How do these unwind-protects stack up against
Friedman and Haynes's? The main point of difference
seems to be that Friedman and Haynes's continuations
trigger postludes based on the relative positions of
their source and target contexts on the continuation
tree; whereas Pitman's continuations trigger postludes
based on explicit user annotation, whether of the
call/cc operator, or of the continuation itself.
Which semantics is preferred may depend on the
particular programming task at hand. Fortunately,
Scheme's standard call/cc and the cob technique can
accommodate both views of unwind-protection.