This made our usage of it about 20% faster. This is done mainly by avoiding making (shorter) copies of the buffer string when reading.