07.01.2013 Views

Lecture Notes in Computer Science 3472

Lecture Notes in Computer Science 3472

Lecture Notes in Computer Science 3472

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

5 Preorder Relations 127<br />

Def<strong>in</strong>ition 5.6. The observable test<strong>in</strong>g preorder is a relation ⊑⊆ Q × Q,<br />

where p ⊑ q iff obs(o, p) ⊆ obs(o, q) for any test o ∈O. The observable test<strong>in</strong>g<br />

equivalence is a relation �⊆ Q × Q, withp � q iff p ⊑ q and q ⊑ p. ⊓⊔<br />

If we restrict the def<strong>in</strong>ition of O (and thus the def<strong>in</strong>ition of the function obs),<br />

we obta<strong>in</strong> a different preorder, and thus a different equivalence. In other words, if<br />

we change the set of possible tests that can be applied to processes (the test<strong>in</strong>g<br />

scenario), then we obta<strong>in</strong> a different classification of processes.<br />

We will present <strong>in</strong> what follows various preorder relations under various test<strong>in</strong>g<br />

scenarios. These preorders correspond to sets of changes imposed on O and<br />

obs, and we shall keep compar<strong>in</strong>g various scenarios with the test<strong>in</strong>g scenario<br />

presented <strong>in</strong> Section 5.2.2. As it turns out, the changes we impose on O are <strong>in</strong><br />

all but one case restrictions (i.e., simplification of the possible tests).<br />

We will <strong>in</strong> most cases present an equivalent modal characterization correspond<strong>in</strong>g<br />

to these restrictions. Such a modal characterization (conta<strong>in</strong><strong>in</strong>g a set<br />

of test<strong>in</strong>g formulae and a satisfaction operator) will <strong>in</strong> essence model exactly the<br />

same th<strong>in</strong>g, but we are able to offer some results that are best shown us<strong>in</strong>g the<br />

modal characterization rather than other techniques.<br />

When we say that a preorder ⊑α makes more dist<strong>in</strong>ction than another preorder<br />

⊑β we mean that there exist processes that are dist<strong>in</strong>guishable under ⊑α<br />

but not under ⊑β. This does not imply that ⊑α and ⊑β are comparable, i.e., it<br />

could be possible that ⊑α makes more dist<strong>in</strong>ction than ⊑β and that ⊑β makes<br />

more dist<strong>in</strong>ction than ⊑α. Whenever ⊑α makes more dist<strong>in</strong>ction than ⊑β but<br />

not the other way around we say that ⊑α is coarser than ⊑β, orthat⊑β is f<strong>in</strong>er<br />

than ⊑α.<br />

5.3 Trace Preorders<br />

We thus beg<strong>in</strong> our discussion on preorder and equivalence relations with what<br />

we believe to be the simplest assumption: we compare two processes by their<br />

trace, i.e., by the sequence of actions they perform. In this section we follow<br />

roughly [vG01, dN87].<br />

We consider that the divergence predicate ↑B of the underly<strong>in</strong>g transition<br />

system is empty (no process diverges). The need for such a strong assumption<br />

will become clear later, when we discover that trace preorders do not cope well<br />

with divergence.<br />

The trace preorder is based on the follow<strong>in</strong>g test<strong>in</strong>g scenario: Weviewa<br />

process as a black box that conta<strong>in</strong>s only one <strong>in</strong>terface to the real world. This<br />

<strong>in</strong>terface is a w<strong>in</strong>dow display<strong>in</strong>g at any given moment the action that is currently<br />

carried out by the process. The process chooses its execution path autonomously,<br />

accord<strong>in</strong>g to the given transition system. As soon as no action is carried out, the<br />

display becomes empty. The observer records a sequence of actions (a trace), or<br />

a sequence of actions followed by an empty w<strong>in</strong>dow (a complete trace). Internal<br />

moves are ignored (<strong>in</strong>deed, by their def<strong>in</strong>ition they are not observable). We<br />

regard two processes as equivalent if we observe the same complete trace us<strong>in</strong>g<br />

our construction for both processes.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!