| Summary: | The main observational equivalences of the untyped lambda-calculus have been
characterized in terms of extensional equalities between B\"ohm trees. It is
well known that the lambda-theory H*, arising by taking as observables the head
normal forms, equates two lambda-terms whenever their B\"ohm trees are equal up
to countably many possibly infinite eta-expansions. Similarly, two lambda-terms
are equal in Morris's original observational theory H+, generated by
considering as observable the beta-normal forms, whenever their B\"ohm trees
are equal up to countably many finite eta-expansions.
The lambda-calculus also possesses a strong notion of extensionality called
"the omega-rule", which has been the subject of many investigations. It is a
longstanding open problem whether the equivalence B-omega obtained by closing
the theory of B\"ohm trees under the omega-rule is strictly included in H+, as
conjectured by Sall\'e in the seventies. In this paper we demonstrate that the
two aforementioned theories actually coincide, thus disproving Sall\'e's
conjecture.
The proof technique we develop for proving the latter inclusion is general
enough to provide as a byproduct a new characterization, based on bounded
eta-expansions, of the least extensional equality between B\"ohm trees.
Together, these results provide a taxonomy of the different degrees of
extensionality in the theory of B\"ohm trees.
|
|---|