Building Graphical Choreographies from Communicating Machinesmrg.doc.ic.ac.uk/talks/2017/05/sussex/sussex-may-2017.pdf · Building Graphical Choreographies from Communicating Machines

Building Graphical Choreographiesfrom Communicating Machines

Principles and Applications

Julien Langebased on works with L. Bocchi, N. Ng, E. Tuosto, and N. Yoshida

May 2017

J. LangeBuilding Graphical Choreographies from Communicating Machines

http://mrg.doc.ic.ac.uk/

Us ∈ Mobility Research Group

www.scribble.org

Online tool : http://scribble.doc.ic.ac.uk/

Interactions with Industries

Interactions with Industries

Selected Publications 2016/2017• [ECOOP’17] Alceste Scala, Raymond Hu, Ornela Darda, NY: A Linear Decomposition of

Multiparty Sessions for Safe Distributed Programming.. • [COORDINATION’17] Keigo Imai, NY and Shoji Yuen: Session-ocaml: a session-based

library with polarities and lenses. • [FoSSaCS’17] Julien Lange , NY: On the Undecidability of Asynchronous Session

Subtyping. • [FASE’17] Raymond Hu , NY: Explicit Connection Actions in Multiparty Session Types. • [CC’17] Rumyana Neykova , NY: Let It Recover: Multiparty Protocol-Induced Recovery. • [POPL’17] Julien Lange , Nicholas Ng , Bernardo Toninho , NY: Fencing off Go: Liveness

and Safety for Channel-based Programming. • [FPL’16] Xinyu Niu , Nicholas Ng , Tomofumi Yuki , Shaojun Wang , NY, Wayne Luk :

EURECA Compilation: Automatic Optimisation of Cycle-Reconfigurable Circuits. • [ECOOP’16] Alceste Scala, NY: Lightweight Session Programming in Scala • [CC’16] Nicholas Ng, NY: Static Deadlock Detection for Concurrent Go by Global

Session Graph Synthesis. • [FASE’16] Raymond Hu, NY: Hybrid Session Verification through Endpoint API

Generation. • [TACAS’16] Julien Lange, NY: Characteristic Formulae for Session Types. • [ESOP’16] Dimitrios Kouzapas, Jorge A. Pérez, NY: On the Relative Expressiveness of

Higher-Order Session Processes. • [POPL’16] Dominic Orchard, NY: Effects as sessions, sessions as effects .

Selected Publications 2016/2017• [ECOOP’17] Alceste Scala, Raymond Hu, Ornela Darda, NY :A Linear

Decomposition of Multiparty Sessions for Safe Distributed Programming. • [COORDINATION’17] Keigo Imai, NY and Shoji Yuen: Session-ocaml: a session-

based library with polarities and lenses. • [FoSSaCS’17] Julien Lange , NY : On the Undecidability of Asynchronous Session

Subtyping. • [FASE’17] Raymond Hu , NY : Explicit Connection Actions in Multiparty Session

Types. • [CC’17] Rumyana Neykova , NY: Let It Recover: Multiparty Protocol-Induced

Recovery. • [POPL’17] Julien Lange , Nicholas Ng , Bernardo Toninho , NY: Fencing off Go:

Liveness and Safety for Channel-based Programming.• [FPL’16] Xinyu Niu , Nicholas Ng , Tomofumi Yuki , Shaojun Wang , NY, Wayne Luk:

EURECA Compilation: Automatic Optimisation of Cycle-Reconfigurable Circuits. • [ECOOP’16] Alceste Scala, NY: Lightweight Session Programming in Scala • [CC’16] Nicholas Ng, NY: Static Deadlock Detection for Concurrent Go by Global

Session Graph Synthesis. • [FASE’16] Raymond Hu, NY: Hybrid Session Verification through Endpoint API

Generation. • [TACAS’16] Julien Lange, NY: Characteristic Formulae for Session Types.• [ESOP’16] Dimitrios Kouzapas, Jorge A. Pérez, NY: On the Relative Expressiveness

of Higher-Order Session Processes. • [POPL’16] Dominic Orchard, NY: Effects as Sessions, Sessions as Effects.

Building Graphical Choreographiesfrom Communicating Machines

Principles and Applications

Julien Langebased on works with L. Bocchi, N. Ng, E. Tuosto, and N. Yoshida

May 2017


separation

alignment

cohesion

Single bird » local behaviour » CFSM

Flock » global behaviour » choreography


separation

alignment

cohesion

Single bird » local behaviour » CFSM

Flock » global behaviour » choreography


Introduction

§ Parts of distributed systems change/evolve, not always in acoordinated way,

§ these changes are often not documented.

§ Service oriented systems are sometimes composed dynamically,

§ it is often unclear how complex the overall system has become.

§ Cognizant’s Zero Deviation Lifecyle.

A global point of view of a distributed system is essential for top-levelmanagement.


Choreography

Local TypeiLocal Type1 Local Typen

ProcessiProcess1 Processn

Validate Validate Validate

Project Project Project

§ Choreography-driven development, cf. Multiparty Session Typestop-down approach (POPL’08 & ESOP’12)

§ Not applicable without a priori knowledge of a choreography

§ Our goal: from Communicating Finite-State Machines to GlobalGraphs


Choreography

Local TypeiLocal Type1 Local Typen

ProcessiProcess1 Processn

Validate Validate Validate

§ Choreography-driven development, cf. Multiparty Session Typestop-down approach (POPL’08 & ESOP’12)

§ Not applicable without a priori knowledge of a choreography§ Our goal: from Communicating Finite-State Machines to Global

Graphs


Background: CFSMs

A0

A1 A2

A3

AB!apple AB!orange

AB!apple

AB!orangeAB!drink

. . . . . .

İ

İ

. . . . . .

N

B0

B1 B2

B3

AB?appleAB?orange

AB?apple

AB?orangeAB?drink

BA!orange

A B

“On Communicating Finite-State Machines”, Brand & Zafiropulo (’83)


Background: CFSMs

A0

A1 A2

A3

AB!apple AB!orange

AB!apple

AB!orangeAB!drink

. . . . . .

İ

İ

. . . . . .

N

B0

B1 B2

B3

AB?appleAB?orange

AB?apple

AB?orangeAB?drink

BA!orange

A B

apple



Background: CFSMs

A0

A1 A2

A3

AB!apple AB!orange

AB!apple

AB!orangeAB!drink

. . . . . .

İ

İ

. . . . . .

N

B0

B1 B2

B3

AB?appleAB?orange

AB?apple

AB?orangeAB?drink

BA!orange

A B

orange apple



Background: CFSMs

A0

A1 A2

A3

AB!apple AB!orange

AB!apple

AB!orangeAB!drink

. . . . . .

İ

İ

. . . . . .

N

B0

B1 B2

B3

AB?appleAB?orange

AB?apple

AB?orangeAB?drink

BA!orange

A B

orange apple



Background: CFSMs

A0

A1 A2

A3

AB!apple AB!orange

AB!apple

AB!orangeAB!drink

. . . . . .

İ

İ

. . . . . .

N

B0

B1 B2

B3

AB?appleAB?orange

AB?apple

AB?orangeAB?drink

BA!orange

A B

orange



Background: CFSMs

A0

A1 A2

A3

AB!apple AB!orange

AB!apple

AB!orangeAB!drink

. . . . . .

İ

İ

. . . . . .

N

B0

B1 B2

B3

AB?appleAB?orange

AB?apple

AB?orangeAB?drink

BA!orange

A B

orange



Background: CFSMs

A0

A1 A2

A3

AB!apple AB!orange

AB!apple

AB!orangeAB!drink

. . . . . .

İ

İ

. . . . . .

N

B0

B1 B2

B3

AB?appleAB?orange

AB?apple

AB?orangeAB?drink

BA!orange

A B

orange orange



Background: CFSMs

A0

A1 A2

A3

AB!apple AB!orange

AB!apple

AB!orangeAB!drink

. . . . . .

İ

İ

. . . . . .

N

B0

B1 B2

B3

AB?appleAB?orange

AB?apple

AB?orangeAB?drink

BA!orange

A B

orange orange orange



Background: CFSMs

A0

A1 A2

A3

AB!apple AB!orange

AB!apple

AB!orangeAB!drink

. . . . . .

İ

İ

. . . . . .

N

B0

B1 B2

B3

AB?appleAB?orange

AB?apple

AB?orangeAB?drink

BA!orange

A B

orange orange orange orange



Background: CFSMs

A0

A1 A2

A3

AB!apple AB!orange

AB!apple

AB!orangeAB!drink

. . . . . .

İ

İ

. . . . . .

N

B0

B1 B2

B3

AB?appleAB?orange

AB?apple

AB?orangeAB?drink

BA!orange

A B

orange orange orange orange orange



Background: CFSMs

A0

A1 A2

A3

AB!apple AB!orange

AB!apple

AB!orangeAB!drink

. . . . . .

İ

İ

. . . . . .

N

B0

B1 B2

B3

AB?appleAB?orange

AB?apple

AB?orangeAB?drink

BA!orange

A B

orange orange orange orange orange orange



Global Graphs

A0 A1

A2 A3

A4Alice

AB!bwin

AC!cwin

BA?sig CA?msg

CA?msg BA?sig

AD!free

B0

B1 B2

Bob

AB?bwin

CB?blose

BC!close

BA!sig

C0

C1C2 C3

C4 C5 Carol

BC?closeCD!busy

AC?cwin

CD!busy

CA!msg

BC?close

AC?cwinCD!busy

CB!blose

CB!blose

D0 D1

Dave

AD?free

CD?busy

AÑC :cwin

CÑB :blose

AÑB :bwin

BÑC :close

CÑD :busy

BÑA :sig

CÑA :msg

AÑD : free

Four Player Game


Global Graphs

A0 A1

A2 A3

A4Alice

AB!bwin

AC!cwin

BA?sig CA?msg

CA?msg BA?sig

AD!free

B0

B1 B2

Bob

AB?bwin

CB?blose

BC!close

BA!sig

C0

C1C2 C3

C4 C5 Carol

BC?closeCD!busy

AC?cwin

CD!busy

CA!msg

BC?close

AC?cwinCD!busy

CB!blose

CB!blose

D0 D1

Dave

AD?free

CD?busy

AÑC :cwin

CÑB :blose

AÑB :bwin

BÑC :close

CÑD :busy

BÑA :sig

CÑA :msg

AÑD : free

Four Player Game


A0 A1

A2 A3

A4Alice

AB!bwin

AC!cwin

BA?sig CA?msg

CA?msg BA?sig

AD!free

B0

B1 B2

Bob

AB?bwin

CB?blose

BC!close

BA!sig

C0

C1C2 C3

C4 C5 Carol

BC?closeCD!busy

AC?cwin

CD!busy

CA!msg

BC?close

AC?cwinCD!busy

CB!blose

CB!blose

D0 D1

Dave

AD?free

CD?busy

CFSMs

AÑC :cwin

CÑB :blose

AÑB :bwin

BÑC :close

CÑD :busy

BÑA :sig

CÑA :msg

AÑD : free

Denielou & Yoshida – ESOP’12


A0 A1

A2 A3

A4Alice

AB!bwin

AC!cwin

BA?sig CA?msg

CA?msg BA?sig

AD!free

B0

B1 B2

Bob

AB?bwin

CB?blose

BC!close

BA!sig

C0

C1C2 C3

C4 C5 Carol

BC?closeCD!busy

AC?cwin

CD!busy

CA!msg

BC?close

AC?cwinCD!busy

CB!blose

CB!blose

D0 D1

Dave

AD?free

CD?busy

CFSMs

AÑC :cwin

CÑB :blose

AÑB :bwin

BÑC :close

CÑD :busy

BÑA :sig

CÑA :msg

AÑD : free

Denielou & Yoshida – ESOP’12

This work

This work


Objectives

Two main objectives:

§ Sound Condition for Safety: generalised multiparty compatibility

If S “ pM1, . . . ,Mkq is compatible then S is “safe”, i.e.,every sent message is eventually received and no deadlock.

§ Construction of a Global Graph:

If G is the global graph constructed from S, then

S “ pM1, . . . ,Mkq ” pGç1, . . . ,Gçkq


The Plan

1. Build TSpSq, the transition system of all synchronous executions2. Check for safety on TSpSq to

§ ensure equivalence between original system and the projectionsof the choreography,

§ guarantee safety (no deadlock, no orphan message)

3. Build a choreography (global graph) from TSpSq, relying on§ the theory of regions, and§ Petri nets.


1. Synchronous Transition System of CFSMs


Synchronous Transition System (TSpSq)

A0 A1

A2 A3

A4Alice

AB!bwin

AC!cwin

BA?sig CA?msg

CA?msg BA?sig

AD!free

B0

B1 B2

Bob

AB?bwin

CB?blose

BC!close

BA!sig

C0

C1C2 C3

C4 C5 Carol

BC?closeCD!busy

AC?cwin

CD!busy

CA!msg

BC?close

AC?cwinCD!busy

CB!blose

CB!blose

D0 D1

Dave

AD?free

CD?busy

Four Player Game (CFSMs)

pA0,B0,C0,D0q

pA0,B0,C2,D1q pA1,B0,C1,D0q

pA1,B0,C4,D1q

pA1,B1,C0,D0q



pA3,B2,C0,D1q

pA2,B0,C5,D1q

pA4,B0,C0,D1q

CÑD :busy

AÑC :cwinAÑB :bwin

AÑC :cwin AÑB :bwin

CÑD :busy

CÑB :blose

CÑB :blose

CÑD :busyBÑC :close

BÑC :close CÑD :busy BÑA :sig

BÑA :sig

CÑA :msgCÑD :busy

CÑA :msg

BÑA :sig

AÑD : free

Synchronous Transition System (TSpSq)


2. Check for Safety: Generalised Multiparty Compatibility (GMC)

1. Representability

2. Branching Property


Checking Compatibility (i) – Representability

Representability§ For each participant: projection of TSpSq « original machine

B0

B1 B2

Bob

AB?bwin

CB?blose

BC!close

BA!sig«

pA0,B0,C0,D0q


pA1,B0,C4,D1q

pA1,B1,C0,D0q



pA3,B2,C0,D1q

pA2,B0,C5,D1q

pA4,B0,C0,D1q

CÑD :busy



CÑD :busy

CÑB :blose

CÑB :blose



BÑA :sig

CÑA :msgCÑD :busy

CÑA :msg

BÑA :sig

AÑD : free

pAÑB :bwinqçB“AB?bwinpBÑA :sigqçB “BA!sigpCÑD :busyqçB“ε




B0

B1 B2

Bob

AB?bwin

CB?blose

BC!close

BA!sig«

pA0,B0,C0,D0q


pA1,B0,C4,D1q

pA1,B1,C0,D0q



pA3,B2,C0,D1q

pA2,B0,C5,D1q

pA4,B0,C0,D1q

CÑD :busy



CÑD :busy

CÑB :blose

CÑB :blose



BÑA :sig

CÑA :msgCÑD :busy

CÑA :msg

BÑA :sig

AÑD : free





B0

B1 B2

Bob

AB?bwin

CB?blose

BC!close

BA!sig«

pA0,B0,C0,D0q


pA1,B0,C4,D1q

pA1,B1,C0,D0q



pA3,B2,C0,D1q

pA2,B0,C5,D1q

pA4,B0,C0,D1q

CÑD :busy



CÑD :busy

CÑB :blose

CÑB :blose



BÑA :sig

CÑA :msgCÑD :busy

CÑA :msg

BÑA :sig

AÑD : free



Checking Compatibility (ii) – Branching Property

Each branchingn1

e e1 in TS must be either



Each branchingn1


§ commuting:

n1

n2 n3

n4

e e1

e1 e



Each branchingn1


§ or, each last node nk

n1 n2 nke e1 e e1 e e1

e1 ek´1

must be a “well-formed” choice, i.e.,§ for each participant

§ it receives a different message in each branch, or§ it is not involved in the choice

§ there is a unique sender

Last node,reachable

from n1, fromwhich e and e1

can be fired.


pA0,B0,C0,D0q


pA1,B0,C4,D1q

pA1,B1,C0,D0q



pA3,B2,C0,D1q

pA2,B0,C5,D1q

pA4,B0,C0,D1q

CÑD :busy



CÑD :busy

CÑB :blose

CÑB :blose



BÑA :sig

CÑA :msgCÑD :busy

CÑA :msg

BÑA :sig

AÑD : free

A : AB!bwin‰ AC!cwin B : AB?bwin‰ CB?bloseD : not involved in the choice C : AC?cwin‰ BC?close


pA0,B0,C0,D0q


pA1,B0,C4,D1q

pA1,B1,C0,D0q



pA3,B2,C0,D1q

pA2,B0,C5,D1q

pA4,B0,C0,D1q

CÑD :busy



CÑD :busy

CÑB :blose

CÑB :blose



BÑA :sig

CÑA :msgCÑD :busy

CÑA :msg

BÑA :sig

AÑD : free



pA0,B0,C0,D0q


pA1,B0,C4,D1q

pA1,B1,C0,D0q



pA3,B2,C0,D1q

pA2,B0,C5,D1q

pA4,B0,C0,D1q

CÑD :busy



CÑD :busy

CÑB :blose

CÑB :blose



BÑA :sig

CÑA :msgCÑD :busy

CÑA :msg

BÑA :sig

AÑD : free



pA0,B0,C0,D0q


pA1,B0,C4,D1q

pA1,B1,C0,D0q



pA3,B2,C0,D1q

pA2,B0,C5,D1q

pA4,B0,C0,D1q

CÑD :busy



CÑD :busy

CÑB :blose

CÑB :blose



BÑA :sig

CÑA :msgCÑD :busy

CÑA :msg

BÑA :sig

AÑD : free



pA0,B0,C0,D0q


pA1,B0,C4,D1q

pA1,B1,C0,D0q



pA3,B2,C0,D1q

pA2,B0,C5,D1q

pA4,B0,C0,D1q

CÑD :busy



CÑD :busy

CÑB :blose

CÑB :blose



BÑA :sig

CÑA :msgCÑD :busy

CÑA :msg

BÑA :sig

AÑD : free



3. Build a global graph


1: From TS to Petri Net

pA0,B0,C0,D0q


pA1,B0,C4,D1q

pA1,B1,C0,D0q



pA3,B2,C0,D1q

pA2,B0,C5,D1q

pA4,B0,C0,D1q

CÑD :busy



CÑD :busy

CÑB :blose

CÑB :blose



BÑA :sig

CÑA :msgCÑD :busy

CÑA :msg

BÑA :sig

AÑD : free

p1

AÑB :bwin

p3

BÑC :close

p5

BÑA :sig

p7

AÑC :cwin

p4

CÑB :blose

p6

CÑA :msg

p8

AÑD : free

p2

CÑD :busy

p9

We use the work of Cortadella et al. (1998), based on the theory ofregions, to derive a safe and extended free-choice Petri net from the

Synchronous Transition System.


2: From Petri Net to One-source Petri Net

p1

AÑB :bwin

p3

BÑC :close

p5

BÑA :sig

p7

AÑC :cwin

p4

CÑB :blose

p6

CÑA :msg

p8

AÑD : free

p2

CÑD :busy

p9

p1

AÑB :bwin

p3

BÑC :close

p5

BÑA :sig

p7

AÑC :cwin

p4

CÑB :blose

p6

CÑA :msg

p8

AÑD : free

p2

CÑD :busy

p9

t0

p0


3: From One-source Petri Net to Joined Petri Net

p1

AÑB :bwin

p3

BÑC :close

p5

BÑA :sig

p7

AÑC :cwin

p4

CÑB :blose

p6

CÑA :msg

p8

AÑD : free

p2

CÑD :busy

p9

t0

p0

p1

AÑB :bwin

p3

BÑC :close

p10

t2

p5

BÑA :sig

p7

AÑC :cwin

p4

CÑB :blose

p6

CÑA :msg

p8

AÑD : free

p2

CÑD :busy

p9

t1

p11

t0

p0


4 & 5 : From Joined Petri Net to Global Graph

p1

AÑB :bwin

p3

BÑC :close

p10

t2

p5

BÑA :sig

p7

AÑC :cwin

p4

CÑB :blose

p6

CÑA :msg

p8

AÑD : free

p2

CÑD :busy

p9

t1

p11

t0

p0

p1

AÑB :bwin

p3

BÑC :close

p10

t2

p5

BÑA :sig

p7

AÑC :cwin

p4

CÑB :blose

p6

CÑA :msg

p8

AÑD : free

p2

CÑD :busy

p9

t1

p11

t0

p0

AÑC :cwin

CÑB :blose

AÑB :bwin

BÑC :close

CÑD :busy

BÑA :sig

CÑA :msg

AÑD : free


Prototype Implementation

https://bitbucket.org/julien-lange/gmc-synthesis


Applications of choreography synthesis

1. Timed systems:

Meeting Deadlines Together (CONCUR 2015)Laura Bocchi, Julien Lange & Nobuko Yoshida

2. Deadlock detection in Go:

Static Deadlock Detection for Go by Global Session GraphSynthesis (CC 2016)

Nicholas Ng & Nobuko Yoshida:

3. Analysis of Erlang libraries:

Choreography-Based Analysis of Distributed MessagePassing Programs (PDP 2016)

Ramsay Taylor, Emilio Tuosto, Neil Walkinshaw, & John Derrick


Meeting Deadlines Together (CONCUR 2015)Laura Bocchi, Julien Lange & Nobuko Yoshida


From Communicating Timed Automatato Timed Choreographies

U1

U0

U2

UW!taskx ă 1x :“ 0

AU?resultx ď 20

W0 W1 W2UW?tasky “ 1

y :“ 0, y 1 :“ 0

WA!datay ă 2^ y 1 ă 10

y :“ 0

WA!stopy ă 2

A0 A1 A2

WA?dataz “ 3z :“ 0

WA?stopz “ 3z :“ 0

AU!resultz ď 5

User pUq Worker pWq Aggregator pAq

§ Each machine owns several clocks (not shared)§ Time elapses at the same pace for each clock


U (x ă 1 |x :“ 0)

task

W (y “ 1 |y :“ 0,y 1 :“ 0)

W (y ă 2^ y 1 ă 10 |y :“ 0)

data

A (z “ 3 |z :“ 0)

W (y ă 2)

stop

A (z “ 3 |z :“ 0)

A (z ď 5)

result

U (x ď 20)

§ Construction as in the untimed setting

§ Additional safety checks for time constraints


Safety checks for CTAs

§ MC: (Generalised) Multiparty Compatibility

§ IE: Interaction Enabling

§ CE: Cycle Enabling

Property S „ pTSpSqçpqpPP Safety Progress Non-Zeno Eventual ReceptionMC X X 7 7 7MC+IE X X X 7 7MC+CE X X 7 X 7MC+IE+CE X X X X X

L. Bocchi, J. Lange, and N. Yoshida. Meeting Deadlines Together.More at www.doc.ic.ac.uk/˜jlange/cta/


Static Deadlock Detection for Go by Global Session GraphSynthesis (CC 2016)

Nicholas Ng & Nobuko Yoshida:


Choreography Synthesis for Go (i)

Some Go facts:

§ Developed by Google for multi-core programming

§ Concurrency model built on CSP (process calculi)

§ Message-passing communication over channels

“Do not communicate by sharing memory; instead, sharememory by communicating” – Effective Go (developer guide)


Choreography Synthesis for Go (ii)

Go has a runtime deadlock detector, but it is very limited!


Choreography Synthesis for Go (iii)

https://www.doc.ic.ac.uk/˜cn06/pub/2016/dingo


Summing up

§ An effective way of checking properties of CFSMs, and whetherone can construct a global graph from them.

§ An algorithm based on the theory of regions.

§ A CFSMs characterisation of well-formed generalised globaltypes.

§ Applications:§ An extension for communicating timed automata§ Deadlock detection in Go§ Analysis of Erlang libraries


Thanks!

Any questions?

https://bitbucket.org/julien-lange/gmc-synthesis


Violating the no-race condition

AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit

AB : εCB : εAC : εBA : ε



AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit

AB : εCB : εAC : εBA : ε



AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit

AB : appleCB : εAC : εBA : ε



AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit

AB : appleCB : εAC : cabbageBA : ε



AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit




AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit

AB : appleCB : dateAC : εBA : ε



AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit

AB : appleCB : dateAC : εBA : ε



AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit




AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit

AB : appleCB : εAC : εBA : banana



AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit




AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit

AB : apple ¨ cakeCB : εAC : εBA : ε



AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit

AB : cakeCB : εAC : εBA : ε



AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit

AB : cakeCB : εAC : εBA : biscuit



AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit


ORPHAN MESSAGE CONFIGURATION!



AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit




AB!apple

AC!cabbage

BA?banana

AB!cake

AC!carrot

BA?banana

AB!apple

BA?biscuit

AB?apple

CB?date

BA!banana

AB?cake

CB?date

BA!banana

AB?apple

BA!biscuit

AC?carrotAC?cabbage

CB!date

A B C

AÑB :apple

AÑC :cabbage

CÑB :date

BÑA :banana

AÑB :cake

AÑC :carrot

CÑB :date

BÑA :banana

AÑB :apple

BÑA :biscuit



References

§ Julien Lange, Emilio Tuosto, and Nobuko Yoshida. “From CommunicatingMachines to Graphical Choreographies”. In: POPL. 2015

§ Laura Bocchi, Julien Lange, and Nobuko Yoshida. “Meeting DeadlinesTogether”. In: CONCUR. 2015

§ Ramsay Taylor et al. “Choreography-Based Analysis of Distributed MessagePassing Programs”. In: PDP 2016

§ Nicholas Ng and Nobuko Yoshida. “Static deadlock detection for concurrent goby global session graph synthesis”. In: CC 2016


Related work (i)

§ Pierre-Malo Denielou and Nobuko Yoshida. “Multiparty Session Types MeetCommunicating Automata”. In: ESOP. LNCS. Springer, 2012

§ Julien Lange and Emilio Tuosto. “Synthesising Choreographies from LocalSession Types”. In: CONCUR. LNCS. Springer, 2012

§ Pierre-Malo Denielou and Nobuko Yoshida. “Multiparty Compatibility inCommunicating Automata: Characterisation and Synthesis of Global SessionTypes”. In: ICALP. LNCS. 2013

§ Laura Bocchi, Weizhen Yang, and Nobuko Yoshida. “Timed Multiparty SessionTypes”. In: CONCUR 2014. 2014

§ Pavel Krcal and Wang Yi. “Communicating Timed Automata: The MoreSynchronous, the More Difficult to Verify”. In: CAV. LNCS. 2006


Related work (ii)

§ Kohei Honda, Nobuko Yoshida, and Marco Carbone. “Multiparty asynchronoussession types”. In: POPL. ACM, 2008

§ Dimitris Mostrous, Nobuko Yoshida, and Kohei Honda. “Global Principal Typingin Partially Commutative Asynchronous Sessions”. In: ESOP. LNCS. Springer,2009

§ Samik Basu, Tevfik Bultan, and Meriem Ouederni. “Synchronizability forVerification of Asynchronously Communicating Systems”. In: VMCAI. LNCS.Springer, 2012

§ Samik Basu, Tevfik Bultan, and Meriem Ouederni. “Deciding choreographyrealizability”. In: POPL. ACM, 2012


Documents

Building Graphical Choreographies from Communicating Machinesmrg.doc.ic.ac.uk/talks/2017/05/sussex/sussex-may-2017.pdf · Building Graphical Choreographies from Communicating Machines