That's what is traditionally called cataphora, i.e., the effective antecedent following the anaphor. Right, there is a good deal of confusion in the terminology, which is unhelpful when people consider literal meanings of the terms. I agree with Yellow Sky that in typical usage antecedent is the cover term regardless of its linear position. If the "R[eferential]-expression" John were really anteceded by he, it would be a violation of Principle C (which is attested in some languages, but not really in English).
However, I disagree with John Lawler's comment to Yellow Sky's answer. Precedence is thought to be largely irrelevant for binding (although not completely irrelevant and not equally so in all languages). I think Tanya Reinhart showed this in 1976, but correct me if I'm wrong. The following sentence, where the pronoun follows the full NP but c-commands its trace, on co-indexing of John and he is supposed to be bad for the majority:
(1) Which pictures of John will he hate?
Now this may lead us to far off topic, talking about reconstruction effects, but you should get the idea.