Retrospect on My Research, 1999 I was peaked'

1
Retrospecton My Research, 1999(I was peaked.)

• Akinori Yonezawa
• Dept. of Information Science
• University of Tokyo

2
Why Retrospect, Now

• Invited to give a talk at Foundations on
  Object-Oriented Languages(FOOL) in Boston, Jan.
  22, 2000.
• getting old, over 50
• time to look back and project myself

3
Personal Research History

• Programming Languages (M1)
• Extensible Language ALGOL N compiler (M1)
• tracing a tree structure without using a stack
• Schorr-Waite- Wegbreit Algorithm reinvented
• Proof Procedure (M2)
• Genzten Style Sequent Calculus (1st Order)
• Herbrands decidable case refound
• Implementing a theorem prover in an assemble lang.

4
Research History (cont.)

• Visit of John McCarthy
• Stanford AI lab.
• no future for Resolution-type Proof Procedure
• Lisp Meta-circular interpreter and operational
  semantics
• Visit of Rod Burstall
• Language POP2
• getting to know Planner Language

5
History (cont-1)

• Visit of Andre Ershov
• introduction of Program Schemata
• introduction of Mixed-Computation (partial
  evaluation)
• question on difficulties in dealing with
  recursion in program schemata

6
History (cont-2)

• Visit of P.H. Winston
• introduction of T. Winograds SHRDL Natural
  Language Understanding system, MIT AI Lab
• introduction of Carl Hewitts Planner Language
• Planner Language
• controversy on KR Declarative vs.
  Procedural
• Procedural interpretation of logic

7
MIT AI Lab LCS

• Minsky School - Procedural Dogma
• Non-Declarative, Non-Logic Approach
• Procedural Interpretation of Logic
• Minskys Frame representation
• Hewitt School
• moving from Planner to Actor
• Actor Universal Modular Actor formalism for
  Knowledge Representation, POPL93

8
AI Lab LCS, My Study

• A. Meyers graduate course on Algorithms
• B. Liskovs seminar
• Structured Programming through Abstract Data
  Types
• developing CLU language
• appraisal of formal specification
• C. Hewitts seminar
• chaos!! , discussions!!, developing Actor Lang

9
My Study

• V. Pratts seminar
• developing dynamic logic
• developing Lingol
• M. Rabins seminar
• developing Randomized Algorithms Framework
• R. Rivests course
• advanced algorithms

10
My Study (cont.)

• J. Dennis course
• developing Data-flow architecture and lang.
• H. Putnams course on Gentzen logic
• R. Jacobson, Stefan Kleene, S. Kripke,...
• Classmates
• David Harel, Elliot Moss, Stan Zdonik
• Bob Scheiffler, Guy Steele, Brian Smith
• Henry Baker, Richard Stallman,..

11
My Research _at_AI Lab LCS

• AI research was too hard for me
• Language Design, Semantics and formal approach
  were easier for me finishing Ph.D.
• suggested by H. Hewitt to work on Meta-Evaluation
  (symbolic evaluation)
• inspired by B. Liskov to work on Formal
  Specification

12
Thesis Topic fixed

• Specification and Verification of Parallel
  Program Based on Message Passing Semantics, 1976
• to-do-list
• formulate parallel computation model
• -----gt ABCM model (1984)
• formal specification language
• formal verification rules

13
Computation Model

• Notion of Concurrent Object
• introducing local state in Hewitts Actor against
  the advice
• consideration on distribution of function over
  messages and concurrent objects
• what to be transmitted on a message
• message tag, values, object-id, code/program
• object should be transmitted!?
• migratory objects
• object in motion should receive messages?

14
Formal Specification Language

• Describing local state of an object
• conceptual representation, representing it by
• conceptual name, pattern of sequences, records,
  sets
• (QUEUE h !tl ), (Mailman !letter)
• (PostOffice c1, c2, c3)
• (c2 is-a (Counter q2)), (q2 is-a (CLERK))

15
Formal Spec. Lang. (cont.)

• Describing interaction of
• an object and a message
• an object and an object
• Describing caused events/effects
• changes in local states of objects
• creation of new objects
• caused message passing

16
A Good Convincing Example

• Advisor instructed to find a Good example
• Cambridge Post Office is chosen
• customers arrive at post office
• choose a post office counter
• buy stamps or do registration
• put stamps on envelops
• put them into the mail box, leave post office
  
• Mail collector collects envelops ..

17
Cambridge Post Office
18
Advices and Warning

• When thesis is completed,
• US advisor does not like my computation model, he
  has his own model, he submitted another paper
• Japanese advisor wants me to get a Japanese
  degree for my safety...

19
After Ph.D. Work (1977)

• Have to come back to Japan
• No timesharing system, no e-mail system
• No student,
• published five papers next four years,
• two from Ph.D. work
• one on program synthesis
• one on monitors and path-expressions ..

20
Created own laboratory (1983)

• One year wasted for preparing lectures
• One research associate (Mr. S) and one student
• Started designing Object-Oriented Concurrent
  Language ABCL
• Sent paper to C. Nygaard (Simula 68)
• Engineered to get invited to Berlin in 84 and to
  Paris in 85

21
ABCL Lang. Publicized

• Accepted in 1st ACM OOPSLA, 1986
• Using EC/US connections, edited a book
  Object-Oriented Concurrent Programming in
  1987, MIT Press
• Booming of Object-Oriented Programming started
• Alliance with Gul Agha, not with Hewitt

22
Reflective Computation

• Has been interested in evolutional computation,
  found Brian Smiths 3-Lisp
• Invited at IFIP congress in Dublin, 1986, visited
  Brussels and found Patie Maes Reflection work.
• Got the idea of introducing Reflection into
  Concurrent Object-Oriented Computing
• suggested T. Watanabe to design ABCL/R, published
  in OOPSLA 1988

23
High Performance Implementation

• Surrounded by high performance people at UT
• Equipped by many Parallel Machines
• S. Matsuoka, K. Taura developed marvelous
  schemes and techniques
• papers appeared in
• Supercomputing, PPoPP, PLDI, ICS

24
Inheritance Anomaly

• Avoided introducing inheritance mechanism into
  ABCL
• No good solution found, which caused to write a
  paper with J.P. Briot in 1989 for ECOOP
• S. Matsuoka ...

25
Formal Semantics of COOL

• Trinity of language research design, semantics
  and high performance implementation
• design done, high performance implementation done
• No good previous semantics
• denotational semantics criticized by D. Scott
• Agha transitional semantics not satisfied

26
Semantics (cont.)

• Overwhelmed by Imperialism of pi-calculus
• fond of and familiar with Gentzen style sequent
  calculus
• encountered with Girard Linear Logic
• reviewed a French paper on Linear concurrent
  logic programming
• suggested N. Kobayashi to use a fragment of
  Linear Logic for a new calculus

27
Semantics (cont.2)

• N. Kobayashi developed a nice framework
• Had visits of R. Milner, CAR Hoare, S. Abramsky
• Submitted to Formal Aspects of Computing,
  accepted and recommended by R. Milner

28
Partial Evaluation

• Not on my performance!!
• In 1976, I wrote a paper, Symbolic Evaluation on
  Parallel Programs based on Message Passing
  Semantics
• Semi-symbolic Evaluation is Partial Evaluation
• the crux is simplification of symbolic
  expressions

29
Mobile Objects

• Used to think about artificial dichotomy of
  objects and messages
• messages can be objects
• object can be messages
• both can move and be transmitted and persistent.
• Works on AMO (Autonomous Mobile Objects) with T.
  Sekiguchi, T. Hashimoto

30
Messages

• Lucky to encounter a pregnant idea early days.
• Lucky to encounter good students.
• Work persistently on a small number of nice
  concepts and notions!! Find them.
• Good luck.