The primary assumption of whatever's in the first page is that types are sets of values right? Integrated shrinking assumes the rules that form the set while ty

The primary assumption of whatever's in the first page is that types are sets of values right? Integrated shrinking assumes the rules that form the set while type based shrinking assumes subsetting. Am I right in reading this? If so.. why are we still using the notion that types = sets of values?