/LMAR=0/XLINE=3/FONT#0=BASL30/FONT#1=BASI30/FONT#2=BASB30/FONT#10=BAXM30/FONT#11=ZERO30/FONT#3=STA200/FONT#4=NGB25
␈↓ ↓H␈↓␈↓βS␈↓∧ Artificial Intelligence Laboratory, STANFORD UNIVERSITY, Stanford, California 94305

␈↓ ↓H␈↓∧Telephone 415 497-4430␈↓ 	ISeptember 16, 1977 




␈↓ ↓H␈↓Mr. Joel Trimble
␈↓ ↓H␈↓Sensor and Control Technology Division
␈↓ ↓H␈↓O≠ce of Naval Research
␈↓ ↓H␈↓800 South Quincy St
␈↓ ↓H␈↓Arlington, VA 22217

␈↓ ↓H␈↓Dear Mr. Trimble:

␈↓ ↓H␈↓        Lowell␈α∞Wood␈α∞has␈α∞asked␈α∞me␈α∞to␈α∞write␈α∞you␈α∞giving␈α∞my␈α∞opinion␈α∞of␈α∞the␈α∞S-1␈α∞computer␈α∞project.
␈↓ ↓H␈↓I␈αhave␈αfollowed␈αthis␈αproject␈αsince␈αits␈αinception.␈α Initially␈αI␈αwas␈αskeptical␈αabout␈αwhether␈αthe␈αgroup
␈↓ ↓H␈↓could␈αdesign␈αa␈αlarge␈αcomputer,␈αbut␈αnow␈αit␈αis␈αclear␈αthat␈αthey␈αcan␈αand␈αthat␈αthe␈αcomputer␈αis␈αlikely␈αto
␈↓ ↓H␈↓work.␈α
 Whether␈α∞it␈α
will␈α∞be␈α
as␈α∞fast␈α
as␈α∞they␈α
expect␈α
still␈α∞remains␈α
to␈α∞be␈α
seen,␈α∞but␈α
I␈α∞see␈α
no␈α∞reason␈α
at
␈↓ ↓H␈↓present for more than the usual worry about any computer design.

␈↓ ↓H␈↓        Ever␈αsince␈αthe␈αarchitecture␈αbecame␈α≡rm,␈αI␈αhave␈αbeen␈αenthusiastic␈αabout␈αit,␈αbecause␈αit␈αis␈αthe
␈↓ ↓H␈↓≡rst large computer architecture to solve two important problems:

␈↓ ↓H␈↓        1.␈αIt␈αhas␈α
a␈αlarge␈αenough␈αaddress␈α
for␈αthe␈αnext␈α≡fteen␈α
years.␈α The␈αD.E.C.␈αPDP-10,␈α
which␈αwe
␈↓ ↓H␈↓use␈αin␈α
our␈αlaboratory,␈αis␈α
nearing␈αthe␈αend␈α
of␈αits␈αutility␈α
for␈αlack␈αof␈α
address,␈αand␈αfor␈α
many␈αreasons,
␈↓ ↓H␈↓the IBM 370 architecture is not a satisfactory successor.

␈↓ ↓H␈↓        2.␈α∞The␈α∞S-1␈α∞embodies␈α∂the␈α∞most␈α∞plausible␈α∞approach␈α∂to␈α∞getting␈α∞e≠cient␈α∞computation␈α∂from␈α∞a
␈↓ ↓H␈↓multi-processor␈α∀computer.␈α∀ Namely,␈α∃each␈α∀processor␈α∀has␈α∀its␈α∃own␈α∀control␈α∀unit,␈α∃each␈α∀processor
␈↓ ↓H␈↓addresses␈α
the␈α
whole␈α
of␈α
memory,␈α
and␈α
tasks␈α
that␈αcan␈α
be␈α
done␈α
in␈α
parallel␈α
are␈α
put␈α
on␈α
queues␈αand␈α
then
␈↓ ↓H␈↓done␈α
by␈αwhatever␈α
processors␈α
become␈αfree.␈α
 I␈α
have␈αlong␈α
been␈α
convinced␈αthat␈α
such␈αa␈α
multi-processor
␈↓ ↓H␈↓computer␈α∞can␈α
be␈α∞programmed␈α
to␈α∞use␈α
all␈α∞processors␈α
e≠ciently␈α∞on␈α
large␈α∞problems␈α
with␈α∞little␈α
more
␈↓ ↓H␈↓di≠culty␈α
than␈αprogramming␈α
a␈αsingle␈α
processor,␈α
with␈αonly␈α
minor␈αadditions␈α
to␈αpresent␈α
programming
␈↓ ↓H␈↓languages,␈α∂and␈α∂often␈α∂with␈α∂only␈α∂minor␈α∂changes␈α∂to␈α∂present␈α∂programs.␈α∂ We␈α∂also␈α∂believe␈α⊂that␈α∂this
␈↓ ↓H␈↓machine organization is suitable for general time-sharing use.

␈↓ ↓H␈↓        I␈α∞anticipate␈α∞that␈α∞the␈α∞S-1␈α∞project␈α∞will␈α∞succeed,␈α∞and␈α∞if␈α∞it␈α∞does,␈α∞we␈α∞would␈α∞hope␈α∞to␈α∞acquire␈α∞a
␈↓ ↓H␈↓system for our Laboratory and for the Stanford University Computer Science Department.


␈↓ ↓H␈↓Sincerely,



␈↓ ↓H␈↓John McCarthy
␈↓ ↓H␈↓Director
␈↓ ↓H␈↓Professor of Computer Science