I thought we were talking about high performance parsing. Of which buffered reads are one. Other is loading entire document into mutable memory, which also has limitations.