1.10 Model 3: Message-Passing Text

1
1.10 Model 3 Message-Passing Text

• We saw
• Diagrams
• Shared-variables text
• Now
• Message-passing text (like CSP)

2

• In the communication statements,
• we name a channel,
• and not the communication partner directly.
• This naming convention
• Leads to better encapsulation of processes,
• Allows many-to-many communication through a
  single channel.

3

• Modifications to shared-variables text language
• Disallow the use of shared variables, omit
  synchronization statements based on them
• Add new communication statements based on msg
  passing.

4
Communication statements

• Channel dcl.
• Send
• Receive
• p. 71

5
Buffering capability

• Unbounded buffering (async)
• Bounded buffering (async)
• No buffering (sync)
• The usual convention in buffering keeping the
  order of sending messages.

6

• For buffered communication, we add a new variable
  ? to the set of state variables ?.
• The domain of this variable ? consists of lists
  of elements of the declared type.
• Transition relations in p. 72, 73

7

• No buffering
• Two parallel statements are considered to be
  matching if they form a send and receive pair on
  the same channel.
• When two matching statements are jointly ready to
  execute, their execution is atomic and
  simultaneous and the effect is equivalent to an
  assignment statement.
• P. 74

8

• No buffering is not the same as buffering with
  N0
• Buffering with N0 does not allow any send
  statement to be executed.
• (not the case in Promela)
• See Fig. 1.15, p. 76
• Problem 1.6

9

• Three examples for producer-consumer
• With bounded buffering
• Fig. 1.16, p.77
• With unbounded buffering
• But the channel should never hold more than N
  pending message
• Fig. 1.17, p.78
• Conditional communication statement
• An additional statement

10
Conditional communication statement

• Conditional send and receive
• P. 79
• Async communication
• Sync communication
• Example Prod-cons without buffering
• Fig. 1.19, p. 81
• Problem 1.7

11
Comparison of sync and async models

• We could implement an async problem, using sync
  constructs
• By introducing a process as a channel
• Sync communication is a more primitive concept
  than async (is it?)
• We need some ack channel, like in Fig. 1.17
• (? extended Rebeca)

12
A Fair Server

• There is a single process, server S, that is
  expected to provide services to N customer
  processes, called P1, , PN.
• The server can serve only one customer at a time.
• A fair server using shared variables, Fig. 1.20
  p.84
• A fair server using message passing, Fig. 1.21
  p.85
• It is different from 1.20

13
1.11 Model 4 Petri Nets

• ?S??Fµ??????????????