Re:Verifying Distributed Systems with Isabelle/HOL, by Marti...
by mjb on Wednesday October 12, 2022 @05:12PM
(#33183329)
> And is that model common across other proof systems like TLA?
Yes, it's pretty common, and is a really nice abstraction.
TLA+ doesn't use that model, although a lot of TLA+ code is implemented that way. TLA+'s model is basically "shared memory" where any step can access or change any state in any way it likes. That's super flexible, but can feel a little too flexible.
P, on the other hand, is based on a model of state machines that communicate with messages. See this example for a clear illustration of how that works: https://p-org.github.io/P/tutorial/twophasecommit/
Yes, it's pretty common, and is a really nice abstraction.
TLA+ doesn't use that model, although a lot of TLA+ code is implemented that way. TLA+'s model is basically "shared memory" where any step can access or change any state in any way it likes. That's super flexible, but can feel a little too flexible.
P, on the other hand, is based on a model of state machines that communicate with messages. See this example for a clear illustration of how that works: https://p-org.github.io/P/tutorial/twophasecommit/