/FONT#0=BASL30/FONT#1=BASI30/FONT#2=NGR40/FONT#3=NGR25/FONT#4=NGR20/FONT#15=STA200
␈↓␈↓↓␈↓α␈↓β␈↓∧␈↓∂␈β↓↓␈↓∂S␈↓α
␈↓ ↓H
␈β↓Q␈↓ ¬	STANFORD UNIVERSITY
␈↓ ↓H
␈↓β␈↓ ¬_STANFORD, CALIFORNIA 94305
␈↓ ↓H
␈↓∧
␈↓ ↓H

␈↓ ↓H

␈↓ ↓H

␈↓ ↓H
ARTIFICIAL INTELLIGENCE LABORATORY␈↓ 
+   Telephone:
␈↓ ↓H
COMPUTER SCIENCE DEPARTMENT␈↓ 
+415-497-4971
␈↓ ↓H
␈↓␈↓ ¬qRalph E. Gorin
␈↓ ↓H
␈↓ ¬}June 17, 1974
␈↓ ↓H

␈↓ ↓H
␈↓β␈↓ ↓HThe␈αviews␈αand␈αconclusions␈αexpressed␈αherein␈αare␈αthose␈αof␈αthe␈αauthor␈αand␈αdo␈αnot␈αnecessarily␈αrepresent
␈↓ ↓H
␈↓ ↓Hthe␈αviews␈αof␈αthe␈αArtificial␈αIntelligence␈αLaboratory␈αor␈αany␈αfunding␈αagency.␈α
This␈α
document␈α
is␈α
intended␈α
for
␈↓ ↓H
␈↓ ↓Hlocal␈α
circulation␈α
only.
␈↓ ↓H

␈↓ ↓H
␈↓␈↓ α_We␈α∞have␈α∞gotten␈α∞by␈α
with␈α
our␈α
present␈α
system␈α
since␈α
1968,␈α
but␈α
we␈α
are␈α
now␈α
faced␈α
with␈α
a␈α
major
␈↓ ↓H
␈↓ ↓Hhardware␈αchange␈α(i.e.,␈αpaging)␈α
that␈α
cannot␈α
be␈α
trivially␈α
added␈α
to␈α
the␈α
present␈α
system.␈α
There␈α
are␈α
several
␈↓ ↓H
␈↓ ↓Hother␈αfeatures␈αwhich␈αappear␈αto␈αbe␈αdesirable␈α
additions,␈α
but␈α
which␈α
would␈α
be␈α
difficult␈α
to␈α
include␈α
in␈α
the
␈↓ ↓H
␈↓ ↓Hpresent␈α∞system.␈α
Finally,␈α
the␈α
present␈α
system␈α
includes␈α
a␈α
number␈α
of␈α
features␈α
that␈α
are␈α
continuing␈α
sinks
␈↓ ↓H
␈↓ ↓Hfor␈αmanpower␈αand␈αsources␈αof␈αsystem␈αbugs,␈αe.g.,␈αprocess␈α
interrupts␈α
which␈α
are␈α
now␈α
very␈α
important,␈α
but
␈↓ ↓H
␈↓ ↓Hwhose␈α
implementation␈α
is␈α
irrepairably␈α
flawed.
␈↓ ↓H

␈↓ ↓H
␈↓α␈↓ ¬IThe Alternatives␈↓
␈↓ ↓H

␈↓ ↓H
␈↓ α_There are four choices open to us:
␈↓ ↓H
␈↓ α_1. postpone any decision
␈↓ ↓H
␈↓ α_2. modify the present system to add the features we desire
␈↓ ↓H
␈↓ α_3. Adapt some other existing system to our needs
␈↓ ↓H
␈↓ α_4. Create a new system suited to the needs we now perceive.
␈↓ ↓H

␈↓ ↓H
␈↓ α_Alternative␈α⊂1␈α⊂can␈α⊂be␈α⊂applied␈α⊂indefinitely;␈α⊂ongoing␈α⊂modification␈α⊂to␈α⊂the␈α⊂existing␈α⊂system␈α⊂will
␈↓ ↓H
␈↓ ↓Hcontinue␈α∞to␈α∞happen.␈α∞However,␈α
until␈α
a␈α
decision␈α
is␈α
reached,␈α
selecting␈α
one␈α
of␈α
the␈α
other␈α
alternatives,␈α
I
␈↓ ↓H
␈↓ ↓Hwill␈α
resist␈α
making␈α
any␈α
major␈α
modifications.
␈↓ ↓H

␈↓ ↓H
␈↓ α_Alternative␈α∂2␈α∂is␈α∂fairly␈α∂attractive;␈α∂we␈α∂already␈α∂know␈α∂quite␈α∂a␈α∂lot␈α∞about␈α∞the␈α∞existing␈α∞system,␈α∞so
␈↓ ↓H
␈↓ ↓Hthere's␈αno␈αtransient␈αwhile␈αwe␈αlearn␈αabout␈αit.␈αAdditions␈αand␈αmodifications␈αcan␈αbe␈αmade␈αgradually,␈αso
␈↓ ↓H
␈↓ ↓Hthere's␈α
no␈α
problem␈α
of␈α
converting␈α
all␈α
the␈α
existing␈α
programs.
␈↓ ↓H

␈↓ ↓H
␈↓ α_The␈α∞problem␈α∞with␈α∞alternative␈α∞2␈α∞is␈α∞that␈α∞the␈α∞structure␈α∞and␈α∞organization␈α
of␈α
the␈α
system␈α
doesn't
␈↓ ↓H
␈↓ ↓Hsuit␈αnew␈αdevelopments;␈αconfining␈αourselves␈αwithin␈αthe␈αexisting␈αstructure␈αmay␈αlead␈αto␈α
a␈α
system␈α
that␈α
is
␈↓ ↓H
␈↓ ↓Hhard␈α
to␈α
maintain,␈α
hard␈α
to␈α
understand,␈αunreliable,␈αand␈αdifficult␈αto␈αmodify.␈αThe␈αexisting␈αstructure␈αis
␈↓ ↓H
␈↓ ↓Hinadequate␈α⊃for␈α⊃the␈α⊃development␈α⊃of␈α⊃process␈α⊃interrupts.␈α⊃The␈α⊃addition␈α⊂of␈α⊂paging␈α⊂would␈α⊂require␈α⊂a
␈↓ ↓H
␈↓ ↓Hsubstantial␈α
rewrite␈α
of␈α
most␈α
of␈α
the␈α
I/O␈α
service␈α
routines.
␈↓ ↓H

␈↓ ↓H
␈↓ α_Alternative␈α∀3␈α∀would␈α∀be␈α∀attractive␈α∪if␈α∪we␈α∪could␈α∪find␈α∪some␈α∪system␈α∪that␈α∪was␈α∪a␈α∪substantial
␈↓ ↓H
␈↓ ↓Himprovement␈α→over␈α→the␈α→existing␈α→system.␈α→There␈α→are␈α→three␈α→other␈α→systems␈α_for␈α_PDP-10s,␈α_ITS,
␈↓ ↓H
␈↓ ↓HTENEX/VIROS,␈α
and␈α
TOPS-10.
␈↓ ↓H

␈↓ ↓H
␈↓ α_TOPS-10␈α∃has␈α∃paging␈α∃(for␈α∃the␈α∃KI-10),␈α∃and␈α∃allows␈α∀virtual␈α∀memory␈α∀capability.␈α∀However,
␈↓ ↓H
␈↓ ↓HTOPS-10␈αhas␈αessentially␈αthe␈αsame␈αkind␈αof␈αstructure␈αas␈α
our␈α
system,␈α
which␈α
means␈α
it␈α
doesn't␈α
do␈α
process
␈↓ ↓H
␈↓ ↓Hinterrupts␈α∞right.␈α∞TOPS-10␈α∞does␈α∞have␈α∞some␈α∞interesting␈α∞file␈α∞system␈α∞features␈α∞and␈α∞it␈α∞is␈α∞thought␈α
to␈α
be
␈↓ ↓H
␈↓ ↓Hfairly␈α
efficient,␈α
but␈α
it␈α
lacks␈α
any␈α
kind␈α
of␈α
display␈α
service.
␈↓ ↓H

␈↓ ↓H
␈↓ α_ITS␈αrequires␈αspecial␈α
hardware␈α
(SC␈α
pager)␈α
which␈α
I␈α
presume␈α
we'd␈α
be␈α
reluctant␈α
to␈α
buy.␈α
Also,␈α
ITS
␈↓ ↓H
␈↓ ↓His␈αessentially␈αuntested␈αon␈αthe␈αkinds␈αof␈αloads␈αwe␈αhave␈αhere.␈αITS␈αhas␈αmany␈αgood␈αideas␈αin␈αit␈αwhich␈αwe
␈↓ ↓H
␈↓ ↓Hshould␈α
adapt␈α
to␈α
our␈α
needs,␈α
but␈α
I␈α
don't␈α
think␈α
it␈α
is␈α
practical␈α
to␈α
consider␈α
actually␈α
running␈α
ITS␈α
here.
␈↓ ↓H

␈↓ ↓H
␈↓ α_TENEX␈α
offers␈αus␈αessentially␈αtwo␈αadvantages,␈αpaging,␈αand␈αcompatibility␈αwith␈αother␈αsites.␈αAlso
␈↓ ↓H
␈↓ ↓Hit␈α
appears␈α
that␈α
ARPA␈α
is␈α
interested␈α
in␈α
the␈α
continuing␈α
development␈α
of␈α
TENEX.
␈↓ ↓H

␈↓ ↓H
␈↓ α_TENEX␈αhas␈αits␈αshare␈αof␈αproblems.␈α
It␈α
is␈α
alleged␈α
to␈α
behave␈α
very␈α
poorly␈α
under␈α
what␈α
we␈α
consider
␈↓ ↓H
␈↓ ↓Hto␈αbe␈αmoderate␈αloading.␈αTENEX␈αhas␈αprocess␈αinterrupts,␈αbut␈αthat␈α
implementation␈α
suffers␈α
in␈α
the␈α
same
␈↓ ↓H
␈↓ ↓Hway␈α∞as␈α∞ours␈α∞(i.e.,␈α∞JSYSs␈α∞are␈α∞not␈α∞interruptable).␈α∞TENEX␈α
uses␈α
fixed␈α
allocation␈α
for␈α
the␈α
number␈α
and
␈↓ ↓H
␈↓ ↓Hsize␈αof␈αuser␈αfile␈αdirectories.␈αTENEX␈αfails␈αto␈αstore␈αsmall␈αfiles␈α
efficiently.␈α
(Our␈α
system␈α
loses␈α
even␈α
worse
␈↓ ↓H
␈↓ ↓Hthan␈αTENEX,␈αbut␈αI␈αscarcely␈αconsider␈αthat␈αto␈αbe␈αan␈αargument␈αin␈αfavor␈αof␈αTENEX).␈αNearly␈αhalf␈αof
␈↓ ↓H
␈↓ ↓Hour␈αfiles␈α(9500␈αof␈α19800␈αfiles)␈αcan␈αbe␈αstored␈αin␈αone␈αpage␈α(512␈αwords)␈αbut␈αin␈αTENEX,␈αthe␈αminimum
␈↓ ↓H
␈↓ ↓Hfile␈α
is␈α
two␈α
pages.
␈↓ ↓H

␈↓ ↓H
␈↓ α_TENEX␈α
doesn't␈α
have␈α
any␈α
facility␈α
for␈α
dynamic␈α
allocation␈αof␈αsystem␈αstorage.␈αSuch␈αa␈αfacility␈αis
␈↓ ↓H
␈↓ ↓Hnecessary␈α
for␈α
the␈α
implementation␈α
of␈α
display␈α
service.␈α
TENEX␈α
lacks␈αany␈αrealtime␈αfacility.␈αThe␈αJSYS
␈↓ ↓H
␈↓ ↓Hinstruction␈α⊃is␈α⊃unaesthetic;␈α⊃the␈α⊃idea␈α⊃of␈α⊃using␈α⊃fixed␈α⊃accumulators␈α⊃to␈α⊃hold␈α⊃arguments␈α⊂is␈α⊂a␈α⊂relic␈α⊂of
␈↓ ↓H
␈↓ ↓Hprogrammers␈α⊗who␈α⊗were␈α⊗familiar␈α⊗with␈α⊗the␈α⊗940␈α⊗and␈α⊗didn't␈α⊗understand␈α⊗what␈α⊗to␈α⊗do␈α⊗with␈α∃16
␈↓ ↓H
␈↓ ↓Haccumulators.␈αJSYS␈αhas␈αprecisely␈αone␈αbenefit,␈αnamely,␈αit␈αdecodes␈αthe␈αdispatch␈αaddress.␈α
But,␈α
decoding
␈↓ ↓H
␈↓ ↓Hthe␈αUUO␈αis␈αusually␈αtrivial␈αwith␈αrespect␈αto␈αthe␈αamount␈αof␈α
time␈α
used␈α
in␈α
processing␈α
it.␈α
JSYS␈α
wastes␈α
the
␈↓ ↓H
␈↓ ↓Hentire␈α
AC␈α
field␈α
(they␈α
say␈α
it's␈α
reserved␈α
for␈α
future␈α
expansion)␈α
and␈α
provides␈α
only␈α
one␈α
kind␈αof␈αsystem
␈↓ ↓H
␈↓ ↓Hcall␈α(with␈αUUOs␈αyou␈αget␈αto␈αdecode␈αthe␈αopcode,␈αthe␈αac␈αand␈αthe␈αaddress␈αindependently,␈αdepending␈αon
␈↓ ↓H
␈↓ ↓Hwhat␈αyou␈αwant).␈αJSYS␈αsaves␈αaccumulators␈α-␈αbut␈αthe␈αKL10␈αwill␈αhave␈αmultiple␈αaccumulators␈αsets␈α(like
␈↓ ↓H
␈↓ ↓Hthe␈α
KI10)␈α
so␈α
that's␈α
irrelevant.␈α
JSYS␈α
supposedly␈αsaves␈αtime␈αtoo,␈αbut␈αyou'd␈αhardly␈αnotice␈αconsidering
␈↓ ↓H
␈↓ ↓Hthe␈α
rest␈α
of␈α
TENEX.
␈↓ ↓H

␈↓ ↓H
␈↓ α_TENEX␈α⊂is␈α⊂unreliable.␈α⊂They␈α⊂crash␈α⊂at␈α⊂least␈α⊂as␈α⊂often␈α⊂as␈α⊂we␈α∂do,␈α∂and␈α∂they␈α∂stay␈α∂down␈α∂longer.
␈↓ ↓H
␈↓ ↓HTENEX␈α
also␈α
seems␈α
particularly␈α
adept␈α
at␈α
wiping␈α
out␈α
the␈α
disk.
␈↓ ↓H

␈↓ ↓H
␈↓ α_Compatibility␈α∞in␈α∞TENEX␈α∞is␈α∞a␈α∞farce.␈α
Everyone␈α
on␈α
the␈α
net␈α
is␈α
running␈α
different␈α
systems,␈α
with
␈↓ ↓H
␈↓ ↓Hdifferent␈α⊃versions␈α⊃of␈α⊃the␈α⊃cusps.␈α⊃One␈α⊃the␈α⊃major␈α⊃arguments␈α⊃that␈α⊂went␈α⊂on␈α⊂at␈α⊂the␈α⊂ARPA-TENEX
␈↓ ↓H
␈↓ ↓Hmeeting␈α⊂was␈α⊂about␈α⊂compatibility,␈α⊂where␈α∂Fields␈α∂was␈α∂attempting␈α∂to␈α∂fit␈α∂everyone␈α∂into␈α∂his␈α∂particular
␈↓ ↓H
␈↓ ↓HProcrustean␈αbed,␈αand␈αeveryone␈αelse␈α(except␈αISI␈αand␈αBBN)␈αwas␈αresisting.␈αGetting␈αinto␈αthat␈αgame␈αis␈αa
␈↓ ↓H
␈↓ ↓Hsure␈α⊃way␈α⊃to␈α⊃lose␈α⊂control␈α⊂over␈α⊂what␈α⊂happens␈α⊂here.␈α⊂(If␈α⊂you␈α⊂think␈α⊂BBN␈α⊂or␈α⊂ISI␈α⊂does␈α⊂a␈α⊂better␈α⊂job
␈↓ ↓H
␈↓ ↓Hrunning␈α
their␈α
machines␈α
than␈α
we␈α
do,␈α
you␈α
ought␈α
to␈α
abdicate␈α
the␈α
management␈α
of␈α
your␈α
system␈α
to␈α
them).
␈↓ ↓H

␈↓ ↓H
␈↓ α_If␈α∂we␈α∂adapt␈α∂TENEX␈α∂to␈α∂our␈α∂needs,␈α∂we'll␈α∂have␈α∞to␈α∞include␈α∞the␈α∞following␈α∞features␈α∞and␈α∞device
␈↓ ↓H
␈↓ ↓Hservices␈α∩(quite␈α∩a␈α∩lot␈α∩of␈α⊃work!):␈α⊃DD␈α⊃and␈α⊃III␈α⊃displays;␈α⊃Keyboard␈α⊃Scanner;␈α⊃old␈α⊃630␈α⊃scanner;␈α⊃IMP
␈↓ ↓H
␈↓ ↓Hinterface;␈α
10-11␈α
interface;␈α
IOP/Librascope;␈α
3330␈α
drivers;␈α
XGP.
␈↓ ↓H

␈↓ ↓H
␈↓ α_The␈αarguments␈αfor␈αand␈αagainst␈αVIROS␈αare␈αsimilar.␈αExcept,␈αDEC␈αwon't␈αgive␈αus␈αVIROS,␈αso␈αif
␈↓ ↓H
␈↓ ↓Hwe␈αwant␈αto␈αgo␈αthat␈αroute,␈αwe␈αcan␈αsit␈αon␈αour␈αthumbs␈αfor␈αthe␈αnext␈αyear␈αor␈αso.␈αIt␈αis␈α
hoped␈α
that␈α
VIROS
␈↓ ↓H
␈↓ ↓Hwill␈αat␈α
least␈α
be␈α
able␈α
to␈α
run␈α
jobs␈α
efficiently.␈α
Since␈α
we've␈α
described␈α
the␈α
kind␈α
of␈α
display␈α
service␈α
and␈α
disk
␈↓ ↓H
␈↓ ↓Hservice␈αwe␈αwant,␈αmaybe␈αVIROS␈αwill␈αcontain␈αthem.␈αHowever,␈αI'd␈αhate␈αto␈α
wait␈α
a␈α
year␈α
for␈α
VIROS␈α
and
␈↓ ↓H
␈↓ ↓Hthen␈α
discover␈α
that␈α
it␈α
needed␈α
another␈α
year's␈α
work␈α
before␈α
it␈α
would␈α
do␈α
what␈α
we␈α
want.
␈↓ ↓H

␈↓ ↓H
␈↓ α_The␈α∂final␈α∂alternative␈α∂is␈α∂to␈α∂write␈α∂a␈α∂new␈α∂system,␈α∞designed␈α∞to␈α∞meet␈α∞the␈α∞specific␈α∞needs␈α∞that␈α∞we
␈↓ ↓H
␈↓ ↓Hforsee,␈α
and␈α
designed␈α
to␈α
be␈α
easily␈α
adapted␈α
to␈α
future␈α
needs␈α
that␈α
are␈α
not␈α
yet␈α
forseen.
␈↓ ↓H

␈↓ ↓H
␈↓ α_The␈α∂major␈α∂disadvantages␈α∂of␈α∂this␈α∂approach␈α∂are␈α∂that␈α∂it␈α∂costs␈α∂money,␈α∞and␈α∞it␈α∞will␈α∞produce␈α∞an
␈↓ ↓H
␈↓ ↓Hincompatible␈αsystem.␈αThe␈αadvantage␈αis␈αthat␈αthe␈αnew␈αsystem␈αwill␈α
more␈α
closely␈α
suit␈α
our␈α
needs␈α
than␈α
any
␈↓ ↓H
␈↓ ↓Hother␈α
system␈α
that␈α
is␈α
available.
␈↓ ↓H

␈↓ ↓H
␈↓ α_The␈α∀new␈α∪system␈α∪will␈α∪be␈α∪able␈α∪to␈α∪"reasonably"␈α∪run␈α∪current␈α∪TENEX␈α∪programs,␈α∪TOPS-10
␈↓ ↓H
␈↓ ↓Hprograms␈α
and␈α
old␈α
Stanford␈α
programs.
␈↓ ↓H

␈↓ ↓H
␈↓ α_Since␈αvery␈αlittle␈αdesign␈αwork␈αhas␈αgone␈αinto␈αthe␈αnew␈αsystem,␈αI␈αcannot␈αestimate␈αwhat␈αit␈αwill␈αcost
␈↓ ↓H
␈↓ ↓Hin␈αterms␈αof␈αmanpower.␈αIt␈αseems␈αlikely␈αthat␈αit␈αwill␈αtake␈α2␈αor␈α
3␈α
people␈α
2␈α
or␈α
3␈α
years␈α
to␈α
get␈α
the␈α
system␈α
up
␈↓ ↓H
␈↓ ↓Hand␈α
running␈α
for␈α
users.␈α
In␈α
addition,␈α
there␈α
will␈αbe␈αsome␈αconversion␈αcosts␈αassociated␈αwith␈αmaking␈αold
␈↓ ↓H
␈↓ ↓Hprograms␈αrun␈αon␈αthe␈αnew␈αsystem.␈αAlso,␈αother␈αpeople␈αwill␈αhave␈αto␈αbe␈αfound␈αto␈αperform␈αthe␈αfunctions
␈↓ ↓H
␈↓ ↓Hof␈αthe␈αpeople␈αwho␈αwill␈αbe␈αworking␈αon␈αthe␈αnew␈αsystem.␈αIn␈αaddition,␈αthe␈αdevelopment␈αof␈αa␈α
new␈α
system
␈↓ ↓H
␈↓ ↓Hwill␈α
eventually␈α
entail␈α
large␈α
quantities␈α
of␈α
system␈α
down␈α
time␈α
(say␈α
six␈α
hours␈α
a␈α
day␈α
for␈α
several␈α
months).
␈↓ ↓H

␈↓ ↓H
␈↓ ∧F␈↓αDesign Goals for a New System␈↓
␈↓ ↓H

␈↓ ↓H
␈↓ α_Efficiency.␈αThe␈αnew␈αsystem␈αshould␈αuse␈αa␈αsmaller␈αfraction␈αof␈αthe␈αCPU␈αthan␈αthe␈αpresent␈αsystem,
␈↓ ↓H
␈↓ ↓Hfor␈α∂similar␈α∂loads.␈α∂Moreover,␈α∂the␈α∂new␈α∂system␈α∂should␈α∂be␈α∂able␈α∞to␈α∞execute␈α∞old␈α∞programs␈α∞with␈α∞only␈α∞a
␈↓ ↓H
␈↓ ↓Hminor␈α
degradation␈α
of␈α
performance.
␈↓ ↓H

␈↓ ↓H
␈↓ α_Reliability.␈α∞The␈α∞new␈α∞system␈α∞should␈α∞be␈α∞designed␈α∞with␈α∞reliability␈α∞in␈α∞mind.␈α∞There␈α∞are␈α∞various
␈↓ ↓H
␈↓ ↓Hways␈αto␈αdo␈αthis,␈αthe␈αoverall␈αdesign␈α
being␈α
the␈α
most␈α
important␈α
method.␈α
Various␈α
consistency␈α
checks␈α
and
␈↓ ↓H
␈↓ ↓Hdebugging␈α∂aids␈α∂can␈α∂(and␈α∂should)␈α∂be␈α∂built␈α∂in␈α∂to␈α∂expedite␈α∂the␈α∞debugging␈α∞of␈α∞the␈α∞system␈α∞when␈α∞it␈α∞is
␈↓ ↓H
␈↓ ↓Hbrought␈α
up,␈α
and␈α
to␈α
enhance␈α
its␈α
reliability.
␈↓ ↓H

␈↓ ↓H
␈↓ α_Display␈α
service.␈α
See␈α
"Goals␈α
for␈α
Timesharing"
␈↓ ↓H

␈↓ ↓H
␈↓ α_File␈α
system.␈α
See␈α
"Goals␈α
for␈α
Timesharing"
␈↓ ↓H

␈↓ ↓H
␈↓ α_Core␈α⊃Management/Paging/Scheduling.␈α⊃The␈α⊃system␈α⊃should␈α⊃be␈α⊂sufficiently␈α⊂modular␈α⊂to␈α⊂allow
␈↓ ↓H
␈↓ ↓Hsubstitute␈αalgorithms␈αto␈αbe␈αtried␈αeasily.␈αAlso,␈αsome␈αperformance␈αevaluation␈α
tools␈α
can␈α
be␈α
built␈α
into␈α
the
␈↓ ↓H
␈↓ ↓Hframework␈α
of␈α
the␈α
system␈α
to␈α
indicate␈α
those␈α
areas␈α
that␈α
can␈α
benefit␈α
from␈α
tuning.
␈↓ ↓H

␈↓ ↓H
␈↓ α_User/Executive␈αInterface.␈αInstead␈αof␈αissuing␈αcommands␈αto␈αthe␈αsystem,␈αa␈αuser␈αissues␈αcommands
␈↓ ↓H
␈↓ ↓Hto␈α
his␈αExecutive.␈αThe␈αExecutive␈αis␈αnot␈αprivileged␈αor␈αspecial␈αin␈αany␈αway,␈αtherefore,␈αa␈αuser␈αcan␈αwrite
␈↓ ↓H
␈↓ ↓Hhis␈α⊃own␈α⊃Exec,␈α⊃to␈α⊃make␈α⊃the␈α⊃system␈α⊃perform␈α⊃in␈α⊂whatever␈α⊂way␈α⊂he␈α⊂wants.␈α⊂Different␈α⊂Execs␈α⊂can␈α⊂be
␈↓ ↓H
␈↓ ↓Hevaluated.␈α
An␈α
Exec␈α
that␈α
is␈α
itself␈α
a␈α
programming␈α
language␈α
could␈α
be␈α
developed.
␈↓ ↓H

␈↓ ↓H
␈↓ α_Process␈α∩interrupts␈α∩will␈α∩be␈α⊃designed␈α⊃in␈α⊃from␈α⊃the␈α⊃ground␈α⊃up,␈α⊃unlike␈α⊃the␈α⊃way␈α⊃they've␈α⊃been
␈↓ ↓H
␈↓ ↓Hineffectually␈α
kludged␈α
into␈α
existing␈α
systems␈α
(except␈α
ITS␈α
which␈α
does␈α
it␈α
right).
␈↓ ↓H

␈↓ ↓H
␈↓ α_Interprocess␈αinterface.␈αA␈αflexible,␈αwell␈αdesigned␈αinterface␈αbetween␈αcooperating␈αprocesses␈αwill␈α
be
␈↓ ↓H
␈↓ ↓Hdesigned.␈α⊂It␈α⊂will␈α⊂include␈α⊂(and␈α⊂surpass)␈α⊂the␈α⊂presently␈α⊂existing␈α⊂MAIL␈α⊂and␈α⊂JOBRD␈α⊂UUOs.␈α∂It␈α∂will
␈↓ ↓H
␈↓ ↓Hprovide␈α∩a␈α∩means␈α⊃for␈α⊃inferior␈α⊃forks␈α⊃to␈α⊃be␈α⊃controlled␈α⊃by␈α⊃their␈α⊃superiors.␈α⊃It␈α⊃will␈α⊃provide␈α⊃flexible
␈↓ ↓H
␈↓ ↓Htrapping␈α
and␈α
reporting␈α
of␈α
error␈α
conditions␈α
in␈α
the␈α
inferior.␈α
Processes␈α
can␈α
interact␈α
as␈α
coroutines.
␈↓ ↓H

␈↓ ↓H
␈↓ α_Program/System␈α∪interface.␈α∪User␈α∪programs␈α∪will␈α∩be␈α∩able␈α∩to␈α∩access␈α∩system␈α∩routines␈α∩in␈α∩some
␈↓ ↓H
␈↓ ↓Hmethod␈α
at␈α
least␈α
as␈α
flexible␈α
as␈α
UUOs.
␈↓ ↓H

␈↓ ↓H
␈↓ α_Realtime␈α∞applications␈α∞will␈α∞be␈α∞handled␈α∞by␈α∞some␈α∞technique␈α∞that␈α∞is␈α
similar␈α
to␈α
Spacewar␈α
Mode,
␈↓ ↓H
␈↓ ↓Hbut␈α
which,␈α
hopefully,␈α
will␈α
not␈α
endanger␈α
the␈α
system␈α
so␈α
much.
␈↓ ↓H

␈↓ ↓H
␈↓ α_There␈α
may␈α
be␈α
some␈α
aspects␈α
of␈α
the␈α
new␈α
system␈α
design␈α
that␈αcan␈αbe␈αconsidered␈αresearch␈αtopics.
␈↓ ↓H
␈↓ ↓HAlmost␈α
certainly␈α
there␈α
is␈α
research␈α
to␈α
be␈α
done␈αin␈αscheduling/␈αcore␈αmanagement.␈αThere␈αmay␈αbe␈αsome
␈↓ ↓H
␈↓ ↓Hresearch␈αtopic␈αin␈αthe␈αdesign␈αand␈αefficient␈αimplementation␈αof␈αtree-structured␈α
files.␈α
There␈α
are␈α
probably
␈↓ ↓H
␈↓ ↓Hother␈α
areas␈α
of␈α
research.
␈↓ ↓H

␈↓ ↓H