28/Dec/2005 Linking the -calculus and B-Method Damien Karkinsky PhD Symposium IFM 2005 ¼
Distributed Systems display: 1.high degree of component autonomy, 2.instantiation & state encapsulation, 3.dynamic reconfiguration. The –calculus interaction model seen as suitable abstraction for such application domains. The B-Method offers separation of concerns when describing state components and a tool. Common semantics between the two might offer a powerful formal framework. Motivation
Differences between the two methods are: 1.actions are non-atomic, operations are atomic, 2. has operational semantics, B denotational semantics. The conventional approach in the research field is to give B operational semantics by: 1.encoding each TERM of B with a TERM of 2.proving that the encoding is sound and complete with respect to D B using bi-simulation. Background
If a common semantics exists how would the MACHINE construct behave in a system of agents? We use B denotational semantics to derive a labelled transition system LTS M. LTS M is then extended and a wrapper is provided enabling interaction with -agents. The design of the wrapper: 1.provides machine instantiation, 2.separates operations into non-atomic actions. Methodology
Methodology (cont.) In the case of machines without I/O & no guards y LTS M val1val2 P1P2 zop h ¤ i op INTER ( ¿ ) + LTS M ¼ z LTS M ………
Methodology (cont.) val1 ¼ P1 zop h ¤ i op INTER LTS M ¿ ? val1 ¼ P1 LTS M
Example on o ® o ® on LTS M
Example (cont.)
Using the wrapper The process can be used in a system of -agents. [[( BEGIN )]]h z i
A problem ( # ¿ ) + ( # ¿ ) + # cp ( z ) b = 2 f n ( C 0 )
Future Work We are interested in investigating controlled interactions between -agents and machine instances. Verification of divergence freedom properties 1.The full -calculus is too open, 2.Machine communication internalised, 3.Bi-simulation does not distinguish divergence.