Datalog does not need/do unification for rule evaluation, as it is just matching variables to values in a single direction. Body literals are matched against the database and the substitutions are applied to the rest of the rule and the head.
Prolog does unification of the proof goal with the rule head. It's necessary there but not with datalog.
While bottom-up evaluation is the norm in datalog, it is not a requirement and there are datalog engines that evaluate top-down or all-at-once.
But I still agree with you about the capitalization. Some formats, like KIF, use a '?' prefix instead, and I've seen some HRF notations that mix the '?' prefix with non-KIF formatting (':-' operator and '.' terminator).