↓required before the ␈↓αreal computer revolution␈↓ happens.

␈↓ α∧␈↓␈↓ αTI␈αbecame␈αinterested␈αin␈αoffice␈αuse␈αof␈αcomputers␈αin␈α1957,␈αand␈αthis␈αwas␈αone␈αof␈αthe␈αmotivations
␈↓ α∧␈↓for␈αmy␈αresearch␈αon␈αtime-sharing␈α-␈αthe␈αmain␈αone␈αbeing␈αuse␈αin␈αartificial␈αintelligence␈αresearch.␈α The
␈↓ α∧␈↓first␈α∞time-sharing␈α∞system␈α∞at␈α∞Stanford␈α∞was␈α∞a␈α∞PDP-1␈α∞in␈α∞1964.␈α∞ For␈α∞that␈α∞we␈α∞provided␈α∞no␈α∞off-line
␈↓ α∧␈↓program␈α∂preparation␈α∂equipment,␈α∂and␈α⊂when␈α∂we␈α∂specified␈α∂a␈α⊂display␈α∂system,␈α∂we␈α∂insisted␈α⊂on␈α∂both
␈↓ α∧␈↓upper␈α
and␈α
lower␈α
case␈α
with␈α
a␈α
view␈α
to␈αusing␈α
the␈α
system␈α
for␈α
preparing␈α
documents␈α
as␈α
well␈α
as␈αother
␈↓ α∧␈↓applications.

␈↓ α∧␈↓␈↓↓A. Office Computing at the Stanford Artificial Intelligence Laboratory␈↓

␈↓ α∧␈↓␈↓ αTThe␈α
Stanford␈α
Artificial␈α
Intelligence␈α
Laboratory␈α
received␈α
its␈α
PDP-6␈α
computer␈α
in␈α
1966,␈αand␈α
it
␈↓ α∧␈↓was␈αplanned␈αto␈αuse␈αthe␈αcomputer␈αfor␈αoffice␈αapplications␈αfrom␈αthe␈αbeginning.␈α All␈αdisplays␈αand␈α
the
␈↓ α∧␈↓printer␈α∂permitted␈α∞upper␈α∂and␈α∞lower␈α∂case␈α∂and␈α∞a␈α∂reasonable␈α∞set␈α∂of␈α∞mathematical␈α∂symbols,␈α∂and␈α∞we
␈↓ α∧␈↓began␈αimproving␈αon-line␈αeditors.␈α The␈αfirst␈αPhD␈αthesis␈αwritten␈αand␈αprinted␈αon␈αthe␈αcomputer␈αwas
␈↓ α∧␈↓in 1971.

␈↓ α∧␈↓␈↓ αTOur␈α⊂progress␈α⊂in␈α⊂office␈α⊂use␈α⊂of␈α⊂the␈α⊂computer␈α⊂was␈α⊂mainly␈α⊂paced␈α⊂by␈α⊂hardware␈α⊂acquisition.
␈↓ α∧␈↓While␈α∞we␈α∞could␈α
print␈α∞documents␈α∞from␈α∞the␈α
beginning,␈α∞there␈α∞was␈α
no␈α∞motivation␈α∞to␈α∞prepare␈α
them
␈↓ α∧␈↓on-line␈α
as␈αlong␈α
as␈αwe␈α
were␈αusing␈α
uppercase-only␈αteletypes␈α
for␈αterminals.␈α
 When␈αwe␈α
acquired␈αour
␈↓ α∧␈↓first␈α∩display␈α⊃system,␈α∩on-line␈α∩preparation␈α⊃of␈α∩documents␈α⊃began,␈α∩but␈α∩until␈α⊃we␈α∩installed␈α∩our␈α⊃60-
␈↓ α∧␈↓terminal␈α∂Datadisc␈α∂display␈α⊂system␈α∂in␈α∂1971,␈α⊂the␈α∂terminals␈α∂were␈α∂in␈α⊂a␈α∂terminal␈α∂room.␈α⊂ Putting␈α∂the
␈↓ α∧␈↓terminals␈α∂in␈α∞offices,␈α∂which␈α∞included␈α∂supplying␈α∞the␈α∂secretaries␈α∞with␈α∂terminals,␈α∞was␈α∂a␈α∂major␈α∞step.
␈↓ α∧␈↓Gradually␈α
more␈α
and␈α∞more␈α
Laboratory␈α
administrative␈α∞files␈α
were␈α
kept␈α∞on␈α
line,␈α
and␈α∞the␈α
secretaries
␈↓ α∧␈↓could␈αhelp␈αprepare␈αpapers.␈α However,␈αsince␈αit␈αalso␈αbecame␈αeasier␈αfor␈αresearchers␈αto␈αenter␈αand␈αedit
␈↓ α∧␈↓their␈α
own␈α
papers,␈α
there␈α
was␈α∞less␈α
typing␈α
per␈α
secretary.␈α
 People␈α∞differ␈α
in␈α
the␈α
extent␈α
to␈α∞which␈α
they
␈↓ α∧␈↓work through typists and secretaries, and an office system should provide for these differences.

␈↓ α∧␈↓␈↓ αTGetting␈αon␈αthe␈αARPA␈αnet␈αgave␈αa␈αbig␈αstimulus␈αto␈αmessage␈αsending␈αinside␈αthe␈αLab␈αas␈αwell␈α
as
␈↓ α∧␈↓over␈α∞the␈α∞net.␈α
 The␈α∞E␈α∞editor␈α
permitted␈α∞more␈α∞effective␈α
use␈α∞of␈α∞the␈α
displays.␈α∞ The␈α∞POX␈α∞and␈α
PUB
␈↓ α∧␈↓document␈α⊂compilers␈α⊂automated␈α⊂many␈α⊂editorial␈α∂aspects␈α⊂of␈α⊂document␈α⊂preparation.␈α⊂ When␈α∂Xerox
␈↓ α∧␈↓gave␈α∪us␈α∪a␈α∩Xerox␈α∪Graphics␈α∪Printer,␈α∩this␈α∪made␈α∪possible␈α∩preparing␈α∪multifont␈α∪documents␈α∩with
␈↓ α∧␈↓arbitrary character sets.

␈↓ α∧␈↓␈↓ αTMany␈α∞people␈α
work␈α∞both␈α
in␈α∞the␈α
Lab␈α∞and␈α
at␈α∞home,␈α
and␈α∞their␈α
easy␈α∞use␈α
of␈α∞office␈α
computing
␈↓ α∧␈↓requires home terminals, of which we now have a fair number.

␈↓ α∧␈↓␈↓ αTDonald␈α∂Knuth's␈α∂TEX␈α∞and␈α∂the␈α∂associated␈α∂acquisition␈α∞of␈α∂high␈α∂quality␈α∂printing␈α∞equipment
␈↓ α∧␈↓have␈α∀substantially␈α∀increased␈α∀the␈α∀documentation␈α∀use.␈α∀ Knuth's␈α∀vigorous␈α∀publicizing␈α∃of␈α∀TEX
␈↓ α∧␈↓including␈αthe␈α
book␈αhas␈αbeen␈α
at␈αleast␈α
as␈αimportant␈αas␈α
the␈αprogram␈α
itself.␈α Even␈αmathematicians␈α
are
␈↓ α∧␈↓beginning␈αto␈αuse␈αour␈αcomputers␈αfor␈αproducing␈αtheses␈αand␈αpapers.␈α They␈αhave␈αalways␈αbeen␈αamong
␈↓ α∧␈↓the␈α
slowest␈α
to␈α∞make␈α
use␈α
of␈α
computer␈α∞facilities.␈α
 This␈α
is␈α
because␈α∞mathematics␈α
is␈α
mostly␈α
done␈α∞at␈α
a
␈↓ α∧␈↓␈↓ u2


␈↓ α∧␈↓high␈α∃level␈α∀of␈α∃abstraction,␈α∀and␈α∃we␈α∀are␈α∃only␈α∀beginning␈α∃to␈α∀develop␈α∃computer␈α∃programs␈α∀that
␈↓ α∧␈↓communicate this abstractly.

␈↓ α∧␈↓␈↓ αTBesides␈αthe␈αmain␈αprograms␈αassociated␈αwith␈αoffice␈αuse,␈αmany␈αauxiliary␈αprograms␈αfor␈αlooking
␈↓ α∧␈↓up data in files and even computing have been developed.

␈↓ α∧␈↓␈↓↓B. Conclusions from Our Experience␈↓

␈↓ α∧␈↓␈↓ αT1.␈α⊃Up␈α⊃time␈α⊃of␈α⊃the␈α⊃computer␈α⊃and␈α⊃safety␈α⊃of␈α⊃its␈α⊃file␈α⊃system␈α⊃determine␈α⊃whether␈α⊃it␈α∩will␈α⊃be
␈↓ α∧␈↓accepted for office use.

␈↓ α∧␈↓␈↓ αT2.␈α⊃Secretaries␈α⊃and␈α⊃other␈α⊃clerical␈α⊃people␈α⊂can␈α⊃use␈α⊃computers␈α⊃even␈α⊃without␈α⊃much␈α⊂training.
␈↓ α∧␈↓Their␈αmotivation␈αto␈αdo␈αso␈αrequires␈αthat␈αthey␈α
have␈αterminals␈αon␈αtheir␈αdesks␈αand␈αthat␈αthe␈α
computer
␈↓ α∧␈↓be␈α
reasonably␈α
reliable.␈α Some␈α
amateur␈α
human␈αengineers␈α
imagined␈α
that␈αthey␈α
might␈α
have␈αa␈α
problem
␈↓ α∧␈↓with␈α
a␈α
keyboard␈α
in␈α
which␈α
the␈α
top␈α
row␈αof␈α
keys␈α
was␈α
displaced␈α
from␈α
standard.␈α
 Like␈α
almost␈α
all␈αof␈α
the
␈↓ α∧␈↓researchers,␈α⊃they␈α⊃never␈α⊃noticed.␈α⊃ For␈α⊃workload␈α⊃reasons,␈α⊃we␈α⊃have␈α⊃had␈α⊃to␈α⊃use␈α⊃many␈α⊃temporary
␈↓ α∧␈↓secretaries␈α∞from␈α∞manpower␈α∞agencies.␈α∞ This␈α∞has␈α∞proved␈α∞unexpectedly␈α∞easy.␈α∞ I␈α∞saw␈α∞one␈α
temporary
␈↓ α∧␈↓typing␈αat␈αa␈αterminal␈αtwenty␈αminutes␈αafter␈αher␈αarrival␈αunder␈αthe␈αsupervision␈αof␈αa␈αregular␈αsecretary
␈↓ α∧␈↓working at another terminal in the same office.

␈↓ α∧␈↓␈↓ αT3.␈αDisplay␈αterminals␈αare␈αmuch␈αbetter␈αand␈αcheaper␈αthan␈αhard␈αcopy␈αterminals.␈α The␈αlatter␈αare
␈↓ α∧␈↓noisy,␈αand␈α
waste␈αpaper␈α
gets␈αspread␈α
around.␈α Of␈α
course,␈αthere␈α
needs␈αto␈α
be␈αa␈α
good␈αaccessible␈α
printer,
␈↓ α∧␈↓but many people print only daily.

␈↓ α∧␈↓␈↓ αT4.␈α∞Having␈α∞many␈α∞display␈α∞terminals,␈α∞provided␈α
they␈α∞have␈α∞at␈α∞least␈α∞minimal␈α∞facilities,␈α∞is␈α
more
␈↓ α∧␈↓important␈α
than␈α
having␈α
a␈α∞few␈α
super␈α
terminals.␈α
 We␈α
will␈α∞shortly␈α
have␈α
a␈α
really␈α∞good␈α
experimental
␈↓ α∧␈↓test␈αof␈αthis␈αproposition,␈αbecause␈αthe␈αComputer␈αScience␈αDepartment␈αnow␈αhas␈αperhaps␈α75␈αterminals
␈↓ α∧␈↓in␈α
offices␈α
and␈α
about␈α
15␈αXerox␈α
Alto␈α
systems␈α
in␈α
terminal␈α
rooms.␈α I␈α
am␈α
betting␈α
that␈α
people␈αwho␈α
have
␈↓ α∧␈↓Datadiscs␈αin␈α
their␈αoffices␈α
will␈αuse␈α
them␈αrather␈α
than␈αtake␈α
their␈αpapers␈α
to␈αa␈α
terminal␈αroom␈αdown␈α
the
␈↓ α∧␈↓hall.

␈↓ α∧␈↓␈↓ αT5.␈α∩Reducing␈α∩the␈α∩noise␈α⊃level␈α∩is␈α∩important.␈α∩ A␈α⊃big␈α∩improvement␈α∩comes␈α∩from␈α⊃eliminating
␈↓ α∧␈↓typewriters, and it will probably prove worthwhile to develop less noisy keyboards.

␈↓ α∧␈↓␈↓ αT6.␈α⊃If␈α⊃they␈α⊃have␈α⊃a␈α⊃proper␈α⊃interactive␈α⊃style,␈α⊃many␈α⊃programs␈α⊃can␈α⊃be␈α⊃used␈α⊃without␈α⊂formal
␈↓ α∧␈↓documentation␈α-␈αwhether␈αa␈α
hard␈αcopy␈αmanual␈αor␈α
full␈αinteractive␈αdocumentation.␈α If␈α
the␈αprogram
␈↓ α∧␈↓is␈αwritten␈αin␈αa␈αstyle␈αfamiliar␈αto␈αour␈αusers␈αand␈αits␈αgeneral␈αcapabilities␈αare␈αknown,␈αmost␈αpeople␈αwill
␈↓ α∧␈↓try␈α∀to␈α∀use␈α∀it␈α∪without␈α∀reading␈α∀the␈α∀manual.␈α∀ For␈α∪example,␈α∀punctuation␈α∀of␈α∀the␈α∀arguments␈α∪of
␈↓ α∧␈↓commands␈αmust␈α
be␈αstandard:␈α
Don't␈αrequire␈α
a␈αcomma␈αin␈α
one␈αplace␈α
and␈αa␈α
semicolon␈αin␈αanother.␈α
 "?"
␈↓ α∧␈↓should always get information about the options available at the present point in the interaction.

␈↓ α∧␈↓␈↓ αT7.␈α
Programs␈αshould␈α
interact␈αat␈α
a␈αsingle␈α
level␈αas␈α
much␈αas␈α
possible.␈α Even␈α
an␈αexperienced␈α
user
␈↓ α∧␈↓often␈αgets␈αlost␈αin␈αa␈αhierarchy␈αof␈αmodes␈αand␈αsubmodes.␈α Menus␈αare␈αbad,␈αbecause␈αas␈αsoon␈αas␈αa␈αuser
␈↓ α∧␈↓gains␈αthe␈αslightest␈αexperience,␈α
he␈αhates␈αhaving␈αthe␈αscreen␈α
cluttered␈αup␈αwith␈αchanging␈αmenus.␈α
 The
␈↓ α∧␈↓information␈αprovided␈αby␈αmenus␈αand␈αquestion-and-answer␈αformats␈αcan␈αbe␈αprovided␈αby␈αletting␈αthe
␈↓ α∧␈↓user␈α⊃say␈α⊃"?"␈α⊃whenever␈α⊃he␈α∩needs␈α⊃to␈α⊃know␈α⊃what␈α⊃his␈α∩options␈α⊃are.␈α⊃ Worst␈α⊃of␈α⊃all␈α∩are␈α⊃interactive
␈↓ α∧␈↓programs␈αthat␈αclear␈αthe␈αinput␈αbuffer␈αbefore␈αaccepting␈αa␈αuser's␈αcommand␈αin␈αa␈αnew␈αsituation.␈α The
␈↓ α∧␈↓even␈α
slightly␈α
experienced␈α
user␈α
will␈α
want␈α
to␈αtype␈α
ahead,␈α
often␈α
totally␈α
ignoring␈α
what␈α
is␈α
on␈αthe␈α
screen,
␈↓ α∧␈↓in order to get the program into a desired state.
␈↓ α∧␈↓␈↓ u3


␈↓ α∧␈↓␈↓ αT8.␈α
Keeping␈αup␈α
with␈α
new␈αprograms␈α
and␈αimprovements␈α
in␈α
programs␈αhas␈α
proved␈αimpossible␈α
so
␈↓ α∧␈↓far.␈α∂ People␈α∞use␈α∂only␈α∂a␈α∞part␈α∂of␈α∂the␈α∞facilities␈α∂of␈α∞our␈α∂interactive␈α∂programs.␈α∞ This␈α∂isn't␈α∂a␈α∞tragedy;
␈↓ α∧␈↓everyone␈α⊂has␈α∂his␈α⊂own␈α⊂appropriate␈α∂balance␈α⊂of␈α∂effort␈α⊂between␈α⊂learning␈α∂about␈α⊂new␈α⊂features␈α∂and
␈↓ α∧␈↓using the old ones.

␈↓ α∧␈↓␈↓ αT9.␈α
Many␈α
jobs␈α
do␈α
not␈α
involve␈α
continuous␈α
use␈α
of␈α
the␈α
computer,␈α
and␈α
it␈α
is␈α
more␈α
important␈α
to
␈↓ α∧␈↓meet␈α
the␈α
needs␈αof␈α
the␈α
casual␈αuser␈α
than␈α
those␈αof␈α
the␈α
beginner.␈α A␈α
person␈α
will␈α
invest␈αconsiderable
␈↓ α∧␈↓effort␈αin␈αfirst␈α
learning␈αhow␈αto␈α
use␈αa␈αcomputer,␈α
but␈αif␈αhe␈αhas␈α
to␈αlearn␈αall␈α
over␈αagain␈αafter␈α
a␈αtwo-
␈↓ α∧␈↓month␈α∩layoff,␈α∪he␈α∩won't␈α∩put␈α∪in␈α∩the␈α∪effort␈α∩a␈α∩second␈α∪time.␈α∩ Computer␈α∩use␈α∪should␈α∩not␈α∪be␈α∩like
␈↓ α∧␈↓instrument flying - requiring lessons if you haven't done it for six hours in the last ninety days.

␈↓ α∧␈↓␈↓ αT10.␈αThe␈αutility␈αof␈αmany␈αproposed␈αapplications␈αof␈αcomputers␈αis␈αlimited␈αby␈αthe␈αwork␈αrequired
␈↓ α∧␈↓to␈α∪put␈α∪the␈α∀information␈α∪in␈α∪the␈α∪computer.␈α∀ The␈α∪prize␈α∪example␈α∪is␈α∀the␈α∪proposal␈α∪to␈α∀use␈α∪home
␈↓ α∧␈↓computers␈α∂to␈α∂keep␈α∂track␈α∂of␈α∂items␈α∂in␈α∂the␈α∂pantry␈α∂and␈α∂warn␈α∂the␈α∂householder␈α∂when␈α∂to␈α∂buy␈α∞more.
␈↓ α∧␈↓Even␈α
if␈αthe␈α
terminal␈αwere␈α
in␈αthe␈α
pantry,␈αit␈α
would␈αbe␈α
too␈αmuch␈α
trouble␈αand␈α
people␈α
would␈αforget.
␈↓ α∧␈↓A␈αbar␈αcode␈αreader␈αin␈αthe␈αpantry␈αmight␈αmake␈αit␈αreasonable,␈αbut␈αthe␈αgeneticists␈αmay␈αhave␈αto␈αbreed
␈↓ α∧␈↓hens that lay eggs with bar codes.

␈↓ α∧␈↓␈↓↓C. Some More Controversial Contentions␈↓

␈↓ α∧␈↓␈↓ αTI␈α∞cannot␈α
claim␈α∞that␈α
these␈α∞contentions␈α
have␈α∞been␈α
verified␈α∞by␈α
experience,␈α∞because␈α∞I␈α
haven't
␈↓ α∧␈↓enough.

␈↓ α∧␈↓␈↓ αT1.␈αExecutives␈αwill␈αuse␈α
computer␈αsystems␈αprovided␈αthey␈α
are␈αgenuinely␈αuseful,␈αbut␈α
when␈αthey
␈↓ α∧␈↓are␈αnot␈αuseful,␈αthe␈αcomplaints␈αwill␈αoften␈α
be␈αmisleading␈αabout␈αthe␈αreal␈αreasons.␈α An␈α
executive␈αwill
␈↓ α∧␈↓always␈α∂be␈α∂a␈α∂casual␈α∞user.␈α∂ Therefore,␈α∂the␈α∂terminal␈α∂must␈α∞be␈α∂unobtrusive␈α∂and␈α∂quiet;␈α∂his␈α∞secretary
␈↓ α∧␈↓must␈αhave␈αone␈αtoo,␈αand␈αif␈α
he␈αworks␈αat␈αhome,␈αthere␈αmust␈αbe␈α
a␈αterminal␈αat␈αhome␈αtoo.␈α In␈α
fact,␈αthe
␈↓ α∧␈↓message␈α∞use␈α∞of␈α
a␈α∞computer␈α∞is␈α∞most␈α
helpful␈α∞out␈α∞of␈α
normal␈α∞working␈α∞hours.␈α∞ I␈α
got␈α∞out␈α∞of␈α∞bed␈α
last
␈↓ α∧␈↓night,␈α
because␈α
I␈α
remembered␈α
a␈α
message␈α
I␈α
had␈α
been␈α
intending␈α
to␈α
send␈α
for␈α
a␈α
week,␈α
sent␈α
it␈α
and␈α
forgot
␈↓ α∧␈↓the matter till I received the reply this morning.

␈↓ α∧␈↓␈↓ αTUsefulness␈α∞for␈α∞executives␈α∂will␈α∞depend␈α∞on␈α∂how␈α∞many␈α∞of␈α∞the␈α∂people␈α∞with␈α∞whom␈α∂they␈α∞must
␈↓ α∧␈↓communicate also have terminals.

␈↓ α∧␈↓␈↓ αTAnyone who does much work at home should have a terminal at home.

␈↓ α∧␈↓␈↓ αT2.␈α∞Idiot-proofing␈α∞programs␈α∞is␈α∞often␈α∞a␈α∞bad␈α∞idea.␈α∞ It␈α∞is␈α∞easy␈α∞for␈α∞the␈α∞designer␈α∞of␈α∞an␈α∞on-line
␈↓ α∧␈↓system␈αto␈α
get␈αinto␈α
a␈αstate␈αof␈α
mind␈αwhere␈α
he␈αregards␈α
the␈αuser␈αas␈α
an␈αidiot␈α
who␈αmust␈α
be␈αprevented
␈↓ α∧␈↓from␈α
making␈α
all␈α
kinds␈α
of␈α
mistakes.␈α
 Indeed␈α
the␈α
books␈α
and␈α
papers␈α
on␈α
interactive␈α
programming␈α
take
␈↓ α∧␈↓this␈α⊂attitude.␈α⊂ In␈α∂fact,␈α⊂the␈α⊂people␈α∂who␈α⊂write␈α⊂about␈α∂supervising␈α⊂programmers␈α⊂take␈α⊂that␈α∂attitude
␈↓ α∧␈↓towards their charges.  However, it has several disadvantages.

␈↓ α∧␈↓␈↓ αTFirst␈α∞it␈α∞must␈α
concentrate␈α∞on␈α∞the␈α∞kinds␈α
of␈α∞mistakes␈α∞that␈α∞can␈α
be␈α∞detected␈α∞and␈α∞prevented␈α
by
␈↓ α∧␈↓bureaucracy␈α∂-␈α∂whether␈α∂it␈α∂be␈α∂the␈α⊂programmed␈α∂bureaucracy␈α∂of␈α∂a␈α∂field␈α∂that␈α∂allows␈α⊂only␈α∂numeric
␈↓ α∧␈↓input␈α∂in␈α∂a␈α∂certain␈α∂range␈α∂or␈α∂the␈α∂administrative␈α∂bureaucracy␈α∂that␈α∂requires␈α∂a␈α∂comment␈α⊂for␈α∂every
␈↓ α∧␈↓statement␈α⊂in␈α⊂a␈α⊂program.␈α⊂ There␈α⊂are␈α⊂many␈α⊂situation␈α⊂in␈α⊂which␈α⊂the␈α⊂bureaucracy␈α⊂spends␈α⊂its␈α∂time
␈↓ α∧␈↓preventing␈α
trivial␈α
errors,␈αwhile␈α
major␈α
substantive␈α
errors␈αare␈α
ignored,␈α
because␈α
the␈αinput␈α
embodying
␈↓ α∧␈↓them is "grammatical" according to the lights of the system.
␈↓ α∧␈↓␈↓ u4


␈↓ α∧␈↓␈↓ αTSecond,␈α
idiot-proofing␈α
takes␈α∞time,␈α
and␈α
it␈α
often␈α∞happens␈α
that␈α
the␈α
idiot-proof␈α∞programs␈α
are
␈↓ α∧␈↓insufficiently␈αdebugged.␈α There␈αis␈αnothing␈αas␈αannoying␈αas␈αtrying␈αto␈αget␈αa␈αprogram␈αto␈αaccept␈αinput
␈↓ α∧␈↓that it is rejecting for trivial reasons.

␈↓ α∧␈↓␈↓ αTThird,␈αidiot-proof␈αprograms␈αare␈αusually␈α
extremely␈αinflexible␈αand␈αare␈αdifficult␈αto␈α
modify␈αto
␈↓ α∧␈↓take new data into account.

␈↓ α∧␈↓␈↓ αTLet␈α∀me␈α∀describe␈α∀an␈α∀experiment␈α∀that␈α∀unfortunately␈α∀was␈α∀never␈α∀carried␈α∀out.␈α∀ A␈α∪certain
␈↓ α∧␈↓university␈α∪found␈α∩its␈α∪on-line␈α∪registration␈α∩system␈α∪terribly␈α∩late,␈α∪full␈α∪of␈α∩bugs,␈α∪and␈α∪expensive␈α∩of
␈↓ α∧␈↓computer␈α⊂resources.␈α⊂ The␈α⊃experiment␈α⊂was␈α⊂to␈α⊂have␈α⊃the␈α⊂clerks␈α⊂prepare␈α⊂the␈α⊃registration␈α⊂material
␈↓ α∧␈↓using␈αan␈αordinary␈α
editor␈α-␈αlabelling␈αthe␈α
items␈αin␈αthe␈α
text␈αof␈αthe␈αrecord.␈α
 The␈αfiles␈αprepared␈αby␈α
the
␈↓ α∧␈↓clerks␈αwould␈αthen␈αbe␈αprocessed␈αby␈αprograms␈αto␈αget␈αit␈αin␈αthe␈αdesired␈αform.␈α Unlike␈αon-line␈αinput-
␈↓ α∧␈↓receiving␈α∞programs,␈α∂the␈α∞processing␈α∂program␈α∞could␈α∂be␈α∞written␈α∂while␈α∞the␈α∂input␈α∞was␈α∂taking␈α∞place,
␈↓ α∧␈↓and␈α
if␈α
bugs␈α
showed␈α
up,␈αthey␈α
could␈α
be␈α
corrected␈α
after␈αthe␈α
fact.␈α
 Even␈α
last-minute␈α
changes␈α
in␈αthe
␈↓ α∧␈↓information␈α⊃to␈α⊃be␈α∩included␈α⊃could␈α⊃be␈α∩accommodated.␈α⊃ The␈α⊃results␈α∩of␈α⊃the␈α⊃data-entry␈α∩could␈α⊃be
␈↓ α∧␈↓printed␈αand␈αchecked␈α
by␈αsupervisors␈αor␈αthe␈α
supervisors␈αcould␈αexamine␈α
them␈αon␈αline.␈α However,␈α
the
␈↓ α∧␈↓university took the "safer" path of buying another computer.

␈↓ α∧␈↓␈↓ αTUnfortunately,␈αthe␈αtask␈αof␈αwriting␈αa␈αcomputer␈αprogram␈αfor␈αothers␈αto␈αuse␈αseems␈αto␈αbring␈αout
␈↓ α∧␈↓the latent tyrant in many people.

␈↓ α∧␈↓␈↓↓D. A Step Further Out␈↓

␈↓ α∧␈↓␈↓ αTThere␈α∂are␈α⊂many␈α∂opportunities␈α∂for␈α⊂expanding␈α∂the␈α∂usefulness␈α⊂of␈α∂computers␈α∂in␈α⊂offices,␈α∂but
␈↓ α∧␈↓many of them require the development of standardized facilities.

␈↓ α∧␈↓␈↓ αT1.␈α⊃The␈α⊃Dialnet␈α⊃project.␈α⊃ Many␈α∩rival␈α⊃networks␈α⊃for␈α⊃interconnecting␈α⊃computers␈α∩have␈α⊃been
␈↓ α∧␈↓developed,␈α
but␈α
in␈α
my␈α
opinion,␈α
the␈α
possibilities␈αof␈α
the␈α
ordinary␈α
dial␈α
telephone␈α
network␈α
have␈αnot␈α
yet
␈↓ α∧␈↓been␈α
fully␈α
exploited.␈α
 That␈α
network␈α
has␈α
the␈α
advantage␈α
that␈α
it␈α
already␈α
connects␈α
all␈α
the␈α
offices␈α
in
␈↓ α∧␈↓the world.

␈↓ α∧␈↓␈↓ αTThe␈α⊃Stanford␈α⊃Artificial␈α⊃Intelligence␈α∩Laboratory␈α⊃is␈α⊃developing␈α⊃the␈α⊃Dialnet␈α∩system.␈α⊃ This
␈↓ α∧␈↓consists␈α⊃of␈α∩a␈α⊃telephone␈α∩dialer␈α⊃and␈α∩suitable␈α⊃modems␈α∩connected␈α⊃to␈α∩our␈α⊃computer␈α∩and␈α⊃software
␈↓ α∧␈↓implementing␈α
the␈α
Dialnet␈α∞protocols.␈α
 Anyone␈α
else␈α∞in␈α
the␈α
world␈α∞can␈α
similarly␈α
equip␈α∞his␈α
computer
␈↓ α∧␈↓and users of any computer equipped with Dialnet can communicate with users of any other.

␈↓ α∧␈↓␈↓ αTSitting␈αat␈αmy␈αterminal␈αI␈αwill␈αbe␈αable␈αto␈αtype␈α"MAIL␈αMIKESMITH@202-666-6666␈αMike␈αare
␈↓ α∧␈↓you␈α⊃free␈α⊃for␈α⊃lunch␈α⊃on␈α⊃Thursday?"␈α⊃Once␈α⊃I␈α∩have␈α⊃done␈α⊃this,␈α⊃I␈α⊃can␈α⊃use␈α⊃my␈α⊃terminal␈α∩for␈α⊃other
␈↓ α∧␈↓purposes.␈α
 My␈αcomputer␈α
calls␈α
a␈αcomputer␈α
at␈αthat␈α
number␈α
and␈αtells␈α
it␈αthat␈α
it␈α
has␈αa␈α
message␈α
for␈αa
␈↓ α∧␈↓user␈αcalled␈α
MIKESMITH.␈α He␈α
gets␈αthe␈α
message␈αimmediately␈α
if␈αhe␈α
is␈αlogged␈α
in␈α-␈α
later␈αotherwise.
␈↓ α∧␈↓We␈αcan␈αdo␈αthis␈α
now␈αfor␈αcomputers␈αon␈α
the␈αARPAnet,␈αbut␈αwhy␈α
go␈αthrough␈αall␈αthat␈α
politics,␈αwhen
␈↓ α∧␈↓the␈α∀telephone␈α∀system␈α∪is␈α∀available?␈α∀ Dialnet␈α∪can␈α∀also␈α∀be␈α∪used␈α∀for␈α∀transferring␈α∀files␈α∪between
␈↓ α∧␈↓computers.

␈↓ α∧␈↓␈↓ αTThe␈α1200␈αbaud␈αlimitation␈αof␈αpresent␈αDialnet␈αis␈αimportant␈αfor␈αsome␈αapplications␈αbut␈αnot␈αfor
␈↓ α∧␈↓messages␈α⊃and␈α⊃transfer␈α⊂of␈α⊃medium-size␈α⊃files␈α⊂like␈α⊃reports.␈α⊃ If␈α⊂one␈α⊃specificies␈α⊃NIGHTMAIL␈α⊂the
␈↓ α∧␈↓telephone cost for a 9000-byte message will be only a little more than the price of a stamp.

␈↓ α∧␈↓␈↓ αT2.␈α∂National␈α∂file-naming␈α∂system.␈α∂ A␈α∂major␈α∂application␈α∂of␈α∂Dialnet␈α∂or␈α⊂other␈α∂inter-computer
␈↓ α∧␈↓␈↓ u5


␈↓ α∧␈↓communication␈αsystems␈αwill␈αbe␈αto␈αtransfer␈αfiles␈αfrom␈αone␈αcomputer␈αto␈αanother.␈α This␈αis␈αdone␈αnow
␈↓ α∧␈↓but␈α∀it␈α∪almost␈α∀always␈α∪involves␈α∀specific␈α∀technical␈α∪arrangements␈α∀between␈α∪the␈α∀managers␈α∀of␈α∪the
␈↓ α∧␈↓computers.␈α∞ In␈α∞order␈α∞to␈α∞transfer␈α∞files␈α∞freely␈α∞(except␈α∞as␈α∞restricted␈α∞by␈α∞password␈α∞fences),␈α∞a␈α
national
␈↓ α∧␈↓file-naming system is required.

␈↓ α∧␈↓␈↓ αT3.␈α∀Describing␈α∀other␈α∪people's␈α∀files.␈α∀ Many␈α∪programming␈α∀languages␈α∀contain␈α∀features␈α∪for
␈↓ α∧␈↓describing␈αdata␈αstructures␈αso␈αthat␈αthe␈αcompiler␈αwill␈αgenerate␈αthem␈αand␈αcompile␈αprograms␈αthat␈αuse
␈↓ α∧␈↓them.␈α However,␈αno␈αone␈αsuch␈α
system␈αwill␈αconquer␈αthe␈αworld␈α
and␈αindeed,␈αif␈αprogress␈αis␈αto␈α
continue,
␈↓ α∧␈↓it␈α∞is␈α∞not␈α∞even␈α∞desirable␈α∂that␈α∞a␈α∞single␈α∞system␈α∞be␈α∂adopted.␈α∞ Therefore␈α∞people␈α∞will␈α∞always␈α∂want␈α∞to
␈↓ α∧␈↓refer to other people's data structures.

␈↓ α∧␈↓␈↓ αTThis␈αcan␈αbe␈αmade␈αpossible␈αby␈αa␈αuniversal␈αsystem␈αfor␈αdescribing␈αexisting␈αfiles␈αwhich␈αcan␈αbe
␈↓ α∧␈↓developed using the techniques for describing grammars and data structures.

␈↓ α∧␈↓␈↓ αT4.␈α∪A␈α∪standardized␈α∀style␈α∪of␈α∪interactive␈α∪programming␈α∀will␈α∪help␈α∪people␈α∪use␈α∀each␈α∪other's
␈↓ α∧␈↓programs.