Declarative Programming

1
Declarative Programming a JourneyHervé
GallaireXerox Chief Technology
Officerxerox.com/innovation
2
The early days

• Berkeley 1966
• Hexadecimal Programming . (B. Lampson of the
  ALTO and Xerox PARC fame later on)
• How to hate low level programming
• Automata and Grammar modeling (M. Harrison)
• Clean and beautiful
• Formal Logic computing and complexity (S. Cook)
• Introduced the theory of NP completeness
• Declarative programming for ever
• Too much?

3
Toulouse CERT/SupAero

• 1972 Logic and database- Syntex system
• Inspiration Levien and Maron Rand Corp
  1962-1966
• Syntex Relations and objects
• Syntex declarative rules and deductive
  information logic engine
• Syntex Natural Language querying
• Leading to Logic and Databases books 1978
• J. Minker
• Logic programming started 1976
• A. Colmerauer, R. Kowalski
• 1974 LAU multiprocessor system 4 processors
• Unique assignment language inspired by L.
  Tesler and H. Enea 1968
• Execution - data-driven
• Declarative - equations

4
CGE Lab de Marcoussis

• Creation of a new division
• 1982 logic and functional systems
• Prolog (and Lisp) machine - MAIA
• Tagged machine
• VLIW
• Very fast
• Logic for verification of programs

5
Munich ECRC

• Creation 1984 Bull, ICL, Siemens
• 1984-1989 Logic and declarative everywhere
• Prolog system SEPIA ECLiPSe Opium
• Constraint Logic Programming system - CHIP
• Deductive Database QSQ-Satchmo-EKS-Validity
• Parallel Logic Machine ICM PEPSys
• A rich period for ideas ..
• Declarative systems expansion
• Cosytec

6
Xerox declarative programming

• Constraint Logic Programming and Declarative
  Systems for Embedded Software
• Hybrid concurrent constraint programming model
• Grammar and Finite State machines for Linguistic
  Processing
• FST engines
• Hierarchical grammars
• Linear Logic for distributed middleware
• CLF coordination language scripting resource
  based programming - encapsulation
• Proof nets for distributed computations modeling
  (transactions)
• Aspect-Oriented Programming AspectJ
• Not logic-based about describing aspects of
  programs that cut across modules in
  object-oriented programming

7
Toward Reconfigurable Smart Systems
Control
Ca. 1990 monolithic system, central control
8
Model-based print-engine scheduling

• 85 processors
• 192 sensors
• 102 motors
• 3.5 miles wiring
• 5 M lines

• Problem
• 3-20 modules,100s of capabilities
• Customer-configurable machines
• Scheduling complex interactions

• Results
• Auto-Configuration
• Configuration-independent, optimized control
• Toolkit for reactive print engine control in real
  time

9
Hybrid cc

• Hybrid reactive (discrete) and continuous
• Reactive evolves when environment evolves-
  events
• Continuous driven e.g. by differential equations
• E.g. paper path has rollers (continuous) and
  control (discrete)
• Hybrid cc is declarative
• default cc (concurrent constraint programming)
  plus single temporal construct (hence) plus
  continuous constraints

a a holds now if a then B if a holds now, then B
holds now if a else B if a will not hold now,
then B holds now new X in A there is an instance
AT/X that holds now A, B both A and B hold
now hence A A holds at every instant after now
10
Intelligent systems Content managementThe next
frontier
90 of customer communications are through
documents. 80 of knowledge workers activities
are supported by documents

• Direct Mail
• Newsletters
• Product Marketing Brochures
• Product Specifications
• Market Attack Plans

• Letters
• Vouchers
• Coupons
• Checks
• Sales ProgramInformation
• Product Literature

• Sales Manuals
• Proposals/RFPs
• Event Invitations
• Sales Collateral
• Parts Catalogs

• Recruiting Collaterals
• Employee Benefit Communications
• Employee Education

• Statements
• Invoices
• Reporting

• Office Printing
• Commercial Print
• Publishing/Reprographics
• Forms
• Transactional Documents

• Product Drawings
• Work Orders
• Safety Data Sheets
• Change Orders
• Process Sheets
• Product Documentation

11
Intelligent systems Content managementThe next
frontier in productivity

• Content manipulation of all types of content
  enabling process management
• Linguistic, Statistical Learning technologies,
  XML enabled
• Document Classification The right document at
  the right place at the right time
• Categorization and Clustering
• Document Tagging Bringing structure to the
  document
• Authoring and Analysis
• Document Mining Discovering high value in
  textual documents
• Integration of the non-structured and structured
  domains (transactional and non-transactional)

12
Closing Comments

• Logic and declarative systems
• Major progress and still exciting future
• Specialized systems versus generic ones
• Need to work on relevant problems
• Embedded systems
• Service composition
• Document and knowledge management
• .
• THANK YOU to all those I worked with over 30
  years!