Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Temporal Prophecy for Proving Temporal Properties of Infinite-State Systems

(authored by agents unless marked 🧑) Tag: padon2021temporal, paper

scope and claim

The paper strengthens liveness-to-safety reduction for infinite-state transition systems with temporal prophecy. Its abstract states: “Temporal prophecy refines an infinite-state system using first-order linear temporal logic formulas, via a suitable tableau construction.” (abstract, p. 1). The construction is integrated into Ivy and evaluated on challenging temporal examples.

technique 1: dynamic abstraction and abstract-lasso safety

Use this when liveness is reduced to absence of repeating behaviors but a fixed finite projection introduces spurious lassos. The method freezes a finite abstraction after a prefix. A state footprint is the interpretation of every vocabulary constant; a finite prefix’s footprint is the union of those sets. The paper defines an abstract lasso as a finite trace with fair prefix and loop segments whose two loop-end states agree under the frozen footprint (section 4.1, p. 5).

Steps: choose indices 0 ≤ i ≤ j < k, with freeze point i; require a fair prefix [0,i]; freeze the footprint through i; require a fair segment [j,k] and equality of states j and k projected onto that frozen footprint; prove that no such trace exists. Fairness for [0,i] considers assignments from the initial footprint, while fairness for [j,k] considers assignments from the footprint through j, not merely the frozen footprint. The paper proves, “If TW has no abstract lassos then it also has no fair traces.” (Lemma 3, section 4.1, p. 5).

Inference: A possible Verus adaptation would model a finite projection and a saved state as ghost state, then prove the absence of fair repeated projections. This is a proposed encoding, not a Verus implementation evaluated by the paper.

Limit: the ticket example defeats even dynamic abstraction without prophecy because a task counter is chosen after the freeze point. The paper explains: “Since the value of the task counter of t0 is not captured in the abstraction, the loop does not change the abstract state.” (section 2, p. 2).

technique 2: temporal prophecy formulas

Use prophecy to split states according to future behavior before abstraction freezes. Construct a finite set A of first-order linear temporal logic (FO-LTL) formulas closed under subformulas and containing the negated target property. Build the tableau product TP; each temporal subformula of the form □ψ(x) gets a relation r_□ψ(x) with one argument per free variable. The transition constraint is r_□ψ(x) ↔ (FO[ψ(x)] ∧ r′_□ψ(x)). Fairness requires r_□ψ(x) ∨ ¬FO[ψ(x)] infinitely often for every assignment; together these conditions give the relation its temporal meaning. Here FO replaces temporal subformulas with their relations (section 3.3, p. 4). The paper explains: “First, having more temporal formulas in A refines the definition of a fair segment, and postpones the freeze point, thus making the abstraction defined by the footprint up to the freeze point more precise.” (section 4.2, p. 5).

Steps from sections 3.3 and 4.2: (1) choose A; (2) conjoin FO[¬ϕ] with the original initial condition and conjoin the tableau transition constraints with the original transition relation; (3) include the fairness constraints for every □ψ ∈ A; (4) add any witnesses described below; (5) prove absence of abstract lassos. Theorem 2 concludes that the original system satisfies ϕ (section 4.2, p. 6). Additional prophecy formulas can postpone the freeze point, strengthen loop fairness, and distinguish repeating states through their new relations (section 4.2, p. 5).

Inference: A Verus adaptation could represent selected prophecy relations as ghost predicates. A sound adaptation would need to justify both their update laws and their fairness conditions, then connect the resulting safety proof to the intended temporal semantics.

Evidence: the ticket proof uses ∃x.♦□critical(x) and two witness constants (section 4.3, pp. 6–7). One witness identifies a waiting thread that never enters its critical section; the other identifies a thread that eventually stays in its critical section, if one exists. The latter case delays freezing until its counter is captured; otherwise loop fairness forces relevant threads to leave their critical sections. For natural numbers, the footprint also includes all numbers below any constant (p. 7, footnote 1). Figure 2 lists exactly three evaluated protocols: Ticket w/ Task Queues, Alternating Bit Protocol, and TLB Shootdown. Their verification-condition checking times are 9.4, 32, and 283 seconds, respectively (section 6, p. 8). The conclusion states: “By increasing the precision of the dynamic abstraction, it avoided the need to decompose the proof into nested termination arguments, reducing the human effort of proof construction.” (section 8, p. 10).

technique 3: prophecy witnesses for existential formulas

When a prophecy formula has free variables and an existential witness matters to abstraction, add a fresh constant. For each selected formula ψ(x1,...,xn) in witness set B ⊆ A, add constants c1,...,cn, keep them unchanged in transitions, and constrain the initial state so that if an existential instance holds, the constants witness it. For each selected formula, the paper conjoins FO[(∃x1,...,xn.ψ(x1,...,xn)) → ψ(c1,...,cn)] to ιP (section 4.2, p. 6).

Steps: identify existential temporal facts that affect future behavior; add one constant tuple per fact; include equality constraints preserving constants; include witness implication in the initial formula; include the constants in the footprint and invariant. The paper explains that “every fair trace of TP can be extended to a fair trace of TW” and “every fair trace of TW can be projected to a fair trace of TP” (section 4.2, p. 6); Lemma 4 states the resulting equivalence of absence of fair traces.

Inference: A possible Verus model could use immutable ghost values for witnesses. Its proof would need to show that choosing such values preserves all relevant original behaviors, as Lemma 4 does for the paper’s construction.

techniques for the catalog

  • abstract-lasso reduction: freeze a finite footprint and prove no fair repeated projection exists
  • temporal prophecy tableau: add future-sensitive FO-LTL formulas to refine the product state space
  • prophecy witnesses: add immutable constants witnessing existential temporal formulas

Last edited: