Are two CCS processes equivalent with respect to weak bisimilarity if and only if they satisfy exactly the same set of HML formulas?

Question

I was skimming this recent paper and I was struck by the following statement:

two processes are equivalent with respect to weak bisimilarity if and only if they satisfy exactly the same set of HML formulas

I have trouble believing it.

I reason that:

weak bisimilarity does not account for tau-transitions and therefore is oblivious to the possibility of livelocks.
I believe the possibility of a livelock can be represented in HML,

and, therefore, a counterexample can be constructed.

Consider these processes:

One = lol.One;
A = lol.omg.A + zomg.A;
B = 'omg.B + 'zomg.B;
Two = (A | B) \ {omg, zomg};

Consider, furthermore, the HML formula

Livelock = max(X. <tau>X);

I believe that this formula is satisfied only by Two, which nevertheless is weakly bisimiliar to One.

I have tried to use the Edinburgh Concurrency Workbench to confirm my reasoning:

agent One = lol.One;
agent A = lol.omg.A + zomg.A;
agent B = 'omg.B + 'zomg.B;
agent Two = (A | B) \ {omg, zomg};
prop Livelock = max(X. <tau>X);
****** 
echo "Is One strongly bisimilar to Two?";
strongeq(One, Two); ********************* outputs "false"
echo "Is One weakly bisimilar to Two?";
eq(One, Two);       ********************* outputs "true"
echo "Is it the case that One |= Livelock?";
checkprop(One,Livelock);  *************** outputs "false"
echo "Is it the case Two |= Livelock?";
checkprop(Two,Livelock);  *************** outputs "true"

So, is it the case that two processes are equivalent with respect to weak bisimilarity if and only if they satisfy exactly the same set of HML formulas?

If so, why is this true, and what's the flaw in the counterexample, above?
If not, what did the authors mean or what sort of context am I missing? (Or was it a typo?)

score 1 · Accepted Answer · answered Apr 04 '19 at 13:21

(This probably does not fully answer, but I think it can still help)

According to their Def. 4, there is no modality $\langle \tau \rangle$ in their HML logic. Their syntax only allows $\langle\!\langle u \rangle\!\rangle$ where $u \in A \cup\{\epsilon\}$, and $\tau$ is assumed not to be in $A$ (Def. 1), hence $u\neq\tau$.

Also, there seems to be no recursion allowed in their logic.

Further, as far as I can see, in the Concurrency Workbench <a> indicates a strong possibility, while <<a>> indicates a weak possibility.

I can't recall which variant of the HML logic precisely corresponds to weak bisimilarity, but I think that it can't involve a strong <tau> modality since otherwise we would distinguish lol and tau.lol.

Are two CCS processes equivalent with respect to weak bisimilarity if and only if they satisfy exactly the same set of HML formulas?

1 Answers1