Modocop in the OASIS project: Active Objects, Semantics, Internet and Security INRIA in SophiaAntipo

1
Modocop in the OASIS project Active Objects,
Semantics, Internet and SecurityINRIA in
Sophia-Antipolishttp//www.inria.fr/oasis

• Overview of the OASIS research
• Proofs of safety properties of distributed Java
  applications
• Platform for static analysis and verification of
  distributed Java

Eric Madelaine, 13-02-2002
2
Oasis at a glance
Semantics and proofs Static analysis / Model
Checking Java/XML Environment
Environments
Specifications
concurrence /distribution security
Objects
Distributed
3
Sun Microsystems
Parallel, Distributed, Concurrent,
Mobile programming

• Transparent distribution, remote object creation
• Method call -gt asynchronous communication
• Futures wait-by-necessity
• Mobility Migration of active objects
• XML descriptors for deployment

4
C3D distributed-//-collaborative
5
Ongoing Work

• Secure and efficient meta-computing
• Security at application level (appli. VPN)
• Group communication
• Formal definition of the ProActive model
• ASP object model a la Abadi-Cardelli
• Equivalence proofs between Sequential and
  Parallel programs
• Deterministic subset of the model
• Behavioural semantics, and relation with the
  object model
• Property verification via model-checking

6
Microsoft W3C
A component generator for domain-specific
languages object-oriented XML-centric

• Openness conform to W3C standards (DTD,
  Schemas)
• Reuse an homogeneous approach (exportable GUI)
• Integration component-based architecture
• Adaptability visitor design patterns AOP
• Environments for Java ( Javacard), Bytecode.

7
(No Transcript)
8
Modocop in the OASIS project

• Overview of the OASIS research
• Proofs of safety properties of distributed Java
  applications
• Platform for static analysis and verification of
  distributed Java

9
Proofs of safety properties of concurrent/distribu
ted Java applications

• Context
• Multi-threaded Java
• JCSP (processes channels a la CSP)
• ProActive (asynchronous messages, creation and
  migration of active objects)
• Specific frameworks AAA, Voyager, Aglets, ...
• Criteria
• High level primitives for distribution/commun
  ication, formal semantics
• Distributed implementation

• ProActive

10
ProActive behavioural semantics

• Active objects processes, parameterised, with
  dynamic creation.
• Distant method calls asynchronous messages,
  stored in request queues.
• Object behaviour each active object controls
  the activation of the requests in its queue.
• Return values asynchronous messages, wait by
  necessity.
• Migration is transparent.

11
ProActive finite models

• Pragmatics
• Model checkers deal with finite representations.
• Hierarchical networks of finite transition
  systems (all analysis functions ought to be
  compositional).
• Abstract and approximate finite structures (as in
  CADP or Bandera).
• Finite representation of request queues.

12
ProActive academic example

• (work done with Rabea Boulifa)
• Dining Philosophers
• Rewritten in ProActive, distributed style
• Finite configuration
• Proof of (absence of) deadlock
• Proof of boundness of requests queues
• Proof of liveness

13
ProActive academic example

• (pictures)

14
Proactive a bigger example

• (work with Tomas Barros)
• Electronic tax services in Chile
• Specification, and reference implementation in
  ProActive
• Finite instantiation already too big for brute
  force reachability analysis
• Properties (of specification)
• deadlock analysis, partial specifications
  (scenarios)
• Properties (of ref. implementation)
• equivalence with the corresponding component in
  the specification

15
Modocop in the OASIS project

• Overview of the OASIS research
• Proofs of safety properties of distributed Java
  applications
• Platform for static analysis and verification of
  distributed Java

16
Platform for verification

• Static analysis, verification and model-checking
  of Java and ProActive applications (source or
  bytecode)
• Generic tools reusable for other OO languages
• Modular external tools connected through
  standard formats or APIs, various functions
  combined for an application
• Practical source level diagnoses

17
Platform for verification

• Schema

18
Conclusion

• Work in Modocop
• Analysis platform for Java and ProActive (with
  Lande)
• Behavioural semantics (R. Boulifa)
• Connection with external verification tools and
  model-checker (E. Madelaine stage DEA)
• Realistic case study