No. There is the original bytestorage for the data that is to be parsed - and is the only thing necessarily taking a chunk of memory - and the sink for the parsed data somewhere else. The state held by the enumerator itself is tiny. A pointer to the source buffer is held in the pos pointer. As the user extracts entities from the parser the pos pointer is just forwarded. There is of course the parsing code inside 'next' itself which was not included, but which knows what entities to expect and forwards the pos by the entity size while checking it does not go over the end.

i.e. enumerating_parser p(source, source_len); ParseResult r; while(p.next(r)){// forward context of r forwards}.