274 16 29MB
English Pages 162 Year 1992
LOGIC PROGl\AMMING
...,, rr...,, 0
...,, -0 ..,., rv -0
00
0 ..,.,
"How I Got Started with Mathematica®" A few years ago, I was working with
haved, the matrices were nonsingular.
about symbolic math.
another engineer at Boeing on a project
But because of the limits of the significant
about the limits of significant digits in
digits we got with Fortran, we were run-
Fortran, C, or any other programming
I
language because with Mathematica I can
hadn't used it, I was willing to try sorting
maintain hundreds of significant digits
out the problem with Mathematica.
with no concerns about inaccuracies or
ning into this problem.
Although
Using Mathe111atica, we discovered
for the Air Force that required extensive symbolic manipulations.
of equations wasn't singular,
partial differential equations.
that it did have a solv-
My col
league said, "Why don't we try Mathe-
able set.
matica? It does exactly what we want to
rewrote our code
do." But I'd never used it.
knowing every
So we
I was impressed.
Fortran and were getting numerical We knew that was a
was able to handle the symbolic manipu
mistake-the equations were well be-
lations we needed and more. [t solved dif-
eql = Coe.fficient[equation'm; the nonmember annual subscription Is $109. see top One Of malllng label tor subscrlpt!On expiration date coded In tour digits: the first two are year, last two, month Of expiration. Microfilm editions through 1980 can be purchased from Microfilm Dept., PubllShlng services Dlv.. waverly Press, 42B East Preston Street, aaltlmore, MD 21202. Microfilm and microfiche are alSo avallable from University Mlcrofllms Interna tional, 300 North Zeeb Road, Dept. PR, Ann Arbor, Ml 4B106; 18001521-o&OO. 51•11• copies are SB to members and $17 to nonmembers. Please send onters prepaid plus $4 tor shipping and handling to ACM onter Dept., P.O. Box 64145, Baltimore, MD 21264. FDr credit cant onters can 1800> 342·6626 Un Baltimore, Alaska, or Canada, call 13011521l-42611. Onter personnel on duty B:154:45 EST. After hours, please leave message and onter personnel wlll return your call.
M
EDITORIAL POINTERS
any people associate logic programming LP with Prolog, the language that was born in France in the early 1970s and gained attention in 1981 as the choice for Japan's ambitious Fifth Generation Computer Project. To others, logic programming brings to mind particular application domains, such as expert systems, natural language processing, compiler writing, VSLI circuit analysis, or relational databases. To yet another group logic programming will probably signify numerous combined endeavors in logic and com puter science that are producing fascinating developments in the areas of massively parallel architectures, operations research, numerical analysis, and the development of large, less error-prone, software packages. We hope this month's issue will provide you with fresh information about the LP field regardless of which group you fit. Communications's coverage begins with Guest Editor Jacques Cohen's overview of current topics in LP and his speculations on the future. JA. Robinson, author of the historical article, is credited with setting the path for many of the developments that followed the publication of his 1965 Journal of the ACM paper, "A Machine· Oriented Logic Based on the Resolution Principle." Following the Robinson article, John Grant and Jack M inker demonstrate the significant impact that logic programming has had on databases, particularly on the understanding of database semantics, on the extension of the concepts of relational databases, and on new techniques and tools for database users. Next, Koichi Furukawa describes the role of logic programming in Japan's Fifth Generation Project. As one of the leaders of that project, Furukawa is in an excellent position to present examples of the latest Japanese efforts. He stresses the application aspects of concurrent variants and extensions of logic pro gramming, which have been vigorously pursued by three groups: one at ICOT, Japan, and the other two at Imperial College, England, and the Weizmann Institute, Israel. The closing article by David S. Warren, one of the early American LP researchers, contrasts with the others in that it focuses on three specific research topics: the bottom-up evaluation of logic programs, abstract interpretations, and partial deduction. As a complement to the special section, Computing Practices features an article on implementing sets in Prolog by Toshinori Munakata. Because of its unique characteristics, Prolog requires special techniques for imple menting sets. Munakata presents a comprehensive overview of both existing and new methods for defining and manipulating sets in the language. Rounding out the issue is an article by Marcia C. Linn and Michael J. Clancy that examines how programmers organize their knowledge and how program design skills are commonly taught. At the core of their article is an empirical investigation that demonstrates the effectiveness of case studies for communicating the complex, multidimensional decisions necessary when designing an effective computer program. Next Month: Special Issue on CASE
WBPA
COMMUNICATIONS OF THE ACM/March 1992/Vol.35, No.3
.:]::.
...
,... mo�
Executive Editor
5
ADVANCING HUMAN CAPABIL/1'lES THROUGH INPORMAT!ON TECHNOLOGY
Founded in 194 7 as the society of the computing community, the
Association for Computing Machinery is dedicated to the development of information processing as a discipline, and to the responsible use of
computers in an increasing diversity of applications. The purposes of ACM are to advance the sciences and arts of information processing, to promote the free interchange of iriformation among specialists and the public, and to develop and maintain the integrity and competence of individuals in the field. Executive Director: Joseph S. DeBlasi
Members-at-Large: V inton G. Cerf (1990-
Deputy Executive Director-Operations
B. Wilson (1988-1992); William A. Wulf
Patricia Ryan;
(1990-1994);
Director, MIS: Joseph P. Sullivan
Central Region, Helen C. Takacs Ouly 1,
1992); Michael R . Garey (1988-1992); Jan
Associate Director: VVayne Graves Director, Office of Financial Services: Michael Lichtenstein; Associate Director, SIG Financial Reporting: John DeLorenzo; Assistant Director, Budgeting and
Regional Representatives:
1991-June 30, 1993); Eastern Region, Gwen Bell Ouly 1, 1991-June 30, 1994); Western Region, Anita Borg Ouly 1, 1991-June 30,
1993); International Region, Maurice V . W ilkes Ouly 1, 1991-;]une 30, 1994)
Financial Planning: Russell Harris; Assistant Director, Financial Opera
Board Chairs: Education Board: A. Joe
tions: Oliver Ali
Turner; SIC Board: Mark Scott Johnson;
Director, Office of Membership:
Publications Board: W illiam B. Gruener;
James M. Adams
Local Activities Board: William B. Poucher
Membership
& Public Relations
Board: Vacant;
Associate Director, Strategic Initiatives: Fred Aronson; Associate Director, Membership and Marketing Services: Lillian Israel Director, Office of Publications: Mark Mandelbaum; Associate Director, CACM, Quarterlies, ACM Press Books, SIG Publishing: Ja net Benton; Associate Director, Computing Reviews, Guide, Database and Electronic Products: Bernard Rous Director, Office of SIG Services: Pegotty Cooper Associate Director, Office of SIG Services Donna Baglio PLAN, ACT, ART, SAM, NUM Program Directors, Office of SIG Services: GRAPH: Lois Blankstein; CHI, Ada, GAS, CAPH, UCCS: Diane Darrow;
SIG Chairs: SIGACT S. Rao Kosaraju;
submitted to the execuuve Editor fOllDWlng the Information fOr Authors printed In the January 1992 Issue.
Stuart C. Shapiro; SIGBIO: Roy Rada; Ephraim P. Glinert; SIGCAS: Ronald E. Anderson; SIGCHI: Austin Henderson; SIGCOMM: A. Lyman Chapin; SIGCPR:
Raymond McLeod, Jr.; SJGCSE: Nell B. Dale; SIGCUE: Janet Azbell; SJGDA: Michael], Lorenzetti; SIGDOC: R . John Brockmann; SIGForth: Irving Montanez; SIGGRAPH: James]. T homas; S!GIR:
Edward A.
Fox;
SIGLJNK: Robert Akscyn;
SJGMETR!CS: Michael K. Molloy; SJGMICRO: Joseph L. Linn; SIGMOD:
Won Kim; SIGNUM: Robert B. Schnabel; SIGOIS: Carson C. Woo; S!GOPS: Henry
M . Levy; SIGPLAN: Stuart I. Feldman; SIGSAC: Daniel P. Faigin; SJGSAM: Keith
SIGSMALLIPC. E.A. Unger; SIGSOFT
Vaught
For Information from Headquarters:
Vice-President: S. Ron Oliver; Secretary:
Membership Services Manager: Linda
Barbara Simons; Treasurer: John H. Esbin;
Berg; Program Manager, Chapter and
Chair, Publications Board: William B.
Local SIG Activities: Beth Olson;
Gruener; Chair, SIC Board: Mark Scott
Program Manager, Education/ACM
Johnson; Past President: Bryan S. Kocher;
Conferences: Don Nowak
G
Su•lsslal 111fo1111tlol: All manuscripts should be
SIGBIT Elias Awad; SIGCAPH:
R ichard N. Taylor; SIGUCCS: Russell S.
ACM Council: President: John R. White;
EDITORIAL COlllTTEE Chair: Peter J. Denning; Christine A. Molltgomerv; Herbert D. SChWetman; Alan Shaw; Glo Wledefhold
Lynne C. Shaw; SIGAPP: Hal Berghel;
MICRO, SMALL/PC: Debbie Hall; APL, METRICS, CUE, CSE: Pat McCarren
PUBLICATIDNS BOARD Chair: Wllllam B. Gruener; Gwen Bell; Hal Berghel; Peter J. Denning; Edward A. Fax; Rlehanl Kleburtz; Peter wegner; Stuart zweben
SIGARCH. Alan Jay Smith; SIGART
0. Geddes; SIGSIM: C. Michael Overstreet;
APP, BIO, FORTH, MOD, OPS, SIM,
COMPUTING PRACTICB EDITORIAL PANEL Chair: Edgar H. Sibley; Eric Clemons; tanee B. Ellot; Chr1stopher Fax; Do�ld Gantz; ChHs Kemerer; Brian NeJmeh; SOI Shatz
SIGAda: Mark S. Gerhardt; SIGAPL:
COMM, SOFT Lisa Ernst; DA, ARCH,
BIT, CPR, LINK, OIS, SAC, DOC, IR,
ARTICLB EDITORIAL PAIE.L Jacques Cohen; Flavlu Cr1stlan; Jack Dongarra; Peter Friedland; James GOOdman; Robert M. Harallck; won Kim; Rob Kling; Mlchael A. tangston; Henry F. Ledganl; Darts Udtke; Halm Mendelson; John RUshby; Doug Terry; Jeffrey s. Vltter
March 1992/Vol .35, No.3/COMMUNICATIONS OFTHE ACM
Advancing human capabilities through information technology
Membership Application
acm
Mai l to: ACM, P.O. Box 12114, Church Street Station, New York, NV 10257
You may use this ACM Membership Application to: 1) Join ACM and take advantage of the special Publication and SIC rates for ACM members; 2) Join the Special Interest Group(s) of your choice at the SIC Member (non-ACM member) rates on the reverse side; individuals joining S!Gs using these rates are not entitled to ACM membership privileges.
Please Print
Personal Information First, Middle Initial, Last Name Address Address Address
City/State/Zip or City/Country I Postal Code (___ )
Day Telephone Number
ACM Membership ACM Membership includes a subscription to the monthly
D
Communications of t/ze ACM
D D
Check member class desired
D D
D
SIG Membership Non-ACM Members Only
Mailing List Optional
Purposes of ACM
Signature Required
Note: Membership dues include
$30.00 ($20.00 for students) toward a subscription to Communications of the ACM.
D
For Office Use Only ________________
$75.00 Voting Member Applicants You must subscribe to the Purposes of ACM; have attained professional stature as demonstrated by intellectual competence and ethical conduct in the arts and sciences of information processing; and must satisfy at least one of the requirements at the right.
1. Bachelor's Degree. Institution:_______________ 2. Equivalent level of education. Institution:___________
3. Four full-time years of experience (attach a brief statement). I attest the above is correct___________ ______
Signature
$75 Associate Member Applicants
You must subscribe to the purposes of ACM. Associate Members may convert to Voting Member status at any time by writing ACM Headquarters for a "Self-Certification" form.
$22.00 Student Member Applicants You must be registered in an accredited educational institution and a Faculty Member must certify your full-time status.
Institution: _______________________ Faculty Member's Signature_________________ Expected Graduation Date: ____ / _ _ __ MM
$70.00 Joint Membership -IEEE-CS Members of the IEEE-CS receive a $5 dues discount. (not Affiliates with "N9" member#'s).
YY
Society: Member#:_____________
$60.00 Joint Membership - International Computing Societies ACS (Australia), ADIG (Guatemala), AFCET (France), AICA (Italy), API (Portugal), ATI (Spain), BCS (United Kingdom), BIRA/IBRA (Belgium), CIPS (Canada), CSIS (Czechoslovakia), CSZ (Zimbabwe), Cl (Germany), HKCS (Hong Kong), ICS (Ireland), IPA (Israel), !PS) (Japan), NCI (Netherlands), NZCS (New Zealand), SCCC (Chile), SCS (Shanghai). Spouse or Retired Membership: For more information, please write to ACM Membership Services Department,
11 West 42nd Street, New York, NY, 10036
SIG Membership only. (Non-ACM Members only.) Membership in special interest group(s) of your choice includes a newsletter subscription. SIC non-ACM members are not eligible for ACM SIC member rates or ACM membership privileges. See "SIC Member (11011-ACM) Rates" 011
reverse.
ACM occasionally makes its membership list available to companies and societies for computer-related mailings. If you wish to restrict the use of your name for these purposes, please check one of the following:
D D
ACM announcements only. ACM and sister society announcements only.
To advance the sciences and arts of information processing; to promote the free interchange of information about the sciences and arts of information processing both among specialists and among the public; and to develop and maintain the integrity and competence of individuals engaged in the practice of information processing.
I
For Office Use: Recruit/Promotion Code
CACM-3/92
I
I hereby affirm that I subscribe to the purposes of ACM and understand that my membership is not transferable.
Signature
(OVER)
Date Effective 10/91
Publications Definitions:
Available to overseas members. Circle desired rate(s) and add to publication subscription and/ or SIG membership rates(s). Air Options:
Partial Air Services -
Air freight to Amsterdam and Dutch surface mail. Available only to Europe, India, Africa, and Mideast. Full Air Service - Air service from U .S. Available to all
overseas locations including Hawaii.
Circle appropriate rate(s) and indicate subtotal. Membership includes a subscription to ACM's flagship publication Communications of the ACM. If air service is desired for member copy of CACM, circle appropriate Air Option. Code #
You will receive the current published edition.
ACM No-Nonsense Guide to Computing Careers . . . 120 . . Transactions on: (all quarterlies) . . . . . . . . . . . . . . . . . 108 . . Mathematical Software /TOMS . Database Systems /TODS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 . . t s s a e a d an PL S : ./ . . . Information Systems /TOIS . . . . . . . . . . . . . . . . . . . . . . . . . 1 13 . . Computer Systems /TOCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 . . Software Engineering & Methodology /TOSEM . . . . . 1 1 5 . . Modeling & Computer Simulation / O MAC S . . . . . . . 1 1 6 . . Letters on Programming Languages and Systems I . . . . . . . . . . . . . . . . . . 117. LOPLAS* . . . . . . . . . . . . . . . . . . . . . . . .
�i°a�hk;/�J� �� � � � �r. �� �� � : : : ii�· .
>t£xpected Publication Date May, 1992
members who are not ACM members do not receive a subscription to Communicntions Note: SIG
of t/1e ACM.
Definitions: Air Options:
Available to overseas members. Circle desired rate(s) and add to publication subscription and / or SIG membership rate(s). Partial Air Service - Air freight to Amsterdam and Dutch surface mail. Available only to Europe, India, Africa, and M ideast. Full Air Service - Air service from U.S. Available to all
overseas locations including Hawaii.
Voting/
Associate
Student
Communications of the ACM (monthly) addt'I subscri ptions only . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 . . . . . . . $25.00. . . . . Membership includes a subscription to CACM Journal of the ACM ( g uarterl y ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 . . . . . . . 22.00 . . . . . Computing Surveys (quarterly) . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 . . . . . . . 15.00 . . . . . Computing Reviews (monthly) . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 . . . . . . . 34.00 . . . . . Collected Algorithms, Vols I , II, Ill, IV, V & 1 yr's quarterl y updating supp lements . . . . . . . . . . . . . . . . . . . . . . . . . 105 . . . . . . . 215.00 . . . . . ACM Guide to Computmg Literature (annual) . . . . . . . 160 . . . . . . . 97.00 . . . . .
T
Special Interest Groups (SIGs)
Add Appropriate Overseas Air Option(s) to Voting/Associate or Student Rate(s) Full Air
Partial
Total
$20.00 . . . . + $30 .00 . . . $60.00 . . . . = $ ___ 17.00 . . . . + 20.00 . . . 25.00 . . . . . = $ ___ 10.00 . . . . + 15.00 . . . 30.00 . . . . . = $ ___ _ _ 29.00 . . . . + 20.00 . . . 50.00 . . . . . = $ _ 190.00 . . . . + 10.00 . . . 20.00 . . . . . = $ ___ 92.00 . . . . no air offered . . . . . . . . = $ ___
20.00 . . . . .
15.00 . . . .
+ 6.00 . . .
26.00 . . . . . 25.00 . . 24.00 . . . . 30.00 . . . 26.00 . . 26.00 22.00 . . 30.00 . .
21 .00 . . 20.00 . . . . 19.00 . . . . 25 . 00 . . 21 . 00 . . 2 1 . 00 . . 17.00 25.00
"
+ 10.00 . . . + 10.00 . . . + 10.00 . . . + 10.00 . . . + 10.00 . . + 10.00 . . + 10.00 . . . + 10.00 . . .
26.00.
21.00 . . . .
+ 10.00 . .
8.00 . . . . . = $ ___ _ 20.00 . . . . . = $ __ 20.00 . . . . . = $ ___ 20.00 . . . . . = $ ___ 20.00 . . . . . = $ ___ 20.00 . . . . = $ ___ _ 20.00 . . . . . = $ __ _ 20.00 . . " $ __ 20.00 . . . . . = $ ___
=
20.00 . . . . . = $ ___
PUBLICATION SUBTOTAL $
Circle appropriate rate(s) and indicate subtotal. SIG Membership includes a Newsletter subscription.
Code # SIGACT (Automata and Computability
Add Appropriate Overseas A ir Option(s) to Voting/ Associate, Student or SIG Member Rate(s)
ACM Members A����§{e Student
SIG Mbrs �g�--���s
Partial
Full Air
Total
Theory) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 001 . . . . . $ 1 5 . 00 . . . . . . $7.50 . . . . . $40.00 . . + $ 1 6 . 00 . . . . $32.00 = $ SIGADA (Ada) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 037 . . . . . . 1 5 . 00 . . . . . . 1 0 . 00 . . . . . . 37.00 . . + . 22.00 . . . . .44.00 = $ SIGAPL (APL) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 032 . . . . . . 1 8 . 00 . . . . . . 1 2.50 . . . . . . 45.00 . . + . . . 8 . 00 . . . . . 20.00 = $ $ SIGAPP (A pplied Com p uting) . . . . . . . . . . . . . . . . . . . . . 042 . . . . . . 1 5 . 00 . . . . . . . . 7.50 . . . . . . 39.00 . . + . . . 5 . 00 . . . . . 1 2 . 00 SIGARCH (Computer Architecture) . . . . . . . . . . . . . . 002 . . . . . . 28.00 . . . . . . 14.00 . . . . . . 54.00 . . + . 22.00 . . . . . 65.00 - $ SIGART (Artificial Intelligence) . . . . . . . . . . . . . . . . . . . . 003 . . . . . . 1 5 . 00 . . . . . . . . 8 . 00 . . . . . . 40.00 . . + . 1 1 .00 . . . . . 22.00 = $ SIGBIT (Business Information Technology) . . . . . 004 . . . . . . 1 8 . 00 . . . . . . 1 2 . 00 . . . . . . 38.00 . . + . . . 6 . 00 . . . . . . . 9 . 00 = $ SIGBIO (Biomedical Computing) . . . . . . . . . . . . . . . . . . 005 . . . . . . 20.00 . . . . . . . . 7.00 . . . . . . 30.00 . . + . 1 0 . 00 . . . . . 1 6 . 00 = $ SIGCAPH (Computers and the "Physically Handica pped-Print Edition). : . . . . . . . . . . . . . . . . . . . . . . 006 . . . . . . 1 2 . 00 . . . . . . . . 5.00 . . . . . . 34.00 . . + . . . 5 . 00 . . . . . . . 7.00 $ SIGCAS (Computers and Society) . . . . . . . . . . . . . . . . . 007 . . . . . . 1 8 . 00 . . . . . . . . 6 . 00 . . . . . . 5 1 . 0 0 . . + . . . 7 . 00 . . . . . 1 0 . 00 - $ SIGCHI (Computer and Human Interaction) . . . 026 . . . . . . 25.00 . . . . . . 1 0 . 00 . . . . . . 47.00 . . + . 1 6 . 00 . . . . . 50.00 = $ SIGCOMM (Data Communication) . . . . . . . . . . . . . . . 008 . . . . . . 22.00 . . . . . . 1 5 . 00 . . . . . . 50.00 . . + . 1 0 . 00 . . . . .40.00 = $ SIGCPR (Computer Personnel Research) . . . . . . . . 0 1 0 . . . . . . 1 5 . 00 . . . . . . . . 5.00 . . . . . . 20.00 . . + . . . 7.00 . . . . . 1 0 . 00 = $ SIGCSE (Computer Science Education) . . . . . . . . . . 0 1 1 . . . . . . 1 6 . 50 . . . . . . . . 7.50 . . . . . . 40.00 . . + . 1 0 . 00 . . . . . 35.00 = $ SIGCUE (Computer Uses in E ducation) . . . . . . . . . 0 1 2 . . . . . . 1 5 . 00 . . . . . . . . 7 . 00 . . . . . . 50.00 . . + . . . 9 . 00 . . . . . 2 0 . 00 = $ SIGDA (Design Automation) . . . . . . . . . . . . . . . . . . . . . . . 013 . . . . . . 1 0 . 00 . . . . . . 1 0 . 00 . . . . . . 1 8 . 00 . . + . . . 7.00 . . . . . 1 5 . 00 = $ SIGDOC (Documentation) . . . . . . . . . . . . . . . . . . . . . . . . . . 033 . . . . . . 1 2 . 00 . . . . . . . . 5 . 00 . . . . . . 34.00 . . + . . . 6 . 50 . . . . . 1 1 . 00 = $ SIGFORTH (Forth) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 039 . . . . . . 20.00 . . . . . . 1 1 .00 . . . . . . 42.00 . . + . . . 5 . 00 . . . . . 1 3 .00 = $ SIGGRAPH (Computer Gra p hics) . . . . . . . . . . . . . . . . 015 . . . . . . 24.00 . . . . . . 1 6 . 00 . . . . . . 59.00 . . + . 1 5 . 00 . . . . . 34.00 = $ SIGIR (Information Retrieval) . . . . . . . . . . . . . . . . . . . . . . . 016 . . . . . . 1 2 . 00 . . . . . . . . 6 . 00 . . . . . . 56.00 . . + . . . 9 . 00 . . . . . 38.00 = $ SIGLINK (Hypertext/Hypermedia) . . . . . . . . . . . . . . . 043 . . . . . . 20.00 . . . . . . 1 2 . 00 . . . . . . 56.00 . . + . . . 5 . 00 . . . . . 1 2 . 00 = $ SIGMETRICS (Measurement & Evaluation) . . . . 019 . . . . . . 20.00 . . . . . . 1 0 . 00 . . . . . . 45.00 . . + . . . 7.00 . . . . . 25.00 = $ SIGMICRO (Microprogramming) . . . . . . . . . . . . . . . . . 020 . . . . . . 1 2 . 00 . . . . . . . . 7.00 . . . . . . 34.00 . . + . . . 4 . 00 . . . . . 1 2 .00 $ SIGMOD (Management of Data) . . . . . . . . . . . . . . . . . 014 . . . . . . 20.00 . . . . . . 1 2 . 00 . . . . . . 25.00 . . + . 1 3 . 50 . . . . . 51 .00 - $ SIGNUM (Numerical Mathematics) . . . . . . . . . . . . . . . 021 . . . . . . 1 5 . 00 . . . . . . . . 7.50 . . . . . . 23.00 . . + . . . 8 . 00 . . . . . 1 1 . 00 = $ SIGOIS (Office Information Systems) . . . . . . . . . . . . 027 . . . . . . 1 5 . 00 . . . . . . . . 7.50 . . . . . . 37.00 . . + . . . 8 . 00 . . . . . 1 7 . 00 = $ SIGOPS (Op erating Sy stems) . . . . . . . . . . . . . . . . . . . . . . . 022 . . . . . . 1 5 . 00 . . . . . . . . 8.00 . . . . . . 35.00 .. + . 14.00 . . . . . 35.00 $ SIGPLAN (Prog ramming Lan g uages) . . . . . . . . . . . . 023 . . . . . . 25.00 . . . . . . 1 0 . 00 . . . . . . 52.00 . . + . 40.00 . . . . . 90.00 = $ FORTRAN FORUM (Fortran) Newsletter Only 038 . . . . . . 1 0 . 00 . . . . . . . . 6.00 . . . . . . 20.00 .. + . . . 6 . 00 . . . . . 1 5 . 00 - $ LISP POINTERS (LISP) Newsletter Only . . . . . . . 040 . . . . . . 1 2 . 00 . . . . . . . . 7.00 . . . . . . 25.00 . . + . . . 6 . 00 . . . . . 1 5 . 00 = $ OOPS MESSENGER (Object-Oriented Prog ramming Systems) Newsletter Only . . . . . . . 041 . . . . . . 1 0 . 00 . . . . . . . . 6 . 00 . . . . . . 20.00 .. + . . . 6.00 . . . . . 1 5 . 00 = $ SIG SA C (Security, Audit & Control) . . . . . . . . . . . . . . 036 . . . . . . 1 5 . 00 . . . . . . . . 5 . 00 . . . . . . 35.00 . . + . . . 7.50 . . . . . 1 5 . 00 = $ SIGSAM (Symbolic & Algebraic Mani p ula�ion) . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 024 . . . . . . 1 5 . 00 . . . . . . . . 5 . 00 . . . . . . 20.00 . . + . . . 8.00 . . . . . 1 2 . 00 $ SIGSIM (S1mulahon) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 025 . . . . . . 1 2 . 00 . . . . . . . . 6 . 00 . . . . . . 34.00 .. + . . . 9 . 00 . . . . . 1 6 . 00 - $ SIGSMALL/PC (Small & Personal Computing Systems & A p plications . . . . . .. . . . . . . . . . . . . . . . . . . . . . . 031 . . . . . . 1 7.00 . . . . . . 1 0 . 00 . . . . . . 38.00 . . + . . . 8 . 00 . . . . . 1 7 . 00 $ SIGSOFT (Software Engmeenng) . . . . . . . . . . . . . . . . . 034 . . . . . . 20.00 . . . . . . 1 0 . 00 . . . . . . 40.00 . . + . 1 3 . 00 . . . . .44.00 - $ SIGUCCS (University & College Computing Services) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 028 . . . . . . 1 5 . 00 . . . . . . . . 5.00 . . . . . . 40.00 .. + . . Free . . . . . . Free = $
'.'.'.
'.'.'
'.'.'
.
'.'.'.
'.'.'
)
'.'.'.
SIG SUBTOTAL
Payment Information Mail this application with your payment to: ACM P.O. Box 12114 Church Street Station New York, NY 10257
ACM accepts payment by personal check, money order, or credit card. For international residents, payment must be in U .S. dollars drawn on a U.S. bank. Please make checks payable to ACM, Inc. I
wish to pay by:
0 American Express
Card
0 Check (make payable to ACM, Inc.) 0 MasterCard
0 VISA
0 Credit Card
#: I I I I I I I I I I I I I I I I ! Card Exp . Date _ / _
Signature:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
ACM Member Dues
(from reverse side)
Publication Subtotal
S I G Sub total GRAND TOTAL
$
$ $
$
$
__ __ __
__
__
__
__ __
__ __
__ __
__
__
__
__ __ __
__ __
__ __ __
__
__
__ __
__ __
__
__
__
__
__
__
__
__
� .. BUSH PUSH . . .
The White House solicited input from several sci entific agencies and consortia when drafting the first National Tech nology Initiative-the Administration's latest � effort to reroute the descending curve of U.S. com petitiveness. Among the topics under discussion were coop erative R&!D partnerships, industry needs, intellectual property laws and collabora tive ideas. Reportedly, one of the top priorities of the Bush cabinet is to reorganize the nation's 726 laboratories to fortify the best return on federal funding. TATTLERS HOTLINE .
.
•
·�
u
.
The first Congressional computer bulletin board for whistleblowers has been announced by Rep. Bob Wise (D-WV) who will act as system operator. The Whistleblower Bulletin Board will provide an alternative means for federal employees and citi zens with information on government waste, fraud and abuse to contact congres sional investigators. The board allows two-way com munications between inves tigative staffers and potential sources; no one will be re quired to use his or her real name. Anyone with a PC and modem can contact the board round-the-clock all year long. Computerized files
COMMUNICATIONS OF THE ACM/March 1992/Vol.35, No.3
j
m
accepted. (202) 225-5527 . LESS THAN SUPEB . . . Britain seems to be paying the price for letting super computing investments lag so far behind the U.S. , Japan, Germany and Canada. Today, the average Mexican aca demic has greater access to supercomputers than a British counterpart, accord ing to a new report from the Parliamentary Office of Science and Technology. Nature magazine reports the three British academic supercomputing canters are already at capacity. Upcom ing environmental projects threaten to overwhelm U.K. systems. One of the penalties of the situation is that an increasing number of British academics are turning to U.S. supercomputing canters where they must make the results of their work avail able to all U.S. researchers. POETIC JUSTICE . . As if tax season pressures weren't .
enough, the In ternal Revenue Service is also dodging Congres sional arrows over the $8 billion, 10year Tax System Modernization (TSM). Congress has been quite critical of the way the IRS has managed and awarded contracts for TSM-an auto mation project racked by delays and arguments. A recent House report states "the persistence of signifi cant management problems in vital areas (of this proj ect) is a warning sign that cannot be ignored." TOO MUCH OF A GOOD THING .
. Another U.S. agency catching heat over poor management of a multi million dollar computer project is the National Insti tutes of Health. Officials at the NIH have been charged with gross overspending for equipment that was not nec essary or even desired. The General Accounting Office reports the NIH installed an elaborate IBM mainframe based computer system worth over $800 million over the last few years. NIH scientists were reportedly never consulted regarding their computer needs. If they had been, the GAO claims the majority of scientists would have been happy with a PC and have no need for the excessive automation effort. .
9
BY LAND AND SEA .
. .A $3,000 voice-activated car navigation system from Sanyo is helping Japanese drivers get where they're going. The driver com mands the system to zoom in and out of a map displayed on a 4-inch color screen sit ting on a stalk next to the driver's seat. The satellite navigation system, which the driver must program before each trip, consists of a series of map overlays stored on a CD-ROM installed in the trunk of the car. Sanyo says the system will be available in the U. S. by year-end. The race for faster sailing performance has also been enhanced by computeriza tion for the past few years. Digital Equipment Corp.'s America's Cup software pro gram has been computing data on sail shape, sail trim, rig stress, wind and overall boat performance during the race. A Digital spokes person told the New York Times that he foresees a day in the near future when competitors will be able to design a boat on a com puter, optimize it on com puter, build the boat and win the race. EC BIDDING .
. A newly adopted directive from the European Community will allow firms to bid across EC borders on telecommunica tions service contracts as of January 1, 1993. The ruling will affect all service tenders worth more than $260,000. Legislation will require the public author ities tendering service con tracts to follow open and non-
10
.
discriminatory procedures. STATUS REPORT . . . For eign-owned and foreign affiliated manufacturing plants in the U.S. use advanced manufacturing technologies such as CAD/CAM and invest more heavily in new equipment than domestic plants do. New research findings from the Rensselaer School of Management also found that new machinery and equipment investment per employee is higher in . foreign-owned and -affiliated plants. Indeed, U. S. manufacturing managers are less likely to adopt advanced technologies or to match the investment pat terns of domestic com petitors that have foreign connections. FBI JOINS CONSORTIUlVI . . .
The Federal Bureau of Investigation has joined The Microelectronics and Com puter Technology Corpora tion (MCC) as an associate member. An MCC spokes person explained that advanced microelectronics and computer designs are expected to play a signifi cant role in the technical support required by FBI HQ and field offices throughout the decade.
OLD NEWS . . . The com puter industry didn't fare any better than most businesses in 1991. In fact, worldwide sales revenues dropped for the first time in eight years. According to the Dataquest Inc.'s annual review, computer industry revenues dropped over 7% last year from 1990's $ll9 billion to $109. 7 billion. The only growth segment was the workstation industry which enjoyed an $8.8 billion year-an 18% increase. IBM is still the biggest PC maker, but its share shrunk by over 14% . Meanwhile, Apple's price slashing strategy paid off with a 10 .7% jump in its marketshare.
DIGESTIVE TRACKS .
. . It seems technology is easier to swallow when you have a real taste for it. Take the beaver, for instance . Just last month a hungry one dined on a Virginia-based fiber optic cable and sent a Washington, D.C. phone system to a grinding halt. Weeks earlier, USA Today reported a Kansas City air traffic control center was paralyzed for four hours when a beaver gnawed through a communications cable. These toothy critters join a growing legion of technological taste-testers which include the rat, whose antics shut down the com puter system at the National Association of Securities Dealers, and the sharkfest that forced AT&T to come up with an underwater repellant for its trans atlantic cable. 13
Ma M ; but this can b e fixed efficiently as follows: = M - ij; while j � L do begin S[Jl : = O; s(j + ijl : = O; j end if j � M then S[Jl : = O;
L Paul Pritchard
Department of Computer Science St. Lucia Queensland 4067 Australia
:=j
Let us further double the speed of Luo's sieve algorithm. The most time consuming part of the algorithm is:
Richard E. James III
3705 Eastwood Circle Santa Clara, CA 95054 �
while j � M do begin S(j] : = O; j : = + ij; ij : = t - ij; end
References l. Pritchard, P. A sublinear additive sieve for finding prime numbers. Commun. ACM 24, 3 (Jan. 198 1 ), 1 823. 2 . Pritchard, P. Fast compact prime number sieves (among others). j. Algorithms 4 , 4 (Dec. 1 983), 332-344.
The following replacement goes about twice as fast because it clears two elements of S in each step, still with only three statements inside the loop:
Technical Correspondence on Xuedong Luo's article first appeared in the Nov. 1 989 Communications, p. 1 367.
while j � M do begin
Wide browsing, precise selection Search by keyword, title, author, publisher, date, reviewer, ACM Category System, and more. Every word of full-text reviews is indexed and searchable. Page images of reviews allow viewing and printing of formulas and tables. ACM-pioneer in comput ing information-backs Computing Archive with superior customer support. CALL NOW (212) 869-7440 Ext. 246 (Sales Dept.) or FAX YOUR ORDER (212) 869-0481
1.D-Enter ----- -Subscription - - - today. --, YES! myComputing Archive Charter I save $50 off the regular Premier Edition price . I receive all annual updates automatically a t the 5% Charter Discount Rate and can cancel my standing order any time after the Premier Edition. 0 ACM member, $799" 0 nonmember, $999"
D
Send me the Premier Edition only. 0 nonmember, $1049" D ACM member, 5849"
0 Payment enclosed (payable to ACM). Bill my D Amex 0 MasterCard D Visa
Track down the most recent 10 years of scholarly, professional, and trade computing literature with Computing Archive: Bibliography nnd Reviews from ACM. Access more than 1 30,000 records on one disc with a PC-based search system you can master in minutes. Co111p11ting professionals, professors, academic n11d industry researchers, students, information professionals: Perform research at your own
pace; don't rush to complete searches online as charges soar... don't struggle with volumes of printed indexes.
Acd. No.
0
Exp. 0.Jte
Sign.Jturc (wquirt.>d to process your order)
I'd like more information. Have ACM's representative call me.
Telephone
ACM ml!mber # (n.-quirl'CI for member price)
Ext.
Name
Title
Org.mi:.-.Jtion Addrl'SS City
�
Zll'
St.Jte
·single use prrces only Network prrc1ng available
L�� Wcs�nd St�, NY. NY��
XT/AT or 100% compatible with 512KB memory. 1 lloppy arive (360KBJ:
+ t;
Furthermore, note that t = 2 S [i] and this is probably easier to com pute than 2k + 6i.
Comprehensive literature searches Whether you're hunting for a specific article or all works of an author, compiling a bibliography of an emerging area, or retracing developments in a research frontier, Computing Archive encom passes every subject: hardware to software, data encryption to computer graphics, with 1 8,000 book citations, more than 75,000 citations of journal articles and conference papers, and citations from 475 scholarly and industry periodicals ... plus 9,000 reviews.
Minimum configuration IBM
+ t;
ACM CD-ROM Publications
1 CD·AOM drive w1lh in1erface 10 your
_ _ _
__:'�A:'.:'.:3� _J
PC. PC DOS or MSOOS version 30 or higher
C
0 M P U T
N G
P
E R S P E C T
V E S
Charles Babbage- The Great uncle of computing?
On December 26, 1791, just over 200 years ago, Charles Babbage was born, the son of a rising banker, at his father's home not far from London's famous inn, the Elephant and Castle. Three months earlier Michael Faraday, son of a black smith, had been born a few hundred yards away. While Faraday's research work on electromagnetism and other subjects singled him out in his own lifetime as one of the most original and produc tive experimental physicists of the nineteeth Century, Charles Babbage was not in the same class. His present fame as an ancestral figure in the history of cornputers is a result of the spectacular success of the digital computer in our own day. This fame, however, is well deserved. Although his practical achievements were limited, he was the first to clearly see what m ight be achieved by the mechanization of computational processes . Babbage had no immediate suc cessor and it was not until the com pletion of the first working digital computers in the middle of the pres ent century that people began to take a renewed interest in his ideas. That there is no direct line of development from Babbage to the present day has led Doran Swade, curator of the Computing and Control collections at the Science Museum in London, to suggest that Babbage should not be regarded as the father or grand father of computing-he is more like a great uncle.
COMMUNICATIONS OF THE ACM/March 1992/Vol.35, No.3
The Difference Engine
B abbage ' s first idea was not a very good one, though he would never adm it it. I t was for a machine he called his "Difference Engine," and which he claimed would enable mathe matical tables to be computed mechanically by steam, as he once put it. The only thing a difference engine can actually do, however, is to compute and tabulate polynomials. It is true that any mathematical func tion can be represented piece-wise by a series of polynomials, but Babbage would have had to do a great deal of paper-and-pencil computation in order to calculate the coefficients to be set on the engine. Therefore, it was more than a lit tle misleading to say that a difference engine could compute mathematical tables. Babbage's defense would have been that he was concerned not so much with reducing human labor as with eliminating human mistakes, and that the Difference Engine would mechanize just those parts of the work in which human errors were particularly likely to occur. Babbage presented his proposals to the British Government in 1823 and it agreed to finance the construe-
Maurice V. Wilkes
tion of the Difference Engine. As accustomed as we are to the idea of governments providing financial support to scientists and inventors, it is easy to overlook that such a grant was almost unprecedented at that time. All went well until 1832 when Babbage had a disagreement with his engineer and work stopped. Things drifted on until 1842 when the government had to decide whether to continue or terminate the project. By then some $80,000 - a significant sum in currency of the time-had been expended and it appeared likely that as much a gain would be required. The government cannot be accused of taking the matter lightly. On the contrary, it made wide con sultations both directly and through the Royal Society. An important role was played by G.B. Airy, Astronomer Royal and head of the Royal Obser vatory at Greenwich. This was the only true scientific establishment possessed by the British government at the time, and Airy was quite used to acting as scientific adviser to the
15
CP
Mau rice v. Wilkes Cleftl and Allan Brom ley, a noted student of Babbage's work. paid a visit to Babbage's grave in London's Kensal Green cemetery on December 26, 1 99 1 . The date marked the 200th anniversary of Babbage's birth.
government in a variety of roles that had little to do with astronomy. Airy was hard-working and his judgment was sound. He may be regarded as the prototype of the senior government scientist of today. After going into the matter very thoroughly and making his own enquiries, Airy advised the government against sinking further money into the Difference Engine and his advice was taken. Babbage was infuriated by the decision, and from that time on had a pronounced chip on his shoulder against the British government and all its works. This is evident in his writings, particularly in his autobi ography published toward the end of his life. Many modern writers have taken what Babbage said at face value and charged the British government of the day with crimi nal shortsightedness. One frequently heard charge can
16
be disposed o f with confidence. In declining to put more money into Babbage's Difference Engine, the government did not delay the de velopment of the digital. computer. The Difference Engine was a spe cial-purpose device for aiding in the computation of mathematical tables, and was in no sense the an cestor of the modern digital com puter. Personally, I do not see how the decision eventually made by the British government can be faulted. It was not true that the Difference Engine was going to revolutionize the making of mathematical tables. This was shown clearly enough some I 0 years later when Georg and Edvard Scheutz, two Swedish engineers, built a successful differ ence engine. Their information was obtained from an article describing Babbage's ideas that had been pub lished in a technical magazine some years earlier. The Scheutz engine did what it was supposed to and was purchased for the Dudley Observa tory in Albany, N.Y. A copy was made for the Registrar General's office in London, an office which corresponds to the U.S. Census
Bureau. The Registrar General's office published a life table printed in part from plates prepared from stereo molds produced by the Dif ference Engine. However, in spite of the personal interest of William Farr who was responsible for the life table, the engine soon dropped out of use. The engine at the Dud ley Observatory was used to com pute a refraction table, but it too faded from use. The exact reasons are not re corded in either case. No doubt, one explanation was that the en gines were not very reliable. They were, I suspect, apt to make occa sional random mistakes. Since Bab bage had stressed so strongly the supposed inability of machines to make mistakes, this was a serious matter. However, it cannot be the whole story. If the engines were imperfect, steps could have been taken to improve them. The truth must be that the Scheutz engines were not seen as heralding a major revolution in the making of tables. The Analytlcal Engine
If Babbage had stopped at his Dif ference Engine, little more would have been heard of him. Instead, he moved on to the "Analytical Engine" which, had it been built, would have been a true general purpose computer (not a stored program computer in the modern sense, but more like the Harvard Mark I). Here Babbage showed vi sion and insight verging on genius. A good deal was published in Babbage's lifetime about this vision and about what the Analytical En gine might do and how it might be used. Information about the me chanical contrivances that Babbage conceived for its implementation remained hidden in his manuscript notebooks. When I examined these in detail in preparation for a lecture to be delivered on the centennial of his death in 1 870, I was taken aback by the richness of what I found. He discusses the design of adding mechanisms with fast-carry propa-
Macch 1992/Vol.35, No.3/COMMUNICATIONS OF THE ACM
HP asks the key software development question:
How do you stay focused in a
complex object-oriented world?
fi'. t.
The HP answer. C++ SoftBench. Sometimes, in the world of object orientecl development, the distractions can be overwhelming. But leave it to Hewlett-Packard to find an easier way to manage the complex, time-consuming tasks that so often slow you clown. C + + SoftBench is an open, integrated development environment. It allows you to stay focused on your project by reducing the number of details that demand your attention. Included in C + + SoftBench is a leading edge toolset with everything from a static analyzer to a debugger especially suited for object-oriented development. There are even tools for graphical browsing and editing of class libraries, as well as automatic source code template generation. But even the most extensive toolset isn't always enough. So C++ SoftBench gives you the ability to expand your working environment. New tools can easily be incorporated, allowing you to customize C + + SoftBench for your specific processes. The C + + SoftBe n c h Motif user interface maintains a consistent look and feel across all tools. You have the ability to program in C as well as C + +. And, of course, C + + SoftBench runs on the top UNIX® platforms, including HP and Sun. That's why C + + SoftBench is a key element of the HP CASEclge Program. To see why it should be a part of
yours, call 1-800-752- 0900, Ext. 2876.
H EW L E T T PAC KA R D
© 1991 Hewlelt-Packard Com1nny
UNIX is a rcgisl.crcd trademark of UNIX System Laborat.orfos Inc. in the U.S.A. and other counlrit•s. CSSEOOa.
CP gation, scales of notation, control systems, and much more. I f Bab bage had published this material, it would have been of great interest in the period before electronic devices displaced mechanical ones. Babbage started working on the mechanical design of the Analytical Engine in 1 834 and continued, with some intermission, for the rest of his life. Even in his later years he was making progress in the direc tions of greater simplicity and greater efficiency. He began a se ries of successful experiments with an early form of die casting. One can understand that he was disin clined to publish while he felt he was still making progress. lt is sur prising, however, that he did not publish an early sketch written in 1 837 which bears all the signs of being intended for publication. It was written before Babbage had become immersed in detail and gives a very clear outline of what the Analytical Engine would have been like. Babbage proposed to control his machine by means of punched cards strung together as in an auto matic loom and behaving very much like punched paper tape. However, anyone who expects to find that Babbage anticipated mod ern ideas in detail will be disap pointed. For example, he had independently stepped card mech anisms to control the passage of variables to and from the store and to control the mill, as he called his arithmetical unit. He never arrived at the idea, so obvious to us, of asso ciating an operation code perma nently with the name of a variable on the same card. Nevertheless, he had the concept of loops in the pro gram, and he recognized that a set of cards for his engine would con stitute a representation of an algo rithm. For lower-level control, Babbage devised a system very similar to modern microprogramming with rotating barrels carrying projecting studs taking the place of the read only memory. The studs controlled
the action of the various parts of the machine and could also cause the barrel to step forward or back ward by any desired amount. Bab bage prepared notations for the exe cution of a variety of algorithms and many of these still exist. Their resemblance to modern micropro grams is very close and shows that Babbage thoroughly understood what we would call the art of micro programming. By contrast, Babbage is far from clear about how programs would be written at the user level. It is possi ble that he did not make any clear distinction in his mind between user-level programming and mi croprogramming. However, one would have expected to find some indication as to what operations would be available to the program mer at user level. In other words, one would have expected there to be some reference to an instruction set and some evidence that he had considered how it should be com posed. Nowhere in the writings that I have seen does he even touch on the matter. For me, this remains a puzzling feature of his work. Not surprisingly, Babbage had some difficulty in communicating to his contemporaries the full breadth of his vision. Toward the end of his life, he wrote that if any man "unwarned by my example" should ever succeed in building an analytical engine, "I have no fear of leaving my reputation in his charge, for he alone will be fully able to appreciate the nature of my efforts and the value of their results." Truer words than these have never been written. It was only when the first digital computers had come into action that the extent of Bab bage's genius became fully appreci ated.
COMMUNICATIONS OF THE ACM/March 1992/Vol.35, No.3
Babbage's Private Life and Personality
Babbage's intense emotional in volvement with the Difference En gine started early. In February 1 829 William Whewell-who was two years junior to Babbage and
was himself destined to achieve dis tinction-writes in a letter to a friend that Babbage's "anxiety about the success and fame of his machine is quite devouring and unhappy." He never admitted that the idea was not as good as he thought it was, and his feeling of resentment against the scientific establishment of the day amounted at times almost to a persecution mania. This feeling did not, how ever, flow over into his social life which was a great success. One senses a certain awkwardness in his dealings with people, but he must also have been a man of great charm. He knew everyone and went everywhere. Carriages crowded the street outside his house when his Saturday evening parties were in progress. Full and successful as his social life was, Babbage was lonely in his private life. After 1 3 years of very happy married life, he had the mis fortune to lose his wife and two of his children within a short time of each other, having lost his father not long before. It took him a long time to recover from this blow. Of his eight children, only three sons survived into adult life. A daughter of whom he was very fond died at the age of I 7. All three sons fol lowed careers overseas. His eldest son, Herschel, went to work on the Continent with Brunel and then migrated permanently to South Australia, where he became an important figure in the colony. The youngest son went into the military service of the East India Company. The middle son, Dugald, was not as successful in life as his brothers. He also went to South Australia. He was somewhat disapproved of by his brother Herschel for mixing with his social inferiors and being too fond, so Herschel said, of a glass of wine. I find my heart warming toward Dugald. To me he seems the most human member of that family of earnest Victorians. a Maurice V. Wilkes . received the ACM Tu.ring Award in 1 967 and is the author of Memoirs of a Computer Pioneer, MIT Press, 1 985.
21
II F
THE ASSOCIATION FOR COMJl'.��� .t:l. � . �� �HIN ERY PROUDLY PRESENTS
TH E C O M PUTER M U SEU M 'S 1 992 C O M PUTER BOWL
R
I
D
A Y
N
T
M
A Y
1
,
1
9
1 E HASSLE IN THE CAST A
O
H
E
T
R
E
C
H
EAST
CHARLIE "JOHANN SEBASTIAN"
N
I
vs
C
A
L
K
N
O
C
K
W EST
O
U
9
2
T
JOHN F. "FUTURE"
C H M A N vs SHOC
Captain, Asset Mana111m1nt Campany
captain, Bachman Information Systems, Inc.
B I L L "THE ELBOW"
JEFF "THE KILLER"
MACH RO N E
KA LB
MasPar Computer Corporation
ZIU·Oavls Publishing Company
D R . DAVI D L.
RUTHANN
l ��lF' NR��N � 1T�f M b�W' �m��l��
Alex. Brown & Sons
Fluent Machln11 Inc.
ANDY "M.C. "
RAPPAPORT
THE
PAUL "BEARER"
EXA M I N E R
S EVE R I N O
B I LL GATES
Microsoft Corporation
The Technology Re11an:h Group, Inc.
W1/lf/11t Communications, Inc.
vs
vs
D R . JOHN E. "KNOCK KNOCK"
VERN "THE ACE"
RAB U RN
Slate Corporation
•
.
. . . • •
.
. . . . . . • • .
.
•
WARN O C K
Adobe Systems, Incorporated
XEROX PALO ALTO RESEARCH CENTER, CALIFORNIA
This is the rematch fans have been waiti ng for all year. Ever since the last action-packed Technical Knockout, the West has been clamoring for another go at the Champions from the East. On May 1 , they'll get their chance to lace 'em up. Join us at live at ringside or at the closed circuit site for all the blow by blow excitement of The Hassle In The Castle The Computer Museum's 4th Annual Computer Bowl. It's sure to be Another Technical Knockout.
-
Presenter
The Association for Computing Machinery
Underwriters
Apple Computer, I nc. Digital Equipment Corporation
Judges
I ntel Corporation
Robertson, Stephens & Co.
The lnternetworking Company
The Computer Bowl is b road cast on the PBS series, Computer Chron icles, hosted by Stewart Che ifet. The C o m p uter Bowl is a project to benefit the educational programs of The Computer Museum, 300 Congress St., Boston, MA 0221 0 . For tickets and spons o r s h i p i n fo rmat i o n 61 7- 426-2800 ext. 346.
Promotion
Engraving
Pamela McCorduck
The Microprocessor
Author
MasPar Computer Corporation
Stratus Computer, I nc.
The Massively Parallel ComputerCompany
The Transaction Processor
Merrill, Pickard, Anderson & Eyre
Visix Software I nc.
The Venture Capital Rrm
Workstation Software Co.
Official S�onsors
Price Waterhouse
Wellfleet Communications, Inc.
The Bank
Radius Inc.
Heidi Roizen T/Maker Company
Founders
Pat Collins Nelson & Dr. David L. Nelson
Bank of Boston
BASF Information Systems The Diskette
The Accounting Firm The Systems Enhancement Company
The Investment Bank
The High Performance
PARTNERS & Simons
HK Graphics
A C M -TH E FI R S T S O C I ETY I N C O M P U T I N G-I S P R O U D T C S P O N S O R
The Machine That [hanged The World a
n Monday eveninCJ. April 6.
1992 at 9:00 p.m. E.S.T.. and on successive Mondays until May 4.
PBS will present The Machine Thal
[hanged the World, 5 pro[jrams on
the history of the electronic computer and its impact on society.
P
roduced by WBGH Boston
!makers of NOVA]. and the British Broadcastin(j Corporation. and
with major funding provided by ACM and Unisys. the series hiCJhliCJhts the
fifty year revolution in computin(j and information technoloCJY-a revolution
E N I A C PHOTOGRAPH C O U RTESY THE COMPUTER M U S E U M
that is still going on.
B
Check your local PBS listings for broadcast times on the following Monday evenings: eCJinninCj with World War II
• April
research and the ENIAC. which was
6-"liianl Brains." covers the wartime events that led to the 1946
debut of ENIAC. the world's first general purpose electronic computer.
co-invented by J. Presper Eckert and • April 13-"Jnventing the Future" examines how the computer rose from
the late John Mauchly [a founder of
obscurity to become the engine that powers business throughout the world.
ACM]. The Machine Thal
• April 20-"The Paperback [ompuler" explores how computers became
[hanged The World follows the
small. affordable and easy to use.
unpredictable course of information technoloCJY from the room sized. data
• April 27-"The Thinking Machine" focuses on the most ambitious goal of
all-creating a computer that will vie with humans in intelligence.
processinCJ centers of the 1960's to desktop personal computers of the
• May 4-"The World al Your Fingertips" looks at the social revolution
wrought by computers-and the price we pay.
19BO's to virtual reality of the 1990's. describing events that have altered
Tune in and celebrate the exciting history of our industry and our profession.
society in profound and totally unexpected ways.
A DVA N C • N G
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . • . • • . . • • • • • • • • . . . . . . . . . . . . . . •
H U MAN
C A PA B • U T • E S
•
THRO UGH
• N F D R M An O N
TE C H N O L O G Y
F u n d l r1g !or The Machine. Thal l:hanqed The World has been provided b y Unisys. lhe National Science Faundatian a nd the 90.000 members of t he ACM.
YES, I WANT TO ATTEND DB/EXPO '92! Please send me: 0
FREE EXHIBIT PASSES
0 Information on conference sessions and early registration DISCOUNTS
P
E R S O N A L
C
O M P U T
N G
Dynabook Revisited Portable computers Past, Present and Future �,���'.:?£;?"
Much of the foundation of modern personal computing was laid over 20 years ago, when Alan Kay and his colleagues at the Xerox Palo Alto Research Center designed the Dyna book, a prototype portable personal computer (see Figure 1). The name "Dynabook" suggested that it would be portable (like a book), and that it was a new, dynamic medium with aspects of paper and pencil, paint, typewriter, musical instrument, and computer simulation. The first commercially impor tant1 portables, the Osborne 1 and the Radio Shack Model 100, had dif ferent goals as shown in Figure 2 . The Osborne was a portable approxi mation of the desktop cornputer of its day. It had the same C PU, bus, and operating system, and ran the same software as its desktop counterpart. The Model 100 was a complemen tary computer-an input device used in conj unction with a desktop ma chine. It had a text editor in ROM , and was widely used by journalists who took notes and drafted articles that were finished on a desktop computer. Today's Portables: Desktop and complementary
get by with the Zeos p o rt able , which has 5MB of memory and a 60MB disk? I ran a few benchmarks corn p a r m g t h e Zeos to AST and Goldstar desktop machines with the same CPU (16 M Hz 386SX) and memory size. As shown in Table 1, the speeds are com parable, with the portable losing some ground in display speed. Based on speed alone, the portable could replace many desktop computers. The major drawbacks to the port able are its input and output devices. The keyboard is somewhat smaller than standard, slowing down typing. This is compounded by the lack of dedicated scrolling, and asterisk keys found on the desktop machines. The portable display is also slower, lacks color, and is more difficult to read, especially when it changes rapidly (for example when moving a mouse). The portable also has no expansion slots, but with a modem and serial and parallel ports, it would satisfy many needs. If these drawbacks dissuade you, a compromise is possible. You could buy a portable computer, plus a stan-
Kay's first goal, portability in a form factor similar to the original mockup, has been achieved. Could a portable
COMMUNICATIONS OF THE ACM/ March 1992/Vol . 3 5 , No.3
Larry Press
1. Original ovnabook Mockup. While Kays goal was a machine this size, the first version of the oynabook, the "interim" oynabook, was built using a desk-sized workstation. In 1975, Kay felt hardware for a porta b l e oynabook would b e "inevitably avail· able in ten years time" !5l. That predic tion remains valid today.
dard keyboard and color monitor. When at your desk, the portable would be closed and plugged into the monitor and desktop keyboard. (Note that the Zeos is faster with an external display than with its LCD.) This combination would cost more than the equivalent desktop com puter, but not as much as a desktop computer plus a portable. If your needs demand a portable, and your budget is limited, consider this hybrid option. The Zeos portable can run Win dows, and generally substitute for a low-end desktop computer, but what about complementary computers, the descendants of the Model 100? ' There were earlier portables like the I B M 50xx and t h e Sony Typecorder, b u t the Osborne 1 and Model 100 were mass-marketed commercial successes.
25
PC There are two interesting options On the other hand, the TI display is hand-held computers and previous 5- 1 0% faster, and the machine out generation portables. performs many LAN servers of the There are several hand-held mid- l 980s. At 4.4 pounds, and 8 1/2 computers that run standard DOS x 1 1 x 1 .4 inches, it is easy to carry, applications. The first on the mar but its battery life is much less than ket was the Poqet computer. The the Poqet, which runs 1 00 hours on Poqet is 1 x 8% x 4 1/. inches and two AA batteries. The Travelmate is easier to use weighs only 1 .2 pounds. It runs for days on AA batteries. It has 5 1 2 KB than the Poqet. The keyboard is as of RAM, and DOS, BASIC, editing, easy to use as the Zeos, and the dis calendar, name and address, calcu play as easy to read. The Travel lator and terminal emulation soft mate is fine for character-based ware in ROM. Instead of disks, the applications such as word proces Poqet has two slots for ROM or sors, compilers, spreadsheets, and RAM cards, and ROM-card ver terminal emulators, but .you will sions of 1 23 and other programs have to leave your graphic user in are available. There is also a serial terface at home on the desk. Per port and a bus-extension port for haps the best thing about the adding an optional disk drive. The Travelmate is its current price keyboard has a standard typewriter under $800. Unless you need the layout, and by the standards of convenience of a hand-held com hand-held computers, it is large. puter, obsolete portables are a bar The display has 25 lines of 80 char gain. acters and 640 x 200 resolution in Tomorrow's system: graphic (CGA) mode. Since it is a standard PC, you can Dynabook plus Dynabase load your favorite software into a Alan Kay's dynamic medium was Poqet RAM card, and leave your intended for children, but I am less office. I took a Poqet to several lec ambitious. I would settle for a port tures, and discovered that it is pos able that helped me with profes sible to touch type on the keyboard, sional communication as well as but typing is slow and error-prone. with the researching, planning and 2 To make matters worse, the display writing of CACM columns . Since desktop computers will is not backlit, and I frequently found myself holding it near my always offer more performance face or maneuvering it to find suit .and higher bandwidth connectivity able light. I did not think the ergo than portables of the same price, I nomic problems were worth the want a portable that complements convenience of machine-readable my desktop system. It will use voice, notes, so I still use pen and paper, a pen/display, and perhaps o. cam but it is a close call between the two era for input and output. Storage and others might argue for its mer will probably be nonrotating, and it will be able to receive wireless com its. An alternative is to buy a pre munication. Most important, it will vious-generation portable. Con be part of a system that includes a sider, for instance, the Texas In desktop computer and local- and struments Travelmate 2000. (The wide-area networks. These systems Sharp PC-6220 and the Com must be designed to work together, puAdd Companion are identical). with functions placed on the appro A few years ago, this was the hottest priate machine. This integration is thing in portables, with a 1 2 M H z illustrated by some of the following 80286, I t o 3MB of memory and a 2The vision of a professional communication 20MB or 40MB hard disk. I t is 3 1 % medium is older than Kay's Dynabook. It slower than the Zeos on non-float dates to Vannevar Bush's 1 945 article [2] speculating on a hypothetical machine to help benchmarks, 44% slower on floating with his scientific reading, writing, and com point, and the disk is 85% slower. munication.
26
examples of system utilization. I write many notes-at lectures, meetings, and the gym . M y notetaking medium i s pen on 8 1/2 x I r sheets of paper. It is an excellent recording medium, well suited for text and sketches. It is terrible, however, for storage and retrieval. Most of these sheets of paper make it back to my office, where they are sorted into various files and piles, and I am afraid few are ever seen again. This is where the desktop system comes into play. It will be the data base machine that stores and re trieves notes taken on the portable, making them part of a lifelong, professional Dynabase. It will also do handwriting and voice recogni tion on many of the notes. It will be many years before a portable com puter comes close to the conven ience of paper for recording notes. I will switch to a portable before that because of the tools running on my desktop computer. I also use sheets of paper for de signing things such as CACM col umns, programs, courses, talks, or house floor plans. These are similar to the other notes, but are more likely to use varying data types like paint-images, draw-images, tables, spreadsheets, symbolic and nu meric equations, and outlines. A portable computer should have dif ferent types of "paper" for differ ent types of data, and the desktop machine should know about them too. While I use a lot of 8 1/2 x 1 1 paper, my favorite note-taking medium is in the margins of books and articles. Sad to say, these notes are even less likely to be retrieved. Writing-while-reading is a very important application. Perhaps the best measure of the value of a book is the number of marginal notes scrawled on it-reading is conver sation. Slate Corporation has an nounced Pen Book, a program that appears to be suited to this applica tion. Books are portable and easy on the eyes, but I would be willing to
March 1992 /Vol.35, No.3/COMMUNICATIONS OF THE ACM
PC trade some of those advantages for the ability to capture marginal notes, and add them to my Dynabase. Marginal notes would be stored along with lecture and de sign notes, and they would be asso ciated with the relevant document and scanned excerpts. They can also be electronically mailed to the author, closing the reading-as conversation loop. Since we increasingly read elec tronic as opposed to printed mate rial, a portable computer also needs enough storage to carry significant chunks of the desktop away from the office. For example, I would like to leave the office with a copy of a book or a paper and all the papers to which it refers. Sony's Data Diskman foreshadows this capability, but it is limited by its use of CD-ROM media. I see the portable as a dynamic inputloutput medium used in con junction with a desktop machine that will maintain the Dynabase. Because of its unique capabilities, the portable will also often be used in the office. Finally, information from other sources such as net works or CD-ROM will also be stored in the Dynabase. The Technology
It will take many years to get a sys tem such as the one just described, but we can look at what is available today and in the near future. There are several technologies underlying the portable, but the hard part is not the portable; it is the Dynabase. The portable will have an inte grated digitizer-display for pen inputloutput. Pen-based computers have been available for several years with little impact, but interest in them is high. Yourdon [8] states "I have seen the future and it is spelled P-E-N." He feels we are in "the formative stages of a new era in computing: the age of pen com puting. Approximately a dozen companies showed pen-based products or prototypes at Comdex this year. There are major system software efforts at GO and Micro-
COMMUNICATIONS OF THE ACM/March 1992/Vol.35, No.3
� : 'I •1
"-" ·- -
1
.�- -
T
;--
.
•
.
-- .
-
"
•
.
2. The Osborne 1 and the Radio Shack Model 100. The Osborne was a portable version of a desktop computer; whereas the Model 100 was a complementary adjunct to a desktop computer.
Table , _
Desktop and Portable Benchmarks (B + {L}) and ({L} + C) ---'> D infer (A U C) ---'> (B U D). We can make a cut inference from two clauses if any only if there is some atom L which is in the ante cedent of one clause and the conse quent of the other. To form the conclusion of the inference, we first 'cut' out L from both places, and then merge the two antecedents into one and two consequents into one. The 'disjoint union' notation X + Y denotes the union X U Y, but also carries the further infor mation that x n y = 0. Example I. From the clauses A B ---'> C D and D E ---'> F G we can infer the clause A B E ---'> C F G by a
cut, eliminating the atom D. The resolution inference pattern generalizes the cut inference pat tern by bringing in unification. The resolution inference pattern has the form : from
A -'> (B + M) and (N + C) ---'> D (A U C)u ---'> (B U D)u infer where u is an mgu of the one part partition {M U N}.
{F(F(F(F(F(F(F(F(x)))))))), F(F(F(F(F(F(F( F(A))))))))} we can merge only one pair of nodes, and generate only one new link, at each iteration of the loop. These successive minimal modifica tions of the graph therefore com prise essentially a sequential pro cess. However, such 'worst cases' are more pathological than typical, and experience suggests that they are rarely met in real applications. Resolution
Once we can compute an mgu for any unifiable partition of a set of expressions (or show the partition not to be unifiable, if that is the
In making a resolution inference, we must first use unification to de duce a pair of instances of the two premises suitable for a cut to be applied. In the special case that M = N = {L}, the mgu of the parti tion {M U N} is the identity substi tution. So in this case, a resolution is the same as a cut. Example 2 . From ---'> P (G(r s) r s) and P(x y u)P(y z v)P(x v w) ---'> P(u z w) we infer P(r z r) ---'> P(s z s) by a resolution in which M = {P(G(r s) r
COMMUNICATIONS OF THE ACM/March 1992/Vol.35. No.3
s)} and N = {P(x y u) ,P(x v w)}, since {M U N} is unifiable with mgu {x = G(r s), y = v = r, u = w = s}.
Example 3 . From P(x y u)P(y z
v)P(x v w) ---'> P(u z w) and P(a b c)P(b d e)P(c d f) ---'> P(a e f) we infer P(x y a)P(y b v)P(x v c)P(b d e)P(c d f) ---'> P(a e f) by a resolution in which M = {P(u z w)} and N = {P(a b c)}, since {M U N} is unifiable with mgu {u = a, z = b, w = c}. From two given clauses, only a fi nite number of clauses can be in ferred by resolution-one for each choice of the 'cut' sets M and N for which the partition {M U N} is uni fiable. If there are no such choices of M and N, then nothing can be inferred from the two clauses by resolution. Resolution Deductions and Proofs
A resolution deduction is a finite tree whose nodes are labeled by clauses, each nonleaf node being labeled by a clause which is inferred by a reso lution inference from the clauses labeling its immediate successors. The conclusion of the deduction is the clause labeling its root, and the premises of the deduction are the clauses labeling its leaves. A resolu tion proof_ is a resolution deduction whose conclusion is false ( = the empty clause). Such a proof estab lishes that the premises are contra dictory (unsatisfiable). If S is any unsatisfiable set of clauses there is always a resolution proof whose premises are all in S. This fact is the completeness of resolution (see [39]. A resolution proof with n + 1 premises can be taken in n + 1 dif ferent ways as a proof of the nega tion of one of its premises from the other n premises. For example, a resolution proof with premises A , B , C can be taken as ( 1 ) a proof of not-A from the premises B and C, (2) a proof of not-B from the prem ises A and C, and (3) a proof of not-C from the premises A and B. P1 - Resolution
A resolution one of whose two premises
51
'
•
err
wn .
Whatever your development applications with a minimum You've heard the horror stories. environment-stand alone PC,LAN number of steps. Frightening tales of how based, or cooperative, workstation And CA-PAN/LCM even provides uncontrolled change can wreak -to mainframe -CA-PAN/LCM host integration to the most tM.VJUI hav�c ii: distributed tells you the who , what, when, popular mainframe library control - appltcation development :::::::--: where and why of all changes systems, ensuring cross- � environments. . CA'scompuring / B CA-P LCM ut made to any type of application platform ynch onization AN ™ � ! Archuecture For rhe 9os. puts those demons to rest. component. and data integrity. Dial 1-800-CALL CAI '-"f It's the world's most advanced "CA-PAN/LCM represents a comprehensive solution to version control and co1!figuration today, and we'll send you your free change and configuration management, with special emphasis on management software, from demo disk and a monster of a security and data integrity. " Computer Language, August 1991 poster. But call right now. Computer Associates. It also analyzes dependencies After all, who knows "For large networked programming shops, CA-PAN/LCM wasjudged best. . . on the basis qf what evil lurks between application components OMPUTER "' multi-user security, ease qfoperation, and in the hearts of r.SSOCIA and determines how to integration. . . .. TES your programs? Software superior by design. construct, maintain and enhance PC Week, Februa1y 1991
-
&!0��
_ _
(i
© Computer Associates International. Inc .. One Computer Associates Plaza. lslandia. NY 1 1 788-700'.J. CA-PAN/LCM is based on software originally developed by Seidl Computer Engineering, Inc. All product names referenced herein are trademarks of their respective companies.
Logic Programming is unconditional is called a PI -resolu tion. Example 2 is a P i -resolution,
Hyperresolution Deductions
A hyperresolution deduction is a fi but Example 3 is not. It turns out nite tree each of whose nodes has a that whenever a set of clauses is label and each of whose nonleaf unsatisfiable, then there is a P l nodes also has a justification. The resolution proof from those prem labels are unconditional clauses, and ises (see [ 40]). In other words, P l the justifications are conditional resolution is also complete: despite clauses. The clause labeling a non its restricted form, P I -resolution is leaf node N is inferred by a hyper j ust as strong as resolution, but its resolution whose unconditional proof-space is sparser than that of premises are the clauses labeling the immediate successors of N, and unrestricted resolution. whose conditional premise is the justification of N. The conclusion of Hyper-Resolution We get an even sparser proof-space the deduction is the clause labeling when we take as the only inference its root. The premises of the deduc rule, instead of the two-premise P l tion are the labels of its leaf nodes resolution, the (p + l )-premise and the justifications of its nonleaf hyper-resolution rule in which exactly nodes. H yperresolution deductions can one of the premises is a conditional clause and all of the other p prem yield only unconditional clauses. ises, together with the conclusion, are Moreover, they can yield only posi unconditional clauses. The hyper tive unconditional clauses, unless the resolution rule is: justification of the root node is a nega-(C 1 + M 1) , . . . , (Cp + Mp), and N 1 + · · · + Np - D infer -(C 1 U· · · U C P U D)a where a is an mgu of the p-part partition {M 1 U N 1 , . . . , Mp U Np} ·
from
The unifiable p-part partition that is the essential ingredient of a hy perresolution is called its kernel. The p + l premises and the kernel together uniquely determine the conclusion. A hyperresolution inference is really a compacted reorganization of a P I -resolution deduction whose conclusion is unconditional. After the reorganization the deduction has had all of its interior nodes sup pressed and has become a single integrated transaction instead of a linked system of many transactions. By reorganizing the reasoning as a single inference, we are simply re garding its conclusion as having been obtained directly (or, to use a traditional logic expression, immedi ately -without any 'mediation') from its premises in one step, rather than 'mediately' as the even tual outcome of several linked P l resolution steps.
56
tive conditional clause and in that
case, but only in that case, the con clusion is a negative unconditional clause; indeed, it is the empty clause. Thus a hyperresolution deduction of the empty clause (a hyperresolution proof) always has exactly one negative conditional clause
among its justifications. As we shall see, it is this feature which adum brates logic programming. Completeness and Local Finiteness of the Resolution Clausal Predicate Calculi
The resolution and hyperresolu tion versions of the clausal predi cate calculus are all complete. Also, both systems are locally finite. This means that, in each system, there are only finitely many deductions of a given size (number of nodes) having a given set of premises (and this number is much smaller for hyperresolution than for resolu tion). By contrast, traditional predi cate calculi are not even locally fi-
nite. This is one reason it is so difficult to make an efficient proof procedure for tradilional predicate calculi. For example, most tradi tional predicate calculi contain the rule of specialization: from '\;/A infer '\:/(AO) , where () is any substitution.
(The sentence VS is the universal closure of the sentence S : the result of prefixing a universal quantifier to S for every free variable in S). With this inference available, there are infinitely many deductions of size 2 which have the same premise '\;/A-one for each different substi tution e. Hyperresolution and Horn Clause Logic
The advantages of hyperresolution are quite striking in the Horn clause predicate calculus. In this subsystem of the clausal predicate calculus every clause is a Horn clause, namely, a clause having at most one conclusion. Hyperresolution then becomes much simpler. Recall the general definition of hyper resolution : -(C 1 + M 1 ) , . . . , + Np - D (Cp + Mp), and N 1 + infer -(C 1 U · · U Cp U D)a , where a is an mgu of the p-part partition {M 1 U N 1 ; . ., Mp U Np} · from
· · ·
·
When all clauses are restricted to having at most one conclusion, the 'cut' sets Mi can only be singletons (say, {AJ) , and the 'remainder' sets Ci must be empty. Consequently, the definition of hyperresolutions for Horn clauses can be restated, in the following much simpler form: from
-A l > . . . , -A p , {B 1 , . . . , B p} - D
infer -Da where a is an mgu of the p-part
partition {{A1 , B 1 } , . . . , {Ap, B p}}
In this restatement of the rule, D and the A's and B's are all atomic sentences. When we combine hy perresolution inferences into mul tiinference deductions, we are in effect treating each particular ap plication of this inference pattern
March 1992 /Vol.35, No.3/COMMUNICATIONS OF THE ACM
·
as though it were a special infer ence rule, 'the {B 1 , . . . , B p} -? D inference rule', stated as:
has the same premise and the same conclusion as D, and conversely. We define ultraresolution inferences
from -7A 1 , . . . , -?A r infer -? Du where u is an mgu of the p-part
partition {{A l > B 1}, . . . , {Ap , B p}}. This is, however, just a pragmatic device to sharpen our understand ing of the very special role that con ditional Horn clauses play in logic programming. U ltraresol utions: Horn Clause Hyperresolution Deductions as Single Inferences
We again apply the idea of making a single inference out of an entire deduction. In the case of hyper resolution, instead of thinking of the conclusion of an entire deduc tion (namely a deduction built from P I -resolution steps and having an unconditional conclusion) as being arrived at stepwise by the perfor mance of each of its inferences sep arately, we think of the whole con struction as one inference step involving a higher and larger-scale inference pattern. We will now treat Horn clause hyperresolution de ductions in a similar way, and thereby arrive at a higher- and larger-scale inference pattern which we call ultraresolution. There is really no need, prag matically, to know the conclusion of every individual inference in a hy perresolution deduction, if all that we are after is the eventual conclu sion of the whole deduction. We can instead characterize that even tual conclusion more directly, by a relationship based only on the structure of the premises of the deduction. By omitting in this way all of the interior stepwise conclu sions we turn the entire hyper resolution deduction into a single in ference, which immediately yields its conclusion from the premises in one integrated step.
is a cover of A(xo) B(xo) -? C(xo) by C, in view of the assignments given by the table: atom
assigned to node
A(xo) B(xo) H(G(x2)) D(x 1 Y 1 ) E(x 1 )
2 3 4 5 6
directly, however, without refer ence to their corresponding hyperresolution deductions. and has the following partition as The ultraresolution rule is its kernel : (where A -? B is a Horn-clause and C is a set of Horn-clauses): from A -? B and C infer -?Bu if there is a cover of A -? B by C whose kernel is unifiable with mgu u. The clause A -? B is the main prem {{A(xo), A(F(x 1 Y 1 ))} , {B(xo, B(x2)}, ise and the clauses in C are the cov {E(x1 ) , E(M)}, {D(x y 1 ) , D(M N)}, 1 ering premises. {H(G(x2)), H (G(x3 ))}. covers and Their Kernels
A cover of a clause A -7 B by a set C of clauses is a certain kind of finite tree with nodes labeled by clauses. The root of the tree is labeled by A -? B, while the other nodes are labeled by variants of clauses in C. The extra condition that makes the tree a cover is that for each node N in the tree, every atom in the ante cedent of the clause labeling N is assigned to a distinct immediate successor of N. The kernel of the cover is the partition : {{X, Y}IY i s the conclusion o f the clause labelling the node to which X is assigned}. Example 4. To illustrate the no tions of a cover and its kernel, con sider the clause: A(xo) B (xo) -? C(xo) and the set C of clauses {E(xi ) D(x 1 Y 1 ) -7 A(F(x 1 Y 1 )), H (G(x2)) -7 B(x2 ) , -7 H (G(x3 )), -7 D( M N) , -7 E(M)}. The labeled tree given by the table: node parent
Ultra resolution
To every hyperresolution deduc tion D there corresponds an ultraresolution inference U which
COMMUNICATIONS O F T H E ACM/March 1992/Vol .35. No.3
I 2 3 4 5 6
I
I 3 2 2
Since this kernel is unifiable, with mgu u = {xo = x2 = X 3 X 1 = M, Y 1 = N},
=
F(M N)),
we can infer the clause -7C(x0)u = -?C(F(M N))
by an ultraresolution which has A(x0) B(xo) -? C(xo) as its main premise and C as its set of covering premises. The intuition behind the notion of a cover of a clause A -? B is that it depicts exactly the pattern of orga nization of the given clauses. If the kernel of the cover is unifiable with mgu u, it guarantees that we can easily relabel the tree so it turns into a hyperresolution deduction, from these clauses as premises, of the same unconditional clause -?Bu that the ultraresolution inference obtains directly from them in one step. In this relabeling, the new label on each leaf node of the tree is the same as the old label. The old
label A(xo) B (xo) -? C(xo) E(x 1 ) D(x 1 Y 1 ) -7 A(F(x 1 Y 1 )) H(G(x2)) -7 B (x2) -7 H(G(x3 )) -?D(M N), -?E(M)
57
Logic Programm ing --
label on each nonleaf node of the cover, however, is removed (it now becomes the justification of the hy perresolution inference at that same node), and the node's new label is the unconditional clause which is inferred by a hyperresolu tion from this j ustification clause together with the new labels on the immediate successors of the node. The following example illustrates the relationship between a hyperresolution deduction and the corresponding ultraresolution inference.
41 (2)
37
40
Example 5. Figure 6 is a hyper
resolution deduction of the uncon U NCLE(TED clause ditional ANN)- from a subset of the fol lowing set of Horn clauses, which comprises a small 'family relation ship' knowledge base. This knowl edge base contains (as its 'defini tions') the following conditional Horn clauses: 1 2 3 4 5 6 7 8 9 10
UNCLE(u x) U NCLE(u x) PARENT(x,y) B ROTHER(b x) SISTER(s x) SIBLI NG(x y) H USBAND(h w) WIFE(w h) FATHER(f x) MOTHER(m x)
FIGURE 6.
-BROTHER(u y) PARENT(y x) -HUSBAND(u s) SISTER(s p) PARENT(p x) -cH ILD(y,x) -SIBLING(b x) MALE(b) -srnLING(s x) FEMALE(s) -DI FFERENT(x y) FATHER(f x) FATHER(f y) MOTHER(m x) MOTHER(m y) -MA RRIED(h w) MALE(h) -MA RRIED(h w) FEMALE (w) -PARENT(f x) MALE(f) -PA RENT(m x) FEMALE(m)
and (as its 'facts') the following un conditional Horn-clauses: 11 12 13 14
C H I LD(JIM JOE)C H I LD(JOE MEG)CHILD(JIM SUE)C H I LD(ANN JOE)-
15 16 17 18
C H ILD(JOE TOM)- 1 9 CHI LD(TOD PAT)C H I LD(AN N SUE)- 20 CHILD(RON PAT)C H I LD(PAT MEG)- 2 1 CHILD(TOD TED)C H ILD(PAT TOM)- 22 CHILD(RON TED)-
23 24 25 26 27
MALE(J IM)MALE(JOE)MALE(TOM)MALE(TED)MALE(TOD)-
29 30 31 32 33
FEMALE(ANN)FEMALE(SUE)FEMALE(MEG)FEMALE(PAT)FEMALE(SAL)-
40 D IFFERENT(a b)- a ¥- b & a, b,
E
35 36 37 38 39
MARRIED(TOM MEG)MARRIED(JOE SUE)MARRIED(TED PAT)MARRI ED(RON SAL)MARRIED( J IM JAN)-
{JIM, JOE, TOM, TED, TOD, RON , ANN, SUE, MEG, PAT, SAL, JAN}
Premise 40 is a 'virtual' definition: it is simply a shorthand way of sup plying 1 32 facts (such as D I F FERENT(JOE A N N)-) whose predicate is DIFFERENT and whose two arguments are distinct constants in the displayed set.
58
March 1992/Vol.35, No.3/COMMUNICATIONS OF THE ACM
From this knowledge base there are, for example, hyperresolution deductions of each of the following unconditional clauses: 41 42 43 44
U NCLE(TED ANN)